| · APPLIED COMPUTE · WORKBOOK ·

Excel model, rendered

The supplied configurator workbook, shown as it is saved: 4 sheets, 1,680 cells, 377 formulas. The equations are written out below, and a parity table checks every output the tool reproduces against the workbook for the default deal.

Download the workbook (.xlsx) Open the live tool

How the workbook is laid out

  1. 01 Deal Setup. The inputs. Shaded cells are the only things to change: customer, model, cycles, throughput, costs, price, margin and the three approval gates.
  2. 02 Economics Capacity. All calculations. Rows 4–28 turn tokens into GPU-hours; rows 31–47 are the week-by-week capacity grid; rows 52–70 compare buying options; rows 73–85 are cost; rows 91–104 are price, margin and the status gate; rows 111–125 are optional production.
  3. 03 Offer and SOW. The customer-facing offer and SOW wording that the tool reproduces in its proposal and SOW documents.
  4. 04 Guide Assumptions. The guide and the assumption sources. The tool shows each input’s meaning and its workbook cell.

The equations, in order

Read top to bottom: each step uses the one before it. The formula column is the exact text saved in the workbook; the value column is what the workbook shows for the default deal.

Step and meaningCellWorkbook formulaSaved value
Training GPU-hours
Tokens the policy update needs, divided by what one GPU does per hour (throughput × proxy factor).
02 Economics Capacity!B18 =IF('01 Deal Setup'!B43<=0,0,B13/('01 Deal Setup'!B43*B6)) 1,280
Total workload GPU-hours
Policy, rollout and evaluation GPU-hours added together across all approved cycles.
02 Economics Capacity!B21 =SUM(B18:B20) 3,360
Wall-clock hours per cycle
Concurrent jobs overlap, so the slower of trainer or rollout sets the pace. Evaluation is always added on top.
02 Economics Capacity!B27 =IF('01 Deal Setup'!B14="Concurrent",MAX(B22/B7,B23/B8),B22/B7+B23/B8)+B24/B9 143.3333
Total programme weeks
Readiness, plus one cadence block per cycle, plus handoff.
02 Economics Capacity!B28 ='01 Deal Setup'!B12+'01 Deal Setup'!B10*'01 Deal Setup'!B11+'01 Deal Setup'!B13 6
Cheapest way to buy missing GPUs
The lowest of the three quotes: pay-as-you-go, a one-week extension, or a new reservation.
02 Economics Capacity!B61 =MIN(B58:B60) 0
Base delivery cost
Engineering, graders, storage, platform and GPU compute, summed before rework.
02 Economics Capacity!B82 =SUM(B77:B81) 96,900
Delivery COGS including rework
Base cost plus the rework allowance (a percentage of base cost). This is the internal cost the price must cover.
02 Economics Capacity!B84 =SUM(B82:B83) 106,590
RL cycle fee
Cycles approved × fee per cycle × size-class factor (0.65× small, 1× standard, 2.5× large).
02 Economics Capacity!B93 =B10*'01 Deal Setup'!B70*IF('01 Deal Setup'!$G$31=1,0.65,IF('01 Deal Setup'!$G$31=2,1,2.5)) 72,000
Margin-floor price
COGS ÷ (1 − target margin). If the itemised price is below this, the workbook adds an adjustment (B99); the tool folds it into the phase fees.
02 Economics Capacity!B98 =B84/MAX(0.01,1-'01 Deal Setup'!B76) 193,800
Quote
Itemised price plus any margin-floor adjustment. This is the number the customer sees, split by phase.
02 Economics Capacity!B100 =SUM(B97,B99) 210,000
Gross margin
1 − COGS ÷ quote.
02 Economics Capacity!B101 =IF(B100=0,0,1-B84/B100) 0.4924
Status gate
HOLD if an input is missing, the model is unverified, capacity is infeasible, the budget is exceeded or margin is under the floor. READY FOR REVIEW only when every check passes.
02 Economics Capacity!B104 =IF(OR('01 Deal Setup'!B43<=0,'01 Deal Setup'!B44<=0,'01 Deal Setup'!B45<=0,'01 Deal Setup'!B98<=0,'01 Deal Setup'!B10<1,'01 Deal Setup'!B10>10,'01 Deal Setup'!B35<0,'01 Deal Setup'!B36<0,'01 Deal Setup'!B37<0,'01 Deal Setup'!B38<0,'01 Deal Setup'!B40<0,'01 Deal Setup'!B82<0,'01 Deal Setup'!B84<0),"HOLD - INPUT",IF('01 Deal Setup'!G30<>"QUALIFIED*","HOLD - MODEL",IF(B70<>"FEASIBLE*","HOLD - CAPACITY",IF(B103<0,"HOLD - BUDGET",IF(B101<'01 Deal Setup'!B77,"HOLD - MARGIN","READY FOR REVIEW"))))) HOLD - MODEL

Parity: the tool against the workbook

For the default deal, 95 mapped outputs are compared cell by cell: 95 match exactly, 0 documented deviations, 0 mismatch. The full 17-scenario check, recalculated in LibreOffice, is in PARITY.md.

OutputCellWorkbookToolStatus
Proxy throughput factor (benchmark required)02!B6 11 match
Planning trainer GPUs02!B7 88 match
Planning rollout GPUs02!B8 88 match
Planning eval GPUs02!B9 44 match
Policy update GPU-hours02!B18 1,2801,280 match
Rollout GPU-hours02!B19 1,9201,920 match
Evaluation GPU-hours02!B20 160160 match
Total workload GPU-hours02!B21 3,3603,360 match
Trainer GPU-hours per cycle02!B22 426.6667426.6667 match
Rollout GPU-hours per cycle02!B23 640640 match
Evaluation GPU-hours per cycle02!B24 4040 match
Baseline evaluation GPU-hours02!B25 4040 match
Peak GPU units in RL02!B26 88 match
RL activity wall-hours / cycle02!B27 143.3333143.3333 match
Total planned weeks02!B28 66 match
Weeks needing new GPUs02!B52 00 match
First shortage week02!B53 00 match
Last shortage week02!B54 00 match
Peak additional GPU units02!B55 00 match
New workload GPU-hours02!B56 00 match
Existing assigned GPU-hours02!B57 3,3603,360 match
PayGo acquisition cost02!B58 00 match
Weekly extension acquisition cost02!B59 00 match
New reservation committed cost02!B60 00 match
Cheapest incremental cash02!B61 00 match
Selected purchase method02!B62 ExistingExisting match
Cash cost of new capacity02!B63 00 match
New GPU-hours paid / committed02!B64 00 match
Unused new commitment hours02!B65 00 match
Managerial allocation of existing02!B66 8,4008,400 match
Existing alternative-use value02!B68 00 match
Economic decision cost (cash + opportunity)02!B69 00 match
Capacity / schedule verdict02!B70 FEASIBLE*FEASIBLE* match
Forward deployed engineering02!B73 28,80028,800 match
ML research engineering02!B74 32,00032,000 match
Infrastructure engineering02!B75 10,20010,200 match
Delivery coordination02!B76 6,0006,000 match
Total attributable engineering02!B77 77,00077,000 match
Graders, sandboxes and external APIs02!B78 6,0006,000 match
Storage and networking02!B79 1,5001,500 match
Direct platform support02!B80 4,0004,000 match
GPU COGS (allocated + new cash)02!B81 8,4008,400 match
Base cost to deliver02!B82 96,90096,900 match
Expected rework / failure allowance02!B83 9,6909,690 match
Risk-adjusted expected COGS02!B84 106,590106,590 match
COGS + opportunity view02!B85 106,590106,590 match
Readiness / baseline02!B91 30,00030,000 match
Engineering and integration02!B92 78,00078,000 match
Authorized RL cycles02!B93 72,00072,000 match
Extra-cycle engineering02!B94 00 match
Independent acceptance / handoff02!B95 30,00030,000 match
Negotiated discount02!B96 0-0 match
Base customer proposal02!B97 210,000210,000 match
Target-margin price floor02!B98 193,800193,800 match
Transparent capacity / risk adjustment02!B99 00 match
SUGGESTED CUSTOMER QUOTE02!B100 210,000210,000 match
Expected deal gross margin02!B101 0.49240.4924 match
Budget ceiling02!B102 260,000260,000 match
Remaining approved budget02!B103 50,00050,000 match
Commercial approval status02!B104 HOLD - MODELHOLD - MODEL match
Input + output equivalent tokens / month02!B111 20,000,00020,000,000 match
Variable serving demand GPU-hours02!B112 200200 match
Warm capacity floor GPU-hours02!B113 1,4401,440 match
Provisioned / allocated serving GPU-hours02!B114 1,4401,440 match
Serving GPU vendor COGS02!B115 5,745.65,745.6 match
Expected monthly serving COGS02!B116 8,245.68,245.6 match
Suggested monthly serving invoice02!B117 00 match
Pre-adjustment serving gross margin02!B118 00 match
Serving size / SLO validation02!B119 NOT SCOPEDNOT SCOPED match
Peak vs current warm capacity02!B120 0.41670.4167 match
Scaling / re-paper recommendation02!B121 NOT SCOPEDNOT SCOPED match
Target-margin serving floor02!B122 00 match
Visible production uplift02!B123 00 match
SUGGESTED MONTHLY SERVING PRICE02!B124 00 match
Expected production gross margin02!B125 00 match
Total model parameters (B)01!B8 3535 match
Customer01!G5 210,000210,000 match
Workflow / task01!G6 106,590106,590 match
Selected open-weight model01!G7 0.450.45 match
Total model parameters (B)01!G8 0.49240.4924 match
Training method01!G9 3,3603,360 match
Approved maximum RL cycles01!G10 ExistingExisting match
Cadence01!G11 00 match
Readiness period01!G12 00 match
Handoff period01!G13 50,00050,000 match
Baseline score01!G16 FEASIBLE*FEASIBLE* match
Target acceptance score01!G17 HOLD - MODELHOLD - MODEL match
Minimum meaningful gain01!G18 NO RESULTS YETNO RESULTS YET match
Consecutive slow cycles01!G19 NOT IN SCOPENOT IN SCOPE match
Include production proposal?01!G23 33 match
Workbook VLOOKUP returns 0 for blank active parameters; brief requires "Not verified", never 0.
Serving contract type01!G24 MoEMoE match
Monthly input tokens01!G25 ListedListed match
Minimum warm replicas01!G28 ValidateValidate match
Customer accountable owner01!G30 HOLD - BENCHMARKHOLD - BENCHMARK match
Hosting region01!G32 PROVISIONAL ESTIMATEPROVISIONAL ESTIMATE match

Every sheet, cell by cell

Full contents as saved, for checking any single cell. Labels come from the row’s text.

01 Deal Setup · 491 cells, 41 formulas
CellLabelSaved valueFormula
A1DEAL CONFIGURATOR
A2EDIT shaded inputs; selected model determines planning proxy, but do not issue quote until model/hardware/throughput gates clear.
A41 CUSTOMER, SCOPE & APPROVAL
F41 CUSTOMER, SCOPE & APPROVALLIVE DEAL OUTPUTS (AUTO-CALCULATED)
A5Customer
B5CustomerExample Enterprise
C5Customertext
D5CustomerReplace with customer legal/entity name.
F5CustomerRecommended quote
G5Customer210,000='02 Economics Capacity'!B100
A6Workflow / task
B6Workflow / taskInternal workflow automation
C6Workflow / tasktext
D6Workflow / taskOne defined use case; material changes require amendment.
F6Workflow / taskForecast delivery COGS
G6Workflow / task106,590='02 Economics Capacity'!B84
A7Selected open-weight model
B7Selected open-weight modelQwen3.6-35B-A3B
C7Selected open-weight modelmodel dropdown
D7Selected open-weight modelNamed options from the supplied AC2 knowledge base. Training/serving eligibility may differ.
F7Selected open-weight modelTarget margin
G7Selected open-weight model0.45=B76
A8Total model parameters (B)
B8Total model parameters (B)35=IFERROR(IF(VLOOKUP(B7,'04 Guide Assumptions'!$F$26:$Q$55,2,FALSE)=0,"Not verified",VLOOKUP(B7,'04 Guide Assumptions'!$F$26:$Q$55,2,FALSE)),"Not verified")
C8Total model parameters (B)billion params
D8Total model parameters (B)Automatically populated; total model weights, not compute cost or active parameters.
F8Total model parameters (B)Expected deal margin
G8Total model parameters (B)0.4924='02 Economics Capacity'!B101
A9Training method
B9Training methodGRPO
C9Training method▼ select
D9Training methodOnly GRPO uses current RL-loop cost logic; other methods trigger HOLD until re-costed.
F9Training methodDemand GPU-hours
G9Training method3,360='02 Economics Capacity'!B21
A10Approved maximum RL cycles
B10Approved maximum RL cycles3
C10Approved maximum RL cyclescycles
D10Approved maximum RL cyclesContracted maximum; no unapproved experimentation.
F10Approved maximum RL cyclesSelected capacity method
G10Approved maximum RL cyclesExisting='02 Economics Capacity'!B62
A11Cadence
B11Cadence1
C11Cadenceweeks per cycle
D11CadenceOne cycle every week by default; one cycle forecast in a week.
F11CadenceNew GPUs at peak
G11Cadence0='02 Economics Capacity'!B55
A12Readiness period
B12Readiness period2
C12Readiness periodweeks
D12Readiness periodTime to baseline, prepare evals, approve grader.
F12Readiness periodProcurement commitment
G12Readiness period0='02 Economics Capacity'!B63
A13Handoff period
B13Handoff period1
C13Handoff periodweeks
D13Handoff periodFinal independent eval, checkpoint handoff and signoff.
F13Handoff periodCustomer budget headroom
G13Handoff period50,000='02 Economics Capacity'!B103
A14RL jobs scheduled
B14RL jobs scheduledSequential
C14RL jobs scheduledSequential / Concurrent
D14RL jobs scheduledSequential avoids overlap capacity; concurrent requires both fleets at once.
A15Success metric
B15Success metricHeld-out pass rate
C15Success metric▼ select
D15Success metricOnly 0–1 scores with higher = better. Customer approves frozen eval and target.
F15Success metricDEAL READINESS / DECISION SIGNALS
A16Baseline score
B16Baseline score0.6
C16Baseline score%
D16Baseline scoreIllustrative pre-training score; use reproducible baseline.
F16Baseline scoreSchedule / availability
G16Baseline scoreFEASIBLE*='02 Economics Capacity'!B70
A17Target acceptance score
B17Target acceptance score0.8
C17Target acceptance score%
D17Target acceptance scoreTarget for recommendation to stop, not guaranteed research outcome.
F17Target acceptance scorePrice / budget
G17Target acceptance scoreHOLD - MODEL='02 Economics Capacity'!B104
A18Minimum meaningful gain
B18Minimum meaningful gain0.01
C18Minimum meaningful gainabsolute points
D18Minimum meaningful gainPlateau threshold per cycle (1 pp = 0.01).
F18Minimum meaningful gainSuggested stop action
G18Minimum meaningful gainNO RESULTS YET=IF(B21="Stop","STOP - CUSTOMER DECISION",IF(B20=0,"NO RESULTS YET",IF(INDEX(G37:G44,B20)>=B17,"REVIEW - TARGET REACHED",IF(B20>=B10,"REVIEW - CYCLE CAP",IF(AND(B20>=B19,INDEX(H37:H44,B20)<B18,INDEX(H37:H44,B20-1)<B18),"REVIEW - PLATEAU","CONTINUE; CUSTOMER APPROVAL")))))
A19Consecutive slow cycles
B19Consecutive slow cycles2
C19Consecutive slow cyclescycles
D19Consecutive slow cyclesReview when 2 successive gains fall below threshold.
F19Consecutive slow cyclesProduction quote (if Yes)
G19Consecutive slow cyclesNOT IN SCOPE=IF(B23="Yes",'02 Economics Capacity'!B124,"NOT IN SCOPE")
A20Cycles with scores entered
B20Cycles with scores entered0=COUNT(G37:G44)
C20Cycles with scores enteredcompleted
D20Cycles with scores enteredScore log right -> automatically counts entered outcomes.
A21Customer continuation decision
B21Customer continuation decisionPending
C21Customer continuation decisionPending / Continue / Stop
D21Customer continuation decisionCustomer is decision maker after each cycle; forecast is advisory.
A22Maximum deal approval budget
B22Maximum deal approval budget260,000
C22Maximum deal approval budget$
D22Maximum deal approval budgetSignals when quote needs scope/budget change.
F22Maximum deal approval budgetMODEL QUALIFICATION | REQUIRED BEFORE QUOTE
A23Include production proposal?
B23Include production proposal?No
C23Include production proposal?Yes / No
D23Include production proposal?Separate prospective monthly quote, not part of training value.
F23Include production proposal?Active parameters (B)
G23Include production proposal?3=IFERROR(VLOOKUP($B$7,'04 Guide Assumptions'!$F$26:$Q$55,3,FALSE),"Not verified")
A24Serving contract type
B24Serving contract typeDedicated
C24Serving contract typeShared / Dedicated
D24Serving contract typeShared usage minimum vs dedicated warm-replica subscription.
F24Serving contract typeArchitecture
G24Serving contract typeMoE=IFERROR(VLOOKUP($B$7,'04 Guide Assumptions'!$F$26:$Q$55,4,FALSE),"Not verified")
A25Monthly input tokens
B25Monthly input tokens20,000,000
C25Monthly input tokenstokens/month
D25Monthly input tokensForecast only; contract includes named allowances.
F25Monthly input tokensRL training status
G25Monthly input tokensListed=IFERROR(VLOOKUP($B$7,'04 Guide Assumptions'!$F$26:$Q$55,5,FALSE),"Not verified")
A26Monthly output tokens
B26Monthly output tokens10,000,000
C26Monthly output tokenstokens/month
D26Monthly output tokensOutputs usually have higher serving cost.
F26Monthly output tokensServing status
G26Monthly output tokensListed=IFERROR(VLOOKUP($B$7,'04 Guide Assumptions'!$F$26:$Q$55,6,FALSE),"Not verified")
A27Peak / average traffic
B27Peak / average traffic3
C27Peak / average trafficx
D27Peak / average trafficCapacity planning proxy; benchmark peak latency before launch.
F27Peak / average trafficCatalog evidence
G27Peak / average trafficModel-named=IFERROR(VLOOKUP($B$7,'04 Guide Assumptions'!$F$26:$Q$55,7,FALSE),"Not verified")
A28Minimum warm replicas
B28Minimum warm replicas2
C28Minimum warm replicasGPUs
D28Minimum warm replicasLatency / availability floor independent of token volume.
F28Minimum warm replicasGPU class to confirm
G28Minimum warm replicasValidate=IFERROR(VLOOKUP($B$7,'04 Guide Assumptions'!$F$26:$Q$55,9,FALSE),"Not verified")
A29Availability target
B29Availability target0.999
C29Availability target%
D29Availability targetPlanning assumption, not an SLA until service order signed.
F29Availability targetParameter source URL
G29Availability targethttps://huggingface.co/Qwen/Qwen3.6-35B-A3B=IFERROR(VLOOKUP($B$7,'04 Guide Assumptions'!$F$26:$Q$55,10,FALSE),"Not verified")
A30Customer accountable owner
B30Customer accountable ownerCustomer product / SME lead
C30Customer accountable ownerowner
D30Customer accountable ownerOwns base model, grader, continue/stop and acceptance decisions.
F30Customer accountable ownerModel readiness
G30Customer accountable ownerHOLD - BENCHMARK=IF($B$9<>"GRPO","HOLD - METHOD RE-SCOPE",IF(OR($B$33="Access pending",$B$33="Not available"),"HOLD - DATA ACCESS",IF(NOT(ISNUMBER($B$8)),"HOLD - PARAMETERS",IF($B$101<>"Yes","HOLD - BENCHMARK",IF($B$102<>"Yes","HOLD - ACCESS",IF($B$104<>"Yes","HOLD - SKU RATE",IF(AND($G$28="B300",$B$103<>"B300"),"HOLD - GPU FAMILY",IF(OR($G$25="Not listed",$G$25="Research only",$G$25="Legacy mention",$G$25="Family - verify",$G$25="Access required"),"HOLD - TRAINING ELIGIBILITY","QUALIFIED*"))))))))
A31Applied Compute delivery owner
B31Applied Compute delivery ownerAC engagement lead
C31Applied Compute delivery ownerowner
D31Applied Compute delivery ownerBuilds, runs, measures, and recommends.
F31Applied Compute delivery ownerPlanning proxy class
G31Applied Compute delivery owner2=IFERROR(VLOOKUP($B$7,'04 Guide Assumptions'!$F$26:$Q$55,8,FALSE),"Not verified")
A32Hosting region
B32Hosting regionUS region TBD
C32Hosting regionregion
D32Hosting regionData residency and supply feasibility must be confirmed.
F32Hosting regionModel suitability
G32Hosting regionPROVISIONAL ESTIMATE=IF(OR($G$25="Not listed",$G$25="Research only",$G$25="Legacy mention"),"TRAINING NOT CONFIRMED",IF($G$25="Family - verify","VARIANT ACCESS CHECK",IF($B$8="Not verified","PARAMETERS NEED VERIFICATION",IF($G$28="Multi-node","MULTI-NODE REVIEW","PROVISIONAL ESTIMATE"))))
A33Data/access readiness
B33Data/access readinessCustomer supplies by kickoff
C33Data/access readiness▼ select
D33Data/access readinessReady / validated or by kickoff are allowed; pending/unavailable block approval.
F33Data/access readinessOPTIONAL: ENTER ACTUAL HELD-OUT EVALS EACH CYCLE
A342 DEMAND + GPU PERFORMANCE ASSUMPTIONS
A35Policy-update tokens per cycle
B35Policy-update tokens per cycle42,666,666.6667
C35Policy-update tokens per cycletokens/cycle
D35Policy-update tokens per cycleTraining tokens; benchmark replaces heuristic.
A36Rollout attempts per cycle
B36Rollout attempts per cycle4,000
C36Rollout attempts per cycleattempts/cycle
D36Rollout attempts per cycleInclude retries, sampled traces and intended exploration.
F36Rollout attempts per cycleCycle #
G36Rollout attempts per cycleActual score
H36Rollout attempts per cycleGain vs prior
I36Rollout attempts per cycleCycle status
A37Mean rollout input tokens
B37Mean rollout input tokens12,000
C37Mean rollout input tokenstokens/attempt
D37Mean rollout input tokensEstimate from a representative agent trace sample.
F37Mean rollout input tokens1
H37Mean rollout input tokens—=IF(G37="","",G37-$B$16)
I37Mean rollout input tokensNOT RUN=IF(G37="","NOT RUN",IF(G37>=$B$17,"TARGET HIT",IF(F37>=$B$10,"BEYOND SCOPE","CHECK CONTINUE")))
A38Mean rollout output tokens
B38Mean rollout output tokens10,000
C38Mean rollout output tokenstokens/attempt
D38Mean rollout output tokensGeneration volume; avoid counting only successful rollouts.
F38Mean rollout output tokens2
H38Mean rollout output tokens—=IF(G38="","",IF(G37="","",G38-G37))
I38Mean rollout output tokensNOT RUN=IF(G38="","NOT RUN",IF(G38>=$B$17,"TARGET HIT",IF(F38>=$B$10,"BEYOND SCOPE","CHECK CONTINUE")))
A39Input-token work factor
B39Input-token work factor0.5
C39Input-token work factorequivalent/token
D39Input-token work factorSimplified prefill weight relative to decode; benchmark required.
F39Input-token work factor3
H39Input-token work factor—=IF(G39="","",IF(G38="","",G39-G38))
I39Input-token work factorNOT RUN=IF(G39="","NOT RUN",IF(G39>=$B$17,"TARGET HIT",IF(F39>=$B$10,"BEYOND SCOPE","CHECK CONTINUE")))
A40Evaluation cases / cycle
B40Evaluation cases / cycle200
C40Evaluation cases / cyclecases/cycle
D40Evaluation cases / cycleHeld-out evaluation repeated at each checkpoint.
F40Evaluation cases / cycle4
H40Evaluation cases / cycle—=IF(G40="","",IF(G39="","",G40-G39))
I40Evaluation cases / cycleNOT RUN=IF(G40="","NOT RUN",IF(G40>=$B$17,"TARGET HIT",IF(F40>=$B$10,"BEYOND SCOPE","CHECK CONTINUE")))
A41Baseline evaluation cases
B41Baseline evaluation cases200
C41Baseline evaluation casescases
D41Baseline evaluation casesRun before training; included in the evaluation budget.
F41Baseline evaluation cases5
H41Baseline evaluation cases—=IF(G41="","",IF(G40="","",G41-G40))
I41Baseline evaluation casesNOT RUN=IF(G41="","NOT RUN",IF(G41>=$B$17,"TARGET HIT",IF(F41>=$B$10,"BEYOND SCOPE","CHECK CONTINUE")))
A42Work-equivalent tokens / eval
B42Work-equivalent tokens / eval20,000
C42Work-equivalent tokens / evaltokens/case
D42Work-equivalent tokens / evalSimplified GPU work incl. judge inference; external graders separately.
F42Work-equivalent tokens / eval6
H42Work-equivalent tokens / eval—=IF(G42="","",IF(G41="","",G42-G41))
I42Work-equivalent tokens / evalNOT RUN=IF(G42="","NOT RUN",IF(G42>=$B$17,"TARGET HIT",IF(F42>=$B$10,"BEYOND SCOPE","CHECK CONTINUE")))
A43Policy throughput (standard)
B43Policy throughput (standard)100,000
C43Policy throughput (standard)tokens/GPU-hour
D43Policy throughput (standard)Unverified benchmark for 32B-class configuration.
F43Policy throughput (standard)7
H43Policy throughput (standard)—=IF(G43="","",IF(G42="","",G43-G42))
I43Policy throughput (standard)NOT RUN=IF(G43="","NOT RUN",IF(G43>=$B$17,"TARGET HIT",IF(F43>=$B$10,"BEYOND SCOPE","CHECK CONTINUE")))
A44Rollout throughput (standard)
B44Rollout throughput (standard)100,000
C44Rollout throughput (standard)equiv tokens/GPU-hour
D44Rollout throughput (standard)Unverified benchmark; input/output mix influences value.
F44Rollout throughput (standard)8
H44Rollout throughput (standard)—=IF(G44="","",IF(G43="","",G44-G43))
I44Rollout throughput (standard)NOT RUN=IF(G44="","NOT RUN",IF(G44>=$B$17,"TARGET HIT",IF(F44>=$B$10,"BEYOND SCOPE","CHECK CONTINUE")))
A45Eval throughput (standard)
B45Eval throughput (standard)100,000
C45Eval throughput (standard)equiv tokens/GPU-hour
D45Eval throughput (standard)Unverified benchmark; grader APIs budgeted separately.
A46Policy GPUs (standard)
B46Policy GPUs (standard)8
C46Policy GPUs (standard)GPUs
D46Policy GPUs (standard)Illustrative 32B distributed training job size.
F46Policy GPUs (standard)Stop rule
G46Policy GPUs (standard)Target OR plateau OR cycle cap. Customer must approve the actual stop/continue decision.
A47Rollout GPUs (standard)
B47Rollout GPUs (standard)8
C47Rollout GPUs (standard)GPUs
D47Rollout GPUs (standard)Illustrative 32B rollout fleet size.
A48Evaluation GPUs (standard)
B48Evaluation GPUs (standard)4
C48Evaluation GPUs (standard)GPUs
D48Evaluation GPUs (standard)Illustrative GPU group for held-out eval.
A493 DIRECT DELIVERY COST ASSUMPTIONS
A50FDE setup hours
B50FDE setup hours40
C50FDE setup hourshours
D50FDE setup hoursOne-off integration, workflow and harness scoping.
A51FDE hours per cycle
B51FDE hours per cycle40
C51FDE hours per cyclehours/cycle
D51FDE hours per cycleDirect delivery load scales with approved loops.
A52FDE loaded cost
B52FDE loaded cost180
C52FDE loaded cost$/hour
D52FDE loaded costIllustrative loaded labor cost, not customer sell rate.
A53Research setup hours
B53Research setup hours40
C53Research setup hourshours
D53Research setup hoursBaseline, grader and model selection support.
A54Research hours per cycle
B54Research hours per cycle40
C54Research hours per cyclehours/cycle
D54Research hours per cycleTraining diagnostics and quality review.
A55Research loaded cost
B55Research loaded cost200
C55Research loaded cost$/hour
D55Research loaded costIllustrative loaded labor cost.
A56Infra setup hours
B56Infra setup hours12
C56Infra setup hourshours
D56Infra setup hoursInitial deployment, storage and capacity provisioning.
A57Infra hours per cycle
B57Infra hours per cycle16
C57Infra hours per cyclehours/cycle
D57Infra hours per cycleRuntime support and operational monitoring.
A58Infra loaded cost
B58Infra loaded cost170
C58Infra loaded cost$/hour
D58Infra loaded costIllustrative loaded labor cost.
A59Delivery setup hours
B59Delivery setup hours16
C59Delivery setup hourshours
D59Delivery setup hoursCustomer milestones and coordinating approvals.
A60Delivery hours per cycle
B60Delivery hours per cycle8
C60Delivery hours per cyclehours/cycle
D60Delivery hours per cycleStatus, review meeting, completion evidence.
A61Delivery loaded cost
B61Delivery loaded cost150
C61Delivery loaded cost$/hour
D61Delivery loaded costIllustrative loaded labor cost.
A62Baseline graders / environments
B62Baseline graders / environments1,500
C62Baseline graders / environments$
D62Baseline graders / environmentsOne-time APIs, sandbox, human quality checks.
A63Grader / environment per cycle
B63Grader / environment per cycle1,500
C63Grader / environment per cycle$/cycle
D63Grader / environment per cycleSome judge and tool costs are NOT GPU based.
A64Storage / networking
B64Storage / networking1,500
C64Storage / networking$
D64Storage / networkingIllustrative checkpoint, transfer, logs and object storage.
A65Direct platform support
B65Direct platform support4,000
C65Direct platform support$
D65Direct platform supportDeal-specific operating cost, not corporate overhead.
A66Expected rework allowance
B66Expected rework allowance0.1
C66Expected rework allowance% COGS
D66Expected rework allowanceExpected delivery variation; forecast, not booked accounting expense.
A674 CUSTOMER QUOTE + MARGIN POLICY
A68Readiness fixed fee
B68Readiness fixed fee30,000
C68Readiness fixed fee$
D68Readiness fixed feeCustomer-facing fixed feasibility, baseline and grader design.
A69Engineering setup fee
B69Engineering setup fee78,000
C69Engineering setup fee$
D69Engineering setup feeCustomer-facing engineering/harness delivery across engagement.
A70RL cycle fee (standard)
B70RL cycle fee (standard)24,000
C70RL cycle fee (standard)$/cycle
D70RL cycle fee (standard)Customer fee per approved baseline-size cycle.
A71Included cycles before added labor
B71Included cycles before added labor3
C71Included cycles before added laborcycles
D71Included cycles before added laborExtra cycle also needs named incremental engineering allocation.
A72Added engineering fee / extra cycle
B72Added engineering fee / extra cycle24,000
C72Added engineering fee / extra cycle$/cycle
D72Added engineering fee / extra cycleOnly for loops above included engineering scope.
A73Independent final validation fee
B73Independent final validation fee30,000
C73Independent final validation fee$
D73Independent final validation feeHeld-out results, reproducibility and model handoff.
A75Negotiated project discount
B75Negotiated project discount0
C75Negotiated project discount%
D75Negotiated project discountDiscount reduces commercial margin; no hidden cost reduction.
A76Target engagement gross margin
B76Target engagement gross margin0.45
C76Target engagement gross margin%
D76Target engagement gross marginIllustrative policy to test, not company-provided target.
A77Minimum acceptable GM
B77Minimum acceptable GM0.35
C77Minimum acceptable GM%
D77Minimum acceptable GMTriggers warning in discount/underpricing scenario.
A785 GPU CAPACITY + PROCUREMENT
A79Use compatible existing GPUs?
B79Use compatible existing GPUs?Yes
C79Use compatible existing GPUs?Yes / No
D79Use compatible existing GPUs?Existing must be spare and topology/region compatible.
A80New capacity acquisition
B80New capacity acquisitionAuto
C80New capacity acquisitionAuto / PayGo / Extend / Reserve / Existing only / BYOC
D80New capacity acquisitionAuto minimizes modeled incremental vendor cash (not risk-adjusted opportunity).
A81Extra capacity can be obtained?
B81Extra capacity can be obtained?No
C81Extra capacity can be obtained?Yes / No
D81Extra capacity can be obtained?Supplier availability NOT verified; No blocks uncovered workloads.
A82PayGo unit price
B82PayGo unit price3.99
C82PayGo unit price$/GPU-hour
D82PayGo unit priceIllustrative public H100 benchmark; not AC internal contracted price.
A83One-week extension unit price
B83One-week extension unit price2.75
C83One-week extension unit price$/GPU-hour
D83One-week extension unit priceIllustrative; extensions must actually be available.
A84New reservation unit price
B84New reservation unit price2.5
C84New reservation unit price$/GPU-hour
D84New reservation unit priceIllustrative price subject to 4-week minimum.
A85Minimum new reservation
B85Minimum new reservation4
C85Minimum new reservationweeks
D85Minimum new reservationReservation cash paid even when unused.
A86Hours per scheduled week
B86Hours per scheduled week168
C86Hours per scheduled weekhours
D86Hours per scheduled weekUpper bound; operations/maintenance reduce usable hours.
A87Existing capacity allocation cost
B87Existing capacity allocation cost2.5
C87Existing capacity allocation cost$/GPU-hour
D87Existing capacity allocation costManagerial COGS for already-paid GPU hours; not incremental cash.
A88Existing capacity opportunity value
B88Existing capacity opportunity value0
C88Existing capacity opportunity value$/GPU-hour
D88Existing capacity opportunity valueValue lost from displacing real alternative; not accounting COGS.
A89New provisioning / setup
B89New provisioning / setup0
C89New provisioning / setup$
D89New provisioning / setupReserved to add one-off cluster integration fees.
A906 OPTIONAL PRODUCTION INFERENCE
A91Shared minimum monthly
B91Shared minimum monthly15,000
C91Shared minimum monthly$/month
D91Shared minimum monthlyIllustrative contracted minimum for managed shared inference.
A92Dedicated warm GPU fee
B92Dedicated warm GPU fee6,000
C92Dedicated warm GPU fee$/GPU-month
D92Dedicated warm GPU feeIllustrative capacity subscription; check utilization and SLA.
A93Input token overage rate
B93Input token overage rate1.2
C93Input token overage rate$/1M tokens
D93Input token overage rateCharge beyond included input token allowance.
A94Output token overage rate
B94Output token overage rate4.5
C94Output token overage rate$/1M tokens
D94Output token overage rateCharge beyond included output token allowance.
A95Included monthly input tokens
B95Included monthly input tokens20,000,000
C95Included monthly input tokenstokens/month
D95Included monthly input tokensCustomer service-order threshold; overage beyond this.
A96Included monthly output tokens
B96Included monthly output tokens10,000,000
C96Included monthly output tokenstokens/month
D96Included monthly output tokensCustomer service-order threshold; overage beyond this.
A97Serving engineering support
B97Serving engineering support2,500
C97Serving engineering support$/month
D97Serving engineering supportDirect production support and observability forecast.
A98Serving throughput (standard)
B98Serving throughput (standard)100,000
C98Serving throughput (standard)equiv tokens/GPU-hour
D98Serving throughput (standard)Must benchmark SLO-compliant serving throughput.
A99Shared capacity share
B99Shared capacity share0.35
C99Shared capacity share%
D99Shared capacity shareAllocated share of shared serving fleet; finance assumption.
A100MODEL AND HARDWARE APPROVAL GATES (scenario prices still calculate when No)
A101AC workload benchmark confirmed?
B101AC workload benchmark confirmed?No
C101AC workload benchmark confirmed?Yes / No
D101AC workload benchmark confirmed?Includes measured trainer, rollout and eval throughput for selected model and GPU class.
A102Training access / licensing approved?
B102Training access / licensing approved?No
C102Training access / licensing approved?Yes / No
D102Training access / licensing approved?Customer + AC confirm weights, license and AC2 training access.
A103GPU accelerator planned
B103GPU accelerator plannedH100
C103GPU accelerator plannedH100 / B200 / B300 / Other
D103GPU accelerator plannedExisting capacity pools must be same SKU family or separately benchmarked.
A104GPU SKU-specific rate confirmed?
B104GPU SKU-specific rate confirmed?No
C104GPU SKU-specific rate confirmed?Yes / No
D104GPU SKU-specific rate confirmed?H100 starter rate must not silently price B200/B300; verify supplier cost and allocation rate.
A1077 EXISTING CAPACITY | SUPPLY TYPE MUST MATCH GPU FAMILY IN B103
A108Pool / contract
B108Pool / contractFree GPU units
C108Pool / contractStart week
D108Pool / contractExpiry week
E108Pool / contractComment
F108Pool / contractRegion / fit
G108Pool / contractGPU family
H108Pool / contract▼ Fit status
A109Prepaid cluster A
B109Prepaid cluster A8
C109Prepaid cluster A1
D109Prepaid cluster A8
E109Prepaid cluster AAlready paid; check topology
F109Prepaid cluster ACompatible
G109Prepaid cluster AH100
A110Other pool B
B110Other pool B0
C110Other pool B3
D110Other pool B5
E110Other pool BOptional second tranche
F110Other pool BCompatible
G110Other pool BH100
A111Other pool C
B111Other pool C0
C111Other pool C5
D111Other pool C8
E111Other pool COptional third tranche
F111Other pool CCompatible
G111Other pool CH100
02 Economics Capacity · 530 cells, 307 formulas
CellLabelSaved valueFormula
A1DEAL ECONOMICS + GPU CAPACITY
A2Automatic calculations - demand, feasible capacity, incremental cash, fully loaded delivery COGS, and recommended quote.
A41 DEMAND -> GPU-HOURS
F41 DEMAND -> GPU-HOURSHOW TO INTERPRET THE COSTS
A5Selected model
B5Selected modelQwen3.6-35B-A3B='01 Deal Setup'!B7
F5Selected modelDemand
G5Selected modelTokens, rollouts, evals
H5Selected modelCustomer-controlled scope
A6Proxy throughput factor (benchmark required)
B6Proxy throughput factor (benchmark required)1=IF('01 Deal Setup'!$G$31=1,2.5,IF('01 Deal Setup'!$G$31=2,1,0.5))
C6Proxy throughput factor (benchmark required)x
F6Proxy throughput factor (benchmark required)Performance
G6Proxy throughput factor (benchmark required)Model-specific throughput
H6Proxy throughput factor (benchmark required)Replace with measured benchmarks
A7Planning trainer GPUs
B7Planning trainer GPUs8=MAX(1,ROUND('01 Deal Setup'!B46*IF('01 Deal Setup'!$G$31=1,0.5,IF('01 Deal Setup'!$G$31=2,1,2)),0))
C7Planning trainer GPUsGPUs
F7Planning trainer GPUsCapacity
G7Planning trainer GPUsPeak GPUs and timing
H7Planning trainer GPUsAvailability is not GPU-hours alone
A8Planning rollout GPUs
B8Planning rollout GPUs8=MAX(1,ROUND('01 Deal Setup'!B47*IF('01 Deal Setup'!$G$31=1,0.5,IF('01 Deal Setup'!$G$31=2,1,2)),0))
C8Planning rollout GPUsGPUs
F8Planning rollout GPUsNew capacity cash
G8Planning rollout GPUsWhat we must newly buy
H8Planning rollout GPUsPayGo vs extension vs reservation
A9Planning eval GPUs
B9Planning eval GPUs4=MAX(1,ROUND('01 Deal Setup'!B48*IF('01 Deal Setup'!$G$31=1,0.5,IF('01 Deal Setup'!$G$31=2,1,2)),0))
C9Planning eval GPUsGPUs
F9Planning eval GPUsAllocated existing
G9Planning eval GPUsAlready-paid GPU value
H9Planning eval GPUsGross-margin lens, not new cash
A10Authorized RL cycles
B10Authorized RL cycles3='01 Deal Setup'!B10
C10Authorized RL cyclescycles
F10Authorized RL cyclesOpportunity
G10Authorized RL cyclesLost alternative contribution
H10Authorized RL cyclesEconomic choice, not accounting COGS
F11Input + output equivalent tokens / monthCustomer quote
G11Input + output equivalent tokens / monthItemized fee + explicit uplift
H11Input + output equivalent tokens / monthMargin target is a proposed policy
A122 TOTAL WORKLOAD (ALL APPROVED CYCLES)
A13Policy update tokens
B13Policy update tokens128,000,000='01 Deal Setup'!B35*B10
C13Policy update tokenstokens
A14RL rollout attempts
B14RL rollout attempts12,000='01 Deal Setup'!B36*B10
C14RL rollout attemptsattempts
F14RL rollout attemptsSupplier rates, GPU fit, commitment dates, and throughput are unverified sample inputs; DO NOT issue a live quote without checking them.
A15Rollout equivalent tokens
B15Rollout equivalent tokens192,000,000=B14*('01 Deal Setup'!B37*'01 Deal Setup'!B39+'01 Deal Setup'!B38)
C15Rollout equivalent tokensequiv tokens
A16Total eval cases
B16Total eval cases800='01 Deal Setup'!B41+B10*'01 Deal Setup'!B40
C16Total eval casescases
A17Eval equivalent tokens
B17Eval equivalent tokens16,000,000=B16*'01 Deal Setup'!B42
C17Eval equivalent tokensequiv tokens
A18Policy update GPU-hours
B18Policy update GPU-hours1,280=IF('01 Deal Setup'!B43<=0,0,B13/('01 Deal Setup'!B43*B6))
C18Policy update GPU-hoursGPU-hours
A19Rollout GPU-hours
B19Rollout GPU-hours1,920=IF('01 Deal Setup'!B44<=0,0,B15/('01 Deal Setup'!B44*B6))
C19Rollout GPU-hoursGPU-hours
A20Evaluation GPU-hours
B20Evaluation GPU-hours160=IF('01 Deal Setup'!B45<=0,0,B17/('01 Deal Setup'!B45*B6))
C20Evaluation GPU-hoursGPU-hours
A21Total workload GPU-hours
B21Total workload GPU-hours3,360=SUM(B18:B20)
C21Total workload GPU-hoursGPU-hours
A22Trainer GPU-hours per cycle
B22Trainer GPU-hours per cycle426.6667=IF(B10=0,0,B18/B10)
C22Trainer GPU-hours per cycleGPUh / cycle
A23Rollout GPU-hours per cycle
B23Rollout GPU-hours per cycle640=IF(B10=0,0,B19/B10)
C23Rollout GPU-hours per cycleGPUh / cycle
A24Evaluation GPU-hours per cycle
B24Evaluation GPU-hours per cycle40=IF('01 Deal Setup'!B45<=0,0,'01 Deal Setup'!B40*'01 Deal Setup'!B42/('01 Deal Setup'!B45*B6))
C24Evaluation GPU-hours per cycleGPUh / cycle
A25Baseline evaluation GPU-hours
B25Baseline evaluation GPU-hours40=IF('01 Deal Setup'!B45<=0,0,'01 Deal Setup'!B41*'01 Deal Setup'!B42/('01 Deal Setup'!B45*B6))
C25Baseline evaluation GPU-hoursGPUh
A26Peak GPU units in RL
B26Peak GPU units in RL8=IF('01 Deal Setup'!B14="Concurrent",B7+B8,MAX(B7,B8))
C26Peak GPU units in RLGPUs
A27RL activity wall-hours / cycle
B27RL activity wall-hours / cycle143.3333=IF('01 Deal Setup'!B14="Concurrent",MAX(B22/B7,B23/B8),B22/B7+B23/B8)+B24/B9
C27RL activity wall-hours / cyclehours
A28Total planned weeks
B28Total planned weeks6='01 Deal Setup'!B12+'01 Deal Setup'!B10*'01 Deal Setup'!B11+'01 Deal Setup'!B13
C28Total planned weeksweeks
A303 WEEK-BY-WEEK CAPACITY (WEEKS 1-16)
B31Week
C31WeekRL cycle?
D31WeekPolicy GPUh
E31WeekRollout GPUh
F31WeekEval GPUh
G31WeekTotal GPUh
H31WeekPeak GPUs
I31WeekExisting GPUs
J31WeekNeed GPUs
K31WeekNew GPUh
L31WeekExisting GPUh
M31Week1wk cost
N31WeekFirst wk helper
O31WeekLast wk helper
P31WeekStatus
B321
C320=IF(AND(B32>'01 Deal Setup'!$B$12,B32<='01 Deal Setup'!$B$12+'01 Deal Setup'!$B$10*'01 Deal Setup'!$B$11,MOD(B32-'01 Deal Setup'!$B$12-1,'01 Deal Setup'!$B$11)=0),1,0)
D320=C32*$B$22
E320=C32*$B$23
F320=IF(B32='01 Deal Setup'!$B$12,$B$25,0)+IF(C32=1,$B$24,0)
G320=SUM(D32:F32)
H320=IF(C32=1,MAX($B$26,$B$9),IF(F32>0,$B$9,0))
I328=IF('01 Deal Setup'!$B$79="No",0,IF(AND(B32>='01 Deal Setup'!$C$109,B32<='01 Deal Setup'!$D$109,'01 Deal Setup'!$F$109="Compatible",'01 Deal Setup'!$G$109='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$109,0)+IF(AND(B32>='01 Deal Setup'!$C$110,B32<='01 Deal Setup'!$D$110,'01 Deal Setup'!$F$110="Compatible",'01 Deal Setup'!$G$110='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$110,0)+IF(AND(B32>='01 Deal Setup'!$C$111,B32<='01 Deal Setup'!$D$111,'01 Deal Setup'!$F$111="Compatible",'01 Deal Setup'!$G$111='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$111,0))
J320=MAX(0,H32-I32)
K320=IF(C32=1,IF('01 Deal Setup'!$B$14="Concurrent",MAX(0,$B$8-I32)*$B$23/$B$8+MAX(0,$B$7-MAX(0,I32-$B$8))*$B$22/$B$7,MAX(0,H32-I32)*SUM(D32:E32)/MAX(1,H32)),0)+IF(F32>0,MAX(0,$B$9-I32)*$F32/$B$9,0)
L320=IF('01 Deal Setup'!$B$80="BYOC",0,G32-K32)
M320=J32*'01 Deal Setup'!$B$86*'01 Deal Setup'!$B$83
N32999=IF(J32>0,B32,999)
O320=IF(J32>0,B32,0)
P32-=IF(AND(J32>0,OR('01 Deal Setup'!$B$80="Existing only",'01 Deal Setup'!$B$81="No")),"UNFUNDED",IF(AND(C32=1,$B$27>'01 Deal Setup'!$B$86),"SLOW CYCLE",IF(G32=0,"-","PLANNED")))
B332
C330=IF(AND(B33>'01 Deal Setup'!$B$12,B33<='01 Deal Setup'!$B$12+'01 Deal Setup'!$B$10*'01 Deal Setup'!$B$11,MOD(B33-'01 Deal Setup'!$B$12-1,'01 Deal Setup'!$B$11)=0),1,0)
D330=C33*$B$22
E330=C33*$B$23
F3340=IF(B33='01 Deal Setup'!$B$12,$B$25,0)+IF(C33=1,$B$24,0)
G3340=SUM(D33:F33)
H334=IF(C33=1,MAX($B$26,$B$9),IF(F33>0,$B$9,0))
I338=IF('01 Deal Setup'!$B$79="No",0,IF(AND(B33>='01 Deal Setup'!$C$109,B33<='01 Deal Setup'!$D$109,'01 Deal Setup'!$F$109="Compatible",'01 Deal Setup'!$G$109='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$109,0)+IF(AND(B33>='01 Deal Setup'!$C$110,B33<='01 Deal Setup'!$D$110,'01 Deal Setup'!$F$110="Compatible",'01 Deal Setup'!$G$110='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$110,0)+IF(AND(B33>='01 Deal Setup'!$C$111,B33<='01 Deal Setup'!$D$111,'01 Deal Setup'!$F$111="Compatible",'01 Deal Setup'!$G$111='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$111,0))
J330=MAX(0,H33-I33)
K330=IF(C33=1,IF('01 Deal Setup'!$B$14="Concurrent",MAX(0,$B$8-I33)*$B$23/$B$8+MAX(0,$B$7-MAX(0,I33-$B$8))*$B$22/$B$7,MAX(0,H33-I33)*SUM(D33:E33)/MAX(1,H33)),0)+IF(F33>0,MAX(0,$B$9-I33)*$F33/$B$9,0)
L3340=IF('01 Deal Setup'!$B$80="BYOC",0,G33-K33)
M330=J33*'01 Deal Setup'!$B$86*'01 Deal Setup'!$B$83
N33999=IF(J33>0,B33,999)
O330=IF(J33>0,B33,0)
P33PLANNED=IF(AND(J33>0,OR('01 Deal Setup'!$B$80="Existing only",'01 Deal Setup'!$B$81="No")),"UNFUNDED",IF(AND(C33=1,$B$27>'01 Deal Setup'!$B$86),"SLOW CYCLE",IF(G33=0,"-","PLANNED")))
B343
C341=IF(AND(B34>'01 Deal Setup'!$B$12,B34<='01 Deal Setup'!$B$12+'01 Deal Setup'!$B$10*'01 Deal Setup'!$B$11,MOD(B34-'01 Deal Setup'!$B$12-1,'01 Deal Setup'!$B$11)=0),1,0)
D34426.6667=C34*$B$22
E34640=C34*$B$23
F3440=IF(B34='01 Deal Setup'!$B$12,$B$25,0)+IF(C34=1,$B$24,0)
G341,106.6667=SUM(D34:F34)
H348=IF(C34=1,MAX($B$26,$B$9),IF(F34>0,$B$9,0))
I348=IF('01 Deal Setup'!$B$79="No",0,IF(AND(B34>='01 Deal Setup'!$C$109,B34<='01 Deal Setup'!$D$109,'01 Deal Setup'!$F$109="Compatible",'01 Deal Setup'!$G$109='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$109,0)+IF(AND(B34>='01 Deal Setup'!$C$110,B34<='01 Deal Setup'!$D$110,'01 Deal Setup'!$F$110="Compatible",'01 Deal Setup'!$G$110='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$110,0)+IF(AND(B34>='01 Deal Setup'!$C$111,B34<='01 Deal Setup'!$D$111,'01 Deal Setup'!$F$111="Compatible",'01 Deal Setup'!$G$111='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$111,0))
J340=MAX(0,H34-I34)
K340=IF(C34=1,IF('01 Deal Setup'!$B$14="Concurrent",MAX(0,$B$8-I34)*$B$23/$B$8+MAX(0,$B$7-MAX(0,I34-$B$8))*$B$22/$B$7,MAX(0,H34-I34)*SUM(D34:E34)/MAX(1,H34)),0)+IF(F34>0,MAX(0,$B$9-I34)*$F34/$B$9,0)
L341,106.6667=IF('01 Deal Setup'!$B$80="BYOC",0,G34-K34)
M340=J34*'01 Deal Setup'!$B$86*'01 Deal Setup'!$B$83
N34999=IF(J34>0,B34,999)
O340=IF(J34>0,B34,0)
P34PLANNED=IF(AND(J34>0,OR('01 Deal Setup'!$B$80="Existing only",'01 Deal Setup'!$B$81="No")),"UNFUNDED",IF(AND(C34=1,$B$27>'01 Deal Setup'!$B$86),"SLOW CYCLE",IF(G34=0,"-","PLANNED")))
B354
C351=IF(AND(B35>'01 Deal Setup'!$B$12,B35<='01 Deal Setup'!$B$12+'01 Deal Setup'!$B$10*'01 Deal Setup'!$B$11,MOD(B35-'01 Deal Setup'!$B$12-1,'01 Deal Setup'!$B$11)=0),1,0)
D35426.6667=C35*$B$22
E35640=C35*$B$23
F3540=IF(B35='01 Deal Setup'!$B$12,$B$25,0)+IF(C35=1,$B$24,0)
G351,106.6667=SUM(D35:F35)
H358=IF(C35=1,MAX($B$26,$B$9),IF(F35>0,$B$9,0))
I358=IF('01 Deal Setup'!$B$79="No",0,IF(AND(B35>='01 Deal Setup'!$C$109,B35<='01 Deal Setup'!$D$109,'01 Deal Setup'!$F$109="Compatible",'01 Deal Setup'!$G$109='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$109,0)+IF(AND(B35>='01 Deal Setup'!$C$110,B35<='01 Deal Setup'!$D$110,'01 Deal Setup'!$F$110="Compatible",'01 Deal Setup'!$G$110='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$110,0)+IF(AND(B35>='01 Deal Setup'!$C$111,B35<='01 Deal Setup'!$D$111,'01 Deal Setup'!$F$111="Compatible",'01 Deal Setup'!$G$111='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$111,0))
J350=MAX(0,H35-I35)
K350=IF(C35=1,IF('01 Deal Setup'!$B$14="Concurrent",MAX(0,$B$8-I35)*$B$23/$B$8+MAX(0,$B$7-MAX(0,I35-$B$8))*$B$22/$B$7,MAX(0,H35-I35)*SUM(D35:E35)/MAX(1,H35)),0)+IF(F35>0,MAX(0,$B$9-I35)*$F35/$B$9,0)
L351,106.6667=IF('01 Deal Setup'!$B$80="BYOC",0,G35-K35)
M350=J35*'01 Deal Setup'!$B$86*'01 Deal Setup'!$B$83
N35999=IF(J35>0,B35,999)
O350=IF(J35>0,B35,0)
P35PLANNED=IF(AND(J35>0,OR('01 Deal Setup'!$B$80="Existing only",'01 Deal Setup'!$B$81="No")),"UNFUNDED",IF(AND(C35=1,$B$27>'01 Deal Setup'!$B$86),"SLOW CYCLE",IF(G35=0,"-","PLANNED")))
B365
C361=IF(AND(B36>'01 Deal Setup'!$B$12,B36<='01 Deal Setup'!$B$12+'01 Deal Setup'!$B$10*'01 Deal Setup'!$B$11,MOD(B36-'01 Deal Setup'!$B$12-1,'01 Deal Setup'!$B$11)=0),1,0)
D36426.6667=C36*$B$22
E36640=C36*$B$23
F3640=IF(B36='01 Deal Setup'!$B$12,$B$25,0)+IF(C36=1,$B$24,0)
G361,106.6667=SUM(D36:F36)
H368=IF(C36=1,MAX($B$26,$B$9),IF(F36>0,$B$9,0))
I368=IF('01 Deal Setup'!$B$79="No",0,IF(AND(B36>='01 Deal Setup'!$C$109,B36<='01 Deal Setup'!$D$109,'01 Deal Setup'!$F$109="Compatible",'01 Deal Setup'!$G$109='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$109,0)+IF(AND(B36>='01 Deal Setup'!$C$110,B36<='01 Deal Setup'!$D$110,'01 Deal Setup'!$F$110="Compatible",'01 Deal Setup'!$G$110='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$110,0)+IF(AND(B36>='01 Deal Setup'!$C$111,B36<='01 Deal Setup'!$D$111,'01 Deal Setup'!$F$111="Compatible",'01 Deal Setup'!$G$111='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$111,0))
J360=MAX(0,H36-I36)
K360=IF(C36=1,IF('01 Deal Setup'!$B$14="Concurrent",MAX(0,$B$8-I36)*$B$23/$B$8+MAX(0,$B$7-MAX(0,I36-$B$8))*$B$22/$B$7,MAX(0,H36-I36)*SUM(D36:E36)/MAX(1,H36)),0)+IF(F36>0,MAX(0,$B$9-I36)*$F36/$B$9,0)
L361,106.6667=IF('01 Deal Setup'!$B$80="BYOC",0,G36-K36)
M360=J36*'01 Deal Setup'!$B$86*'01 Deal Setup'!$B$83
N36999=IF(J36>0,B36,999)
O360=IF(J36>0,B36,0)
P36PLANNED=IF(AND(J36>0,OR('01 Deal Setup'!$B$80="Existing only",'01 Deal Setup'!$B$81="No")),"UNFUNDED",IF(AND(C36=1,$B$27>'01 Deal Setup'!$B$86),"SLOW CYCLE",IF(G36=0,"-","PLANNED")))
B376
C370=IF(AND(B37>'01 Deal Setup'!$B$12,B37<='01 Deal Setup'!$B$12+'01 Deal Setup'!$B$10*'01 Deal Setup'!$B$11,MOD(B37-'01 Deal Setup'!$B$12-1,'01 Deal Setup'!$B$11)=0),1,0)
D370=C37*$B$22
E370=C37*$B$23
F370=IF(B37='01 Deal Setup'!$B$12,$B$25,0)+IF(C37=1,$B$24,0)
G370=SUM(D37:F37)
H370=IF(C37=1,MAX($B$26,$B$9),IF(F37>0,$B$9,0))
I378=IF('01 Deal Setup'!$B$79="No",0,IF(AND(B37>='01 Deal Setup'!$C$109,B37<='01 Deal Setup'!$D$109,'01 Deal Setup'!$F$109="Compatible",'01 Deal Setup'!$G$109='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$109,0)+IF(AND(B37>='01 Deal Setup'!$C$110,B37<='01 Deal Setup'!$D$110,'01 Deal Setup'!$F$110="Compatible",'01 Deal Setup'!$G$110='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$110,0)+IF(AND(B37>='01 Deal Setup'!$C$111,B37<='01 Deal Setup'!$D$111,'01 Deal Setup'!$F$111="Compatible",'01 Deal Setup'!$G$111='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$111,0))
J370=MAX(0,H37-I37)
K370=IF(C37=1,IF('01 Deal Setup'!$B$14="Concurrent",MAX(0,$B$8-I37)*$B$23/$B$8+MAX(0,$B$7-MAX(0,I37-$B$8))*$B$22/$B$7,MAX(0,H37-I37)*SUM(D37:E37)/MAX(1,H37)),0)+IF(F37>0,MAX(0,$B$9-I37)*$F37/$B$9,0)
L370=IF('01 Deal Setup'!$B$80="BYOC",0,G37-K37)
M370=J37*'01 Deal Setup'!$B$86*'01 Deal Setup'!$B$83
N37999=IF(J37>0,B37,999)
O370=IF(J37>0,B37,0)
P37-=IF(AND(J37>0,OR('01 Deal Setup'!$B$80="Existing only",'01 Deal Setup'!$B$81="No")),"UNFUNDED",IF(AND(C37=1,$B$27>'01 Deal Setup'!$B$86),"SLOW CYCLE",IF(G37=0,"-","PLANNED")))
B387
C380=IF(AND(B38>'01 Deal Setup'!$B$12,B38<='01 Deal Setup'!$B$12+'01 Deal Setup'!$B$10*'01 Deal Setup'!$B$11,MOD(B38-'01 Deal Setup'!$B$12-1,'01 Deal Setup'!$B$11)=0),1,0)
D380=C38*$B$22
E380=C38*$B$23
F380=IF(B38='01 Deal Setup'!$B$12,$B$25,0)+IF(C38=1,$B$24,0)
G380=SUM(D38:F38)
H380=IF(C38=1,MAX($B$26,$B$9),IF(F38>0,$B$9,0))
I388=IF('01 Deal Setup'!$B$79="No",0,IF(AND(B38>='01 Deal Setup'!$C$109,B38<='01 Deal Setup'!$D$109,'01 Deal Setup'!$F$109="Compatible",'01 Deal Setup'!$G$109='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$109,0)+IF(AND(B38>='01 Deal Setup'!$C$110,B38<='01 Deal Setup'!$D$110,'01 Deal Setup'!$F$110="Compatible",'01 Deal Setup'!$G$110='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$110,0)+IF(AND(B38>='01 Deal Setup'!$C$111,B38<='01 Deal Setup'!$D$111,'01 Deal Setup'!$F$111="Compatible",'01 Deal Setup'!$G$111='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$111,0))
J380=MAX(0,H38-I38)
K380=IF(C38=1,IF('01 Deal Setup'!$B$14="Concurrent",MAX(0,$B$8-I38)*$B$23/$B$8+MAX(0,$B$7-MAX(0,I38-$B$8))*$B$22/$B$7,MAX(0,H38-I38)*SUM(D38:E38)/MAX(1,H38)),0)+IF(F38>0,MAX(0,$B$9-I38)*$F38/$B$9,0)
L380=IF('01 Deal Setup'!$B$80="BYOC",0,G38-K38)
M380=J38*'01 Deal Setup'!$B$86*'01 Deal Setup'!$B$83
N38999=IF(J38>0,B38,999)
O380=IF(J38>0,B38,0)
P38-=IF(AND(J38>0,OR('01 Deal Setup'!$B$80="Existing only",'01 Deal Setup'!$B$81="No")),"UNFUNDED",IF(AND(C38=1,$B$27>'01 Deal Setup'!$B$86),"SLOW CYCLE",IF(G38=0,"-","PLANNED")))
B398
C390=IF(AND(B39>'01 Deal Setup'!$B$12,B39<='01 Deal Setup'!$B$12+'01 Deal Setup'!$B$10*'01 Deal Setup'!$B$11,MOD(B39-'01 Deal Setup'!$B$12-1,'01 Deal Setup'!$B$11)=0),1,0)
D390=C39*$B$22
E390=C39*$B$23
F390=IF(B39='01 Deal Setup'!$B$12,$B$25,0)+IF(C39=1,$B$24,0)
G390=SUM(D39:F39)
H390=IF(C39=1,MAX($B$26,$B$9),IF(F39>0,$B$9,0))
I398=IF('01 Deal Setup'!$B$79="No",0,IF(AND(B39>='01 Deal Setup'!$C$109,B39<='01 Deal Setup'!$D$109,'01 Deal Setup'!$F$109="Compatible",'01 Deal Setup'!$G$109='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$109,0)+IF(AND(B39>='01 Deal Setup'!$C$110,B39<='01 Deal Setup'!$D$110,'01 Deal Setup'!$F$110="Compatible",'01 Deal Setup'!$G$110='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$110,0)+IF(AND(B39>='01 Deal Setup'!$C$111,B39<='01 Deal Setup'!$D$111,'01 Deal Setup'!$F$111="Compatible",'01 Deal Setup'!$G$111='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$111,0))
J390=MAX(0,H39-I39)
K390=IF(C39=1,IF('01 Deal Setup'!$B$14="Concurrent",MAX(0,$B$8-I39)*$B$23/$B$8+MAX(0,$B$7-MAX(0,I39-$B$8))*$B$22/$B$7,MAX(0,H39-I39)*SUM(D39:E39)/MAX(1,H39)),0)+IF(F39>0,MAX(0,$B$9-I39)*$F39/$B$9,0)
L390=IF('01 Deal Setup'!$B$80="BYOC",0,G39-K39)
M390=J39*'01 Deal Setup'!$B$86*'01 Deal Setup'!$B$83
N39999=IF(J39>0,B39,999)
O390=IF(J39>0,B39,0)
P39-=IF(AND(J39>0,OR('01 Deal Setup'!$B$80="Existing only",'01 Deal Setup'!$B$81="No")),"UNFUNDED",IF(AND(C39=1,$B$27>'01 Deal Setup'!$B$86),"SLOW CYCLE",IF(G39=0,"-","PLANNED")))
B409
C400=IF(AND(B40>'01 Deal Setup'!$B$12,B40<='01 Deal Setup'!$B$12+'01 Deal Setup'!$B$10*'01 Deal Setup'!$B$11,MOD(B40-'01 Deal Setup'!$B$12-1,'01 Deal Setup'!$B$11)=0),1,0)
D400=C40*$B$22
E400=C40*$B$23
F400=IF(B40='01 Deal Setup'!$B$12,$B$25,0)+IF(C40=1,$B$24,0)
G400=SUM(D40:F40)
H400=IF(C40=1,MAX($B$26,$B$9),IF(F40>0,$B$9,0))
I400=IF('01 Deal Setup'!$B$79="No",0,IF(AND(B40>='01 Deal Setup'!$C$109,B40<='01 Deal Setup'!$D$109,'01 Deal Setup'!$F$109="Compatible",'01 Deal Setup'!$G$109='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$109,0)+IF(AND(B40>='01 Deal Setup'!$C$110,B40<='01 Deal Setup'!$D$110,'01 Deal Setup'!$F$110="Compatible",'01 Deal Setup'!$G$110='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$110,0)+IF(AND(B40>='01 Deal Setup'!$C$111,B40<='01 Deal Setup'!$D$111,'01 Deal Setup'!$F$111="Compatible",'01 Deal Setup'!$G$111='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$111,0))
J400=MAX(0,H40-I40)
K400=IF(C40=1,IF('01 Deal Setup'!$B$14="Concurrent",MAX(0,$B$8-I40)*$B$23/$B$8+MAX(0,$B$7-MAX(0,I40-$B$8))*$B$22/$B$7,MAX(0,H40-I40)*SUM(D40:E40)/MAX(1,H40)),0)+IF(F40>0,MAX(0,$B$9-I40)*$F40/$B$9,0)
L400=IF('01 Deal Setup'!$B$80="BYOC",0,G40-K40)
M400=J40*'01 Deal Setup'!$B$86*'01 Deal Setup'!$B$83
N40999=IF(J40>0,B40,999)
O400=IF(J40>0,B40,0)
P40-=IF(AND(J40>0,OR('01 Deal Setup'!$B$80="Existing only",'01 Deal Setup'!$B$81="No")),"UNFUNDED",IF(AND(C40=1,$B$27>'01 Deal Setup'!$B$86),"SLOW CYCLE",IF(G40=0,"-","PLANNED")))
B4110
C410=IF(AND(B41>'01 Deal Setup'!$B$12,B41<='01 Deal Setup'!$B$12+'01 Deal Setup'!$B$10*'01 Deal Setup'!$B$11,MOD(B41-'01 Deal Setup'!$B$12-1,'01 Deal Setup'!$B$11)=0),1,0)
D410=C41*$B$22
E410=C41*$B$23
F410=IF(B41='01 Deal Setup'!$B$12,$B$25,0)+IF(C41=1,$B$24,0)
G410=SUM(D41:F41)
H410=IF(C41=1,MAX($B$26,$B$9),IF(F41>0,$B$9,0))
I410=IF('01 Deal Setup'!$B$79="No",0,IF(AND(B41>='01 Deal Setup'!$C$109,B41<='01 Deal Setup'!$D$109,'01 Deal Setup'!$F$109="Compatible",'01 Deal Setup'!$G$109='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$109,0)+IF(AND(B41>='01 Deal Setup'!$C$110,B41<='01 Deal Setup'!$D$110,'01 Deal Setup'!$F$110="Compatible",'01 Deal Setup'!$G$110='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$110,0)+IF(AND(B41>='01 Deal Setup'!$C$111,B41<='01 Deal Setup'!$D$111,'01 Deal Setup'!$F$111="Compatible",'01 Deal Setup'!$G$111='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$111,0))
J410=MAX(0,H41-I41)
K410=IF(C41=1,IF('01 Deal Setup'!$B$14="Concurrent",MAX(0,$B$8-I41)*$B$23/$B$8+MAX(0,$B$7-MAX(0,I41-$B$8))*$B$22/$B$7,MAX(0,H41-I41)*SUM(D41:E41)/MAX(1,H41)),0)+IF(F41>0,MAX(0,$B$9-I41)*$F41/$B$9,0)
L410=IF('01 Deal Setup'!$B$80="BYOC",0,G41-K41)
M410=J41*'01 Deal Setup'!$B$86*'01 Deal Setup'!$B$83
N41999=IF(J41>0,B41,999)
O410=IF(J41>0,B41,0)
P41-=IF(AND(J41>0,OR('01 Deal Setup'!$B$80="Existing only",'01 Deal Setup'!$B$81="No")),"UNFUNDED",IF(AND(C41=1,$B$27>'01 Deal Setup'!$B$86),"SLOW CYCLE",IF(G41=0,"-","PLANNED")))
B4211
C420=IF(AND(B42>'01 Deal Setup'!$B$12,B42<='01 Deal Setup'!$B$12+'01 Deal Setup'!$B$10*'01 Deal Setup'!$B$11,MOD(B42-'01 Deal Setup'!$B$12-1,'01 Deal Setup'!$B$11)=0),1,0)
D420=C42*$B$22
E420=C42*$B$23
F420=IF(B42='01 Deal Setup'!$B$12,$B$25,0)+IF(C42=1,$B$24,0)
G420=SUM(D42:F42)
H420=IF(C42=1,MAX($B$26,$B$9),IF(F42>0,$B$9,0))
I420=IF('01 Deal Setup'!$B$79="No",0,IF(AND(B42>='01 Deal Setup'!$C$109,B42<='01 Deal Setup'!$D$109,'01 Deal Setup'!$F$109="Compatible",'01 Deal Setup'!$G$109='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$109,0)+IF(AND(B42>='01 Deal Setup'!$C$110,B42<='01 Deal Setup'!$D$110,'01 Deal Setup'!$F$110="Compatible",'01 Deal Setup'!$G$110='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$110,0)+IF(AND(B42>='01 Deal Setup'!$C$111,B42<='01 Deal Setup'!$D$111,'01 Deal Setup'!$F$111="Compatible",'01 Deal Setup'!$G$111='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$111,0))
J420=MAX(0,H42-I42)
K420=IF(C42=1,IF('01 Deal Setup'!$B$14="Concurrent",MAX(0,$B$8-I42)*$B$23/$B$8+MAX(0,$B$7-MAX(0,I42-$B$8))*$B$22/$B$7,MAX(0,H42-I42)*SUM(D42:E42)/MAX(1,H42)),0)+IF(F42>0,MAX(0,$B$9-I42)*$F42/$B$9,0)
L420=IF('01 Deal Setup'!$B$80="BYOC",0,G42-K42)
M420=J42*'01 Deal Setup'!$B$86*'01 Deal Setup'!$B$83
N42999=IF(J42>0,B42,999)
O420=IF(J42>0,B42,0)
P42-=IF(AND(J42>0,OR('01 Deal Setup'!$B$80="Existing only",'01 Deal Setup'!$B$81="No")),"UNFUNDED",IF(AND(C42=1,$B$27>'01 Deal Setup'!$B$86),"SLOW CYCLE",IF(G42=0,"-","PLANNED")))
B4312
C430=IF(AND(B43>'01 Deal Setup'!$B$12,B43<='01 Deal Setup'!$B$12+'01 Deal Setup'!$B$10*'01 Deal Setup'!$B$11,MOD(B43-'01 Deal Setup'!$B$12-1,'01 Deal Setup'!$B$11)=0),1,0)
D430=C43*$B$22
E430=C43*$B$23
F430=IF(B43='01 Deal Setup'!$B$12,$B$25,0)+IF(C43=1,$B$24,0)
G430=SUM(D43:F43)
H430=IF(C43=1,MAX($B$26,$B$9),IF(F43>0,$B$9,0))
I430=IF('01 Deal Setup'!$B$79="No",0,IF(AND(B43>='01 Deal Setup'!$C$109,B43<='01 Deal Setup'!$D$109,'01 Deal Setup'!$F$109="Compatible",'01 Deal Setup'!$G$109='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$109,0)+IF(AND(B43>='01 Deal Setup'!$C$110,B43<='01 Deal Setup'!$D$110,'01 Deal Setup'!$F$110="Compatible",'01 Deal Setup'!$G$110='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$110,0)+IF(AND(B43>='01 Deal Setup'!$C$111,B43<='01 Deal Setup'!$D$111,'01 Deal Setup'!$F$111="Compatible",'01 Deal Setup'!$G$111='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$111,0))
J430=MAX(0,H43-I43)
K430=IF(C43=1,IF('01 Deal Setup'!$B$14="Concurrent",MAX(0,$B$8-I43)*$B$23/$B$8+MAX(0,$B$7-MAX(0,I43-$B$8))*$B$22/$B$7,MAX(0,H43-I43)*SUM(D43:E43)/MAX(1,H43)),0)+IF(F43>0,MAX(0,$B$9-I43)*$F43/$B$9,0)
L430=IF('01 Deal Setup'!$B$80="BYOC",0,G43-K43)
M430=J43*'01 Deal Setup'!$B$86*'01 Deal Setup'!$B$83
N43999=IF(J43>0,B43,999)
O430=IF(J43>0,B43,0)
P43-=IF(AND(J43>0,OR('01 Deal Setup'!$B$80="Existing only",'01 Deal Setup'!$B$81="No")),"UNFUNDED",IF(AND(C43=1,$B$27>'01 Deal Setup'!$B$86),"SLOW CYCLE",IF(G43=0,"-","PLANNED")))
B4413
C440=IF(AND(B44>'01 Deal Setup'!$B$12,B44<='01 Deal Setup'!$B$12+'01 Deal Setup'!$B$10*'01 Deal Setup'!$B$11,MOD(B44-'01 Deal Setup'!$B$12-1,'01 Deal Setup'!$B$11)=0),1,0)
D440=C44*$B$22
E440=C44*$B$23
F440=IF(B44='01 Deal Setup'!$B$12,$B$25,0)+IF(C44=1,$B$24,0)
G440=SUM(D44:F44)
H440=IF(C44=1,MAX($B$26,$B$9),IF(F44>0,$B$9,0))
I440=IF('01 Deal Setup'!$B$79="No",0,IF(AND(B44>='01 Deal Setup'!$C$109,B44<='01 Deal Setup'!$D$109,'01 Deal Setup'!$F$109="Compatible",'01 Deal Setup'!$G$109='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$109,0)+IF(AND(B44>='01 Deal Setup'!$C$110,B44<='01 Deal Setup'!$D$110,'01 Deal Setup'!$F$110="Compatible",'01 Deal Setup'!$G$110='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$110,0)+IF(AND(B44>='01 Deal Setup'!$C$111,B44<='01 Deal Setup'!$D$111,'01 Deal Setup'!$F$111="Compatible",'01 Deal Setup'!$G$111='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$111,0))
J440=MAX(0,H44-I44)
K440=IF(C44=1,IF('01 Deal Setup'!$B$14="Concurrent",MAX(0,$B$8-I44)*$B$23/$B$8+MAX(0,$B$7-MAX(0,I44-$B$8))*$B$22/$B$7,MAX(0,H44-I44)*SUM(D44:E44)/MAX(1,H44)),0)+IF(F44>0,MAX(0,$B$9-I44)*$F44/$B$9,0)
L440=IF('01 Deal Setup'!$B$80="BYOC",0,G44-K44)
M440=J44*'01 Deal Setup'!$B$86*'01 Deal Setup'!$B$83
N44999=IF(J44>0,B44,999)
O440=IF(J44>0,B44,0)
P44-=IF(AND(J44>0,OR('01 Deal Setup'!$B$80="Existing only",'01 Deal Setup'!$B$81="No")),"UNFUNDED",IF(AND(C44=1,$B$27>'01 Deal Setup'!$B$86),"SLOW CYCLE",IF(G44=0,"-","PLANNED")))
B4514
C450=IF(AND(B45>'01 Deal Setup'!$B$12,B45<='01 Deal Setup'!$B$12+'01 Deal Setup'!$B$10*'01 Deal Setup'!$B$11,MOD(B45-'01 Deal Setup'!$B$12-1,'01 Deal Setup'!$B$11)=0),1,0)
D450=C45*$B$22
E450=C45*$B$23
F450=IF(B45='01 Deal Setup'!$B$12,$B$25,0)+IF(C45=1,$B$24,0)
G450=SUM(D45:F45)
H450=IF(C45=1,MAX($B$26,$B$9),IF(F45>0,$B$9,0))
I450=IF('01 Deal Setup'!$B$79="No",0,IF(AND(B45>='01 Deal Setup'!$C$109,B45<='01 Deal Setup'!$D$109,'01 Deal Setup'!$F$109="Compatible",'01 Deal Setup'!$G$109='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$109,0)+IF(AND(B45>='01 Deal Setup'!$C$110,B45<='01 Deal Setup'!$D$110,'01 Deal Setup'!$F$110="Compatible",'01 Deal Setup'!$G$110='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$110,0)+IF(AND(B45>='01 Deal Setup'!$C$111,B45<='01 Deal Setup'!$D$111,'01 Deal Setup'!$F$111="Compatible",'01 Deal Setup'!$G$111='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$111,0))
J450=MAX(0,H45-I45)
K450=IF(C45=1,IF('01 Deal Setup'!$B$14="Concurrent",MAX(0,$B$8-I45)*$B$23/$B$8+MAX(0,$B$7-MAX(0,I45-$B$8))*$B$22/$B$7,MAX(0,H45-I45)*SUM(D45:E45)/MAX(1,H45)),0)+IF(F45>0,MAX(0,$B$9-I45)*$F45/$B$9,0)
L450=IF('01 Deal Setup'!$B$80="BYOC",0,G45-K45)
M450=J45*'01 Deal Setup'!$B$86*'01 Deal Setup'!$B$83
N45999=IF(J45>0,B45,999)
O450=IF(J45>0,B45,0)
P45-=IF(AND(J45>0,OR('01 Deal Setup'!$B$80="Existing only",'01 Deal Setup'!$B$81="No")),"UNFUNDED",IF(AND(C45=1,$B$27>'01 Deal Setup'!$B$86),"SLOW CYCLE",IF(G45=0,"-","PLANNED")))
B4615
C460=IF(AND(B46>'01 Deal Setup'!$B$12,B46<='01 Deal Setup'!$B$12+'01 Deal Setup'!$B$10*'01 Deal Setup'!$B$11,MOD(B46-'01 Deal Setup'!$B$12-1,'01 Deal Setup'!$B$11)=0),1,0)
D460=C46*$B$22
E460=C46*$B$23
F460=IF(B46='01 Deal Setup'!$B$12,$B$25,0)+IF(C46=1,$B$24,0)
G460=SUM(D46:F46)
H460=IF(C46=1,MAX($B$26,$B$9),IF(F46>0,$B$9,0))
I460=IF('01 Deal Setup'!$B$79="No",0,IF(AND(B46>='01 Deal Setup'!$C$109,B46<='01 Deal Setup'!$D$109,'01 Deal Setup'!$F$109="Compatible",'01 Deal Setup'!$G$109='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$109,0)+IF(AND(B46>='01 Deal Setup'!$C$110,B46<='01 Deal Setup'!$D$110,'01 Deal Setup'!$F$110="Compatible",'01 Deal Setup'!$G$110='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$110,0)+IF(AND(B46>='01 Deal Setup'!$C$111,B46<='01 Deal Setup'!$D$111,'01 Deal Setup'!$F$111="Compatible",'01 Deal Setup'!$G$111='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$111,0))
J460=MAX(0,H46-I46)
K460=IF(C46=1,IF('01 Deal Setup'!$B$14="Concurrent",MAX(0,$B$8-I46)*$B$23/$B$8+MAX(0,$B$7-MAX(0,I46-$B$8))*$B$22/$B$7,MAX(0,H46-I46)*SUM(D46:E46)/MAX(1,H46)),0)+IF(F46>0,MAX(0,$B$9-I46)*$F46/$B$9,0)
L460=IF('01 Deal Setup'!$B$80="BYOC",0,G46-K46)
M460=J46*'01 Deal Setup'!$B$86*'01 Deal Setup'!$B$83
N46999=IF(J46>0,B46,999)
O460=IF(J46>0,B46,0)
P46-=IF(AND(J46>0,OR('01 Deal Setup'!$B$80="Existing only",'01 Deal Setup'!$B$81="No")),"UNFUNDED",IF(AND(C46=1,$B$27>'01 Deal Setup'!$B$86),"SLOW CYCLE",IF(G46=0,"-","PLANNED")))
B4716
C470=IF(AND(B47>'01 Deal Setup'!$B$12,B47<='01 Deal Setup'!$B$12+'01 Deal Setup'!$B$10*'01 Deal Setup'!$B$11,MOD(B47-'01 Deal Setup'!$B$12-1,'01 Deal Setup'!$B$11)=0),1,0)
D470=C47*$B$22
E470=C47*$B$23
F470=IF(B47='01 Deal Setup'!$B$12,$B$25,0)+IF(C47=1,$B$24,0)
G470=SUM(D47:F47)
H470=IF(C47=1,MAX($B$26,$B$9),IF(F47>0,$B$9,0))
I470=IF('01 Deal Setup'!$B$79="No",0,IF(AND(B47>='01 Deal Setup'!$C$109,B47<='01 Deal Setup'!$D$109,'01 Deal Setup'!$F$109="Compatible",'01 Deal Setup'!$G$109='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$109,0)+IF(AND(B47>='01 Deal Setup'!$C$110,B47<='01 Deal Setup'!$D$110,'01 Deal Setup'!$F$110="Compatible",'01 Deal Setup'!$G$110='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$110,0)+IF(AND(B47>='01 Deal Setup'!$C$111,B47<='01 Deal Setup'!$D$111,'01 Deal Setup'!$F$111="Compatible",'01 Deal Setup'!$G$111='01 Deal Setup'!$B$103),'01 Deal Setup'!$B$111,0))
J470=MAX(0,H47-I47)
K470=IF(C47=1,IF('01 Deal Setup'!$B$14="Concurrent",MAX(0,$B$8-I47)*$B$23/$B$8+MAX(0,$B$7-MAX(0,I47-$B$8))*$B$22/$B$7,MAX(0,H47-I47)*SUM(D47:E47)/MAX(1,H47)),0)+IF(F47>0,MAX(0,$B$9-I47)*$F47/$B$9,0)
L470=IF('01 Deal Setup'!$B$80="BYOC",0,G47-K47)
M470=J47*'01 Deal Setup'!$B$86*'01 Deal Setup'!$B$83
N47999=IF(J47>0,B47,999)
O470=IF(J47>0,B47,0)
P47-=IF(AND(J47>0,OR('01 Deal Setup'!$B$80="Existing only",'01 Deal Setup'!$B$81="No")),"UNFUNDED",IF(AND(C47=1,$B$27>'01 Deal Setup'!$B$86),"SLOW CYCLE",IF(G47=0,"-","PLANNED")))
A494 EXISTING VS NEW CAPACITY ECONOMICS
A52Weeks needing new GPUs
B52Weeks needing new GPUs0=COUNTIF(J32:J47,">0")
C52Weeks needing new GPUsweeks
A53First shortage week
B53First shortage week0=IF(B52=0,0,MIN(N32:N47))
C53First shortage weekweek
A54Last shortage week
B54Last shortage week0=MAX(O32:O47)
C54Last shortage weekweek
A55Peak additional GPU units
B55Peak additional GPU units0=MAX(J32:J47)
C55Peak additional GPU unitsGPUs
A56New workload GPU-hours
B56New workload GPU-hours0=SUM(K32:K47)
C56New workload GPU-hoursGPU-hours
A57Existing assigned GPU-hours
B57Existing assigned GPU-hours3,360=SUM(L32:L47)
C57Existing assigned GPU-hoursGPU-hours
A58PayGo acquisition cost
B58PayGo acquisition cost0=B56*'01 Deal Setup'!B82+IF(B55>0,'01 Deal Setup'!B89,0)
C58PayGo acquisition cost$
A59Weekly extension acquisition cost
B59Weekly extension acquisition cost0=SUM(M32:M47)+IF(B55>0,'01 Deal Setup'!B89,0)
C59Weekly extension acquisition cost$
A60New reservation committed cost
B60New reservation committed cost0=B55*MAX('01 Deal Setup'!B85,IF(B52=0,0,B54-B53+1))*'01 Deal Setup'!B86*'01 Deal Setup'!B84+IF(B55>0,'01 Deal Setup'!B89,0)
C60New reservation committed cost$
A61Cheapest incremental cash
B61Cheapest incremental cash0=MIN(B58:B60)
C61Cheapest incremental cash$
A62Selected purchase method
B62Selected purchase methodExisting=IF('01 Deal Setup'!B80="BYOC","BYOC",IF(B55=0,"Existing",IF('01 Deal Setup'!B80="Auto",IF(B58<=MIN(B59,B60),"PayGo",IF(B59<=B60,"Extend","Reserve")),'01 Deal Setup'!B80)))
A63Cash cost of new capacity
B63Cash cost of new capacity0=IF(OR(B62="BYOC",B62="Existing only"),0,IF(B62="PayGo",B58,IF(B62="Extend",B59,B60)))
C63Cash cost of new capacity$
A64New GPU-hours paid / committed
B64New GPU-hours paid / committed0=IF(B62="Extend",SUM(J32:J47)*'01 Deal Setup'!B86,IF(B62="Reserve",B55*MAX('01 Deal Setup'!B85,IF(B52=0,0,B54-B53+1))*'01 Deal Setup'!B86,IF(B62="PayGo",B56,0)))
C64New GPU-hours paid / committedGPU-hours
A65Unused new commitment hours
B65Unused new commitment hours0=MAX(0,B64-B56)
C65Unused new commitment hoursGPU-hours
A66Managerial allocation of existing
B66Managerial allocation of existing8,400=IF(B62="BYOC",0,B57*'01 Deal Setup'!B87)
C66Managerial allocation of existing$
A67New capacity cash exposure
B67New capacity cash exposure0=B63
C67New capacity cash exposure$
A68Existing alternative-use value
B68Existing alternative-use value0=IF(B62="BYOC",0,B57*'01 Deal Setup'!B88)
C68Existing alternative-use value$
A69Economic decision cost (cash + opportunity)
B69Economic decision cost (cash + opportunity)0=B67+B68
C69Economic decision cost (cash + opportunity)$
A70Capacity / schedule verdict
B70Capacity / schedule verdictFEASIBLE*=IF(AND('01 Deal Setup'!G28="B300",'01 Deal Setup'!B103<>"B300"),"HARDWARE MISMATCH",IF(B28>16,"EXTEND CALENDAR",IF(B27>'01 Deal Setup'!B86,"SCHEDULE RISK",IF(COUNTIF(P32:P47,"UNFUNDED")>0,"UNFUNDED CAPACITY","FEASIBLE*"))))
A725 FULL DEAL COGS (MANAGERIAL COST VIEW)
A73Forward deployed engineering
B73Forward deployed engineering28,800=('01 Deal Setup'!B50+B10*'01 Deal Setup'!B51)*'01 Deal Setup'!B52
C73Forward deployed engineering$
A74ML research engineering
B74ML research engineering32,000=('01 Deal Setup'!B53+B10*'01 Deal Setup'!B54)*'01 Deal Setup'!B55
C74ML research engineering$
A75Infrastructure engineering
B75Infrastructure engineering10,200=('01 Deal Setup'!B56+B10*'01 Deal Setup'!B57)*'01 Deal Setup'!B58
C75Infrastructure engineering$
A76Delivery coordination
B76Delivery coordination6,000=('01 Deal Setup'!B59+B10*'01 Deal Setup'!B60)*'01 Deal Setup'!B61
C76Delivery coordination$
A77Total attributable engineering
B77Total attributable engineering77,000=SUM(B73:B76)
C77Total attributable engineering$
A78Graders, sandboxes and external APIs
B78Graders, sandboxes and external APIs6,000='01 Deal Setup'!B62+'01 Deal Setup'!B63*B10
C78Graders, sandboxes and external APIs$
A79Storage and networking
B79Storage and networking1,500='01 Deal Setup'!B64
C79Storage and networking$
A80Direct platform support
B80Direct platform support4,000='01 Deal Setup'!B65
C80Direct platform support$
A81GPU COGS (allocated + new cash)
B81GPU COGS (allocated + new cash)8,400=B66+B63
C81GPU COGS (allocated + new cash)$
A82Base cost to deliver
B82Base cost to deliver96,900=SUM(B77:B81)
C82Base cost to deliver$
A83Expected rework / failure allowance
B83Expected rework / failure allowance9,690=B82*'01 Deal Setup'!B66
C83Expected rework / failure allowance$
A84Risk-adjusted expected COGS
B84Risk-adjusted expected COGS106,590=SUM(B82:B83)
C84Risk-adjusted expected COGS$
A85COGS + opportunity view
B85COGS + opportunity view106,590=B84+B68
C85COGS + opportunity view$
A896 RECOMMENDED PRICE / GROSS MARGIN
A91Readiness / baseline
B91Readiness / baseline30,000='01 Deal Setup'!B68
C91Readiness / baseline$
A92Engineering and integration
B92Engineering and integration78,000='01 Deal Setup'!B69
C92Engineering and integration$
A93Authorized RL cycles
B93Authorized RL cycles72,000=B10*'01 Deal Setup'!B70*IF('01 Deal Setup'!$G$31=1,0.65,IF('01 Deal Setup'!$G$31=2,1,2.5))
C93Authorized RL cycles$
A94Extra-cycle engineering
B94Extra-cycle engineering0=MAX(0,B10-'01 Deal Setup'!B71)*'01 Deal Setup'!B72
C94Extra-cycle engineering$
A95Independent acceptance / handoff
B95Independent acceptance / handoff30,000='01 Deal Setup'!B73
C95Independent acceptance / handoff$
A96Negotiated discount
B96Negotiated discount0=-SUM(B91:B95)*'01 Deal Setup'!B75
C96Negotiated discount$
A97Base customer proposal
B97Base customer proposal210,000=SUM(B91:B96)
C97Base customer proposal$
A98Target-margin price floor
B98Target-margin price floor193,800=B84/MAX(0.01,1-'01 Deal Setup'!B76)
C98Target-margin price floor$
A99Transparent capacity / risk adjustment
B99Transparent capacity / risk adjustment0=MAX(0,B98-B97)
C99Transparent capacity / risk adjustment$
A100SUGGESTED CUSTOMER QUOTE
B100SUGGESTED CUSTOMER QUOTE210,000=SUM(B97,B99)
C100SUGGESTED CUSTOMER QUOTE$
A101Expected deal gross margin
B101Expected deal gross margin0.4924=IF(B100=0,0,1-B84/B100)
C101Expected deal gross margin%
A102Budget ceiling
B102Budget ceiling260,000='01 Deal Setup'!B22
C102Budget ceiling$
A103Remaining approved budget
B103Remaining approved budget50,000=B102-B100
C103Remaining approved budget$
A104Commercial approval status
B104Commercial approval statusHOLD - MODEL=IF(OR('01 Deal Setup'!B43<=0,'01 Deal Setup'!B44<=0,'01 Deal Setup'!B45<=0,'01 Deal Setup'!B98<=0,'01 Deal Setup'!B10<1,'01 Deal Setup'!B10>10,'01 Deal Setup'!B35<0,'01 Deal Setup'!B36<0,'01 Deal Setup'!B37<0,'01 Deal Setup'!B38<0,'01 Deal Setup'!B40<0,'01 Deal Setup'!B82<0,'01 Deal Setup'!B84<0),"HOLD - INPUT",IF('01 Deal Setup'!G30<>"QUALIFIED*","HOLD - MODEL",IF(B70<>"FEASIBLE*","HOLD - CAPACITY",IF(B103<0,"HOLD - BUDGET",IF(B101<'01 Deal Setup'!B77,"HOLD - MARGIN","READY FOR REVIEW")))))
A1097 PRODUCTION INFERENCE - SEPARATE SERVICE ORDER
A111Input + output equivalent tokens / month
B111Input + output equivalent tokens / month20,000,000='01 Deal Setup'!B25*'01 Deal Setup'!B39+'01 Deal Setup'!B26
C111Input + output equivalent tokens / monthequiv tokens
A112Variable serving demand GPU-hours
B112Variable serving demand GPU-hours200=IF(OR('01 Deal Setup'!B98<=0,B6<=0),0,B111/('01 Deal Setup'!B98*B6))
C112Variable serving demand GPU-hoursGPU-hours
A113Warm capacity floor GPU-hours
B113Warm capacity floor GPU-hours1,440='01 Deal Setup'!B28*720*IF('01 Deal Setup'!B24="Shared",'01 Deal Setup'!B99,1)
C113Warm capacity floor GPU-hoursGPU-hours
A114Provisioned / allocated serving GPU-hours
B114Provisioned / allocated serving GPU-hours1,440=MAX(B112,B113)
C114Provisioned / allocated serving GPU-hoursGPU-hours
A115Serving GPU vendor COGS
B115Serving GPU vendor COGS5,745.6=IF(B62="BYOC",0,B114*'01 Deal Setup'!B82)
C115Serving GPU vendor COGS$
A116Expected monthly serving COGS
B116Expected monthly serving COGS8,245.6=B115+'01 Deal Setup'!B97
C116Expected monthly serving COGS$
A117Suggested monthly serving invoice
B117Suggested monthly serving invoice0=IF('01 Deal Setup'!B23="No",0,IF('01 Deal Setup'!B24="Shared",MAX('01 Deal Setup'!B91,MAX(0,'01 Deal Setup'!B25-'01 Deal Setup'!B95)/1000000*'01 Deal Setup'!B93+MAX(0,'01 Deal Setup'!B26-'01 Deal Setup'!B96)/1000000*'01 Deal Setup'!B94),MAX('01 Deal Setup'!B91,'01 Deal Setup'!B28*'01 Deal Setup'!B92)+MAX(0,'01 Deal Setup'!B25-'01 Deal Setup'!B95)/1000000*'01 Deal Setup'!B93+MAX(0,'01 Deal Setup'!B26-'01 Deal Setup'!B96)/1000000*'01 Deal Setup'!B94))
C117Suggested monthly serving invoice$
A118Pre-adjustment serving gross margin
B118Pre-adjustment serving gross margin0=IF(B117=0,0,1-B116/B117)
C118Pre-adjustment serving gross margin%
A119Serving size / SLO validation
B119Serving size / SLO validationNOT SCOPED=IF('01 Deal Setup'!B23="No","NOT SCOPED",IF('01 Deal Setup'!G26<>"Listed","HOLD - SERVING ACCESS",IF('01 Deal Setup'!B101<>"Yes","HOLD - SERVING BENCHMARK",IF('01 Deal Setup'!B27>3,"BENCHMARK PEAK FIRST","BENCHMARK / SLO SIGNOFF"))))
A120Peak vs current warm capacity
B120Peak vs current warm capacity0.4167=IF(B113=0,0,B112*'01 Deal Setup'!B27/B113)
C120Peak vs current warm capacityx
A121Scaling / re-paper recommendation
B121Scaling / re-paper recommendationNOT SCOPED=IF('01 Deal Setup'!B23="No","NOT SCOPED",IF(B120>=1,"AMEND CAPACITY / REPLICAS",IF(B120>=0.8,"START SCALE REVIEW","WITHIN PLANNING ENVELOPE")))
A122Target-margin serving floor
B122Target-margin serving floor0=IF('01 Deal Setup'!B23="No",0,B116/MAX(0.01,1-'01 Deal Setup'!B76))
C122Target-margin serving floor$
A123Visible production uplift
B123Visible production uplift0=MAX(0,B122-B117)
C123Visible production uplift$
A124SUGGESTED MONTHLY SERVING PRICE
B124SUGGESTED MONTHLY SERVING PRICE0=B117+B123
C124SUGGESTED MONTHLY SERVING PRICE$
A125Expected production gross margin
B125Expected production gross margin0=IF(B124=0,0,1-B116/B124)
C125Expected production gross margin%
03 Offer and SOW · 127 cells, 29 formulas
CellLabelSaved valueFormula
A1CUSTOMER OFFER + STATEMENT OF WORK
A2Customer: Example Enterprise | Workflow: Internal workflow automation="Customer: "&'01 Deal Setup'!B5&" | Workflow: "&'01 Deal Setup'!B6
A3PRELIMINARY - DO NOT SEND CUSTOMER: HOLD - MODEL=IF('02 Economics Capacity'!B104="READY FOR REVIEW","DRAFT - commercial/legal review still required","PRELIMINARY - DO NOT SEND CUSTOMER: "&'02 Economics Capacity'!B104)
A4A COMMERCIAL ORDER: WHAT THE CUSTOMER PAYS
A5Paid service
B5Paid serviceQuantity / unit
C5Paid servicePrice ($)
D5Paid serviceCustomer receives
A6Readiness and baseline
B6Readiness and baselineOnce
C6Readiness and baseline30,000='02 Economics Capacity'!B91
D6Readiness and baselineWorkflow definition, model/eval plan, baseline score and feasibility gate
A7Engineering + integration
B7Engineering + integrationFixed engagement
C7Engineering + integration78,000='02 Economics Capacity'!B92
D7Engineering + integrationHarness integration, grader implementation, research and infra support
A8Approved RL cycles
B8Approved RL cyclesCycles x proposed model rate
C8Approved RL cycles72,000='02 Economics Capacity'!B93
D8Approved RL cyclesScheduled train / rollout / measure / review cycles
A9Extra-cycle engineering
B9Extra-cycle engineeringAdditional cycles only
C9Extra-cycle engineering0='02 Economics Capacity'!B94
D9Extra-cycle engineeringIncremental assigned engineers above included loop count
A10Final independent verification
B10Final independent verificationOnce
C10Final independent verification30,000='02 Economics Capacity'!B95
D10Final independent verificationHoldout evaluation, checkpoint, report and handoff
A11Discount
B11DiscountOnce
C11Discount0='02 Economics Capacity'!B96
D11DiscountExplicit negotiated discount
A12Proposed price adjustment*
B12Proposed price adjustment*If needed
C12Proposed price adjustment*0='02 Economics Capacity'!B99
D12Proposed price adjustment**Requires explicit customer agreement; not an automatic surcharge.
A13TOTAL PROPOSED TRAINING FEE
C13TOTAL PROPOSED TRAINING FEE210,000='02 Economics Capacity'!B100
A15Selected model
B15Selected modelQwen3.6-35B-A3B='01 Deal Setup'!B7
D15Selected modelTotal params: 35.0B; MoE. Exact hardware requires benchmark.="Total params: "&IF(ISNUMBER('01 Deal Setup'!B8),TEXT('01 Deal Setup'!B8,"#,##0.0")&"B","unverified")&"; "&'01 Deal Setup'!G24&". Exact hardware requires benchmark."
A16Authorized RL cycles
B16Authorized RL cycles3='01 Deal Setup'!B10
A17Cycle schedule
B17Cycle scheduleEvery 1 week(s)="Every "&'01 Deal Setup'!B11&" week(s)"
A18Planned duration
B18Planned duration6='02 Economics Capacity'!B28
C18Planned durationweeks
A19Production inference
B19Production inferenceNOT INCLUDED=IF('01 Deal Setup'!B23="Yes","OPTIONAL SEPARATE ORDER","NOT INCLUDED")
A21B OPTIONAL SERVING COMMERCIALS (NOT IN TRAINING TOTAL)
A22Service type
B22Service typeDedicated='01 Deal Setup'!B24
D22Service typeShared usage minimum or dedicated replica commitment
A23Monthly serving proposal
B23Monthly serving proposal0='02 Economics Capacity'!B124
D23Monthly serving proposalZero when production option is No
A24Monthly input allowance
B24Monthly input allowance20,000,000='01 Deal Setup'!B95
D24Monthly input allowanceExcess charged per million input tokens
A25Monthly output allowance
B25Monthly output allowance10,000,000='01 Deal Setup'!B96
D25Monthly output allowanceExcess charged per million output tokens
A26Peak / replica assumption
B26Peak / replica assumption2='01 Deal Setup'!B28
D26Peak / replica assumptionReplicas and peak traffic must be benchmarked
A27Serving next step
B27Serving next stepNOT SCOPED='02 Economics Capacity'!B121
D27Serving next stepReview at 80% of capacity; amendment to add GPUs/replicas
A28Serving capacity adjustment
B28Serving capacity adjustment0='02 Economics Capacity'!B123
D28Serving capacity adjustmentAny added committed capacity requires a separately approved order.
A29Serving performance gate
B29Serving performance gateNOT SCOPED='02 Economics Capacity'!B119
D29Serving performance gateLoad test and service levels must be approved before contracting.
A30C SCOPE: RESPONSIBILITY, ACCEPTANCE & PHASE GATES
A31Phase + timing
B31Phase + timingApplied Compute delivers
C31Phase + timingCustomer owns / approves
D31Phase + timingEvidence to accept
A320 · Readiness
B320 · ReadinessScope workflow, assess data quality, compare base model options and estimate capacity.
C320 · ReadinessProvides data, SME and decides initial base model and success metric.
D320 · ReadinessSigned workflow/eval charter; dataset sample; frozen baseline score and review findings.
A331 · Graders + harness
B331 · Graders + harnessBuild environment, grader version, rollout instrumentation and held-out evaluation.
C331 · Graders + harnessApproves graders, thresholds, privacy assumptions and success/stop rules.
D331 · Graders + harnessVersioned grader/harness, reproducible baseline run ID, acceptance signoff.
A342 · RL cycle (repeat)
B342 · RL cycle (repeat)Train checkpoint; sample rollouts; compute cost, reward and held-out score; investigate failures.
C342 · RL cycle (repeat)Reviews score and research evidence; explicitly chooses Stop or Continue.
D342 · RL cycle (repeat)Checkpoint/run ID, GPU and token ledger, graded metrics, failure samples, signed cycle decision.
A353 · Independent handoff
B353 · Independent handoffEvaluate untouched holdout; package checkpoint/weights, traces, config, results and runbook.
C353 · Independent handoffAccepts delivery by checking objective artifacts, not guaranteed quality uplift.
D353 · Independent handoffReproducibility report, model registry/version, delivered files and signed handoff.
A364 · Inference pilot (opt.)
B364 · Inference pilot (opt.)Benchmark representative traffic, SLO throughput, warm replicas, cost and autoscale policy.
C364 · Inference pilot (opt.)Approves traffic forecast, deployment posture, SLA and commercial service order.
D364 · Inference pilot (opt.)Load-test results, capacity envelope and approved serving order before production.
A39D STOPPING RULES + DECISION AUTHORITY
A40Success score required
B40Success score required0.8='01 Deal Setup'!B17
D40Success score requiredA score threshold, plateau, cycle cap or budget limit triggers a review; only the customer authorizes continuing/ending work.
A41Plateau trigger
B41Plateau trigger<1.0% gain for 2 cycles="<"&TEXT('01 Deal Setup'!B18,"0.0%")&" gain for "&'01 Deal Setup'!B19&" cycles"
A42Max preauthorized cycles
B42Max preauthorized cycles3='01 Deal Setup'!B10
A43Current review status
B43Current review statusNO RESULTS YET='01 Deal Setup'!G18
A44Formal decision owner
B44Formal decision ownerCustomer product / SME lead='01 Deal Setup'!B30
A47E AMENDMENTS AND SCALE TRIGGERS
A48New base model or method
B48New base model or methodRe-benchmark model fit, memory, unit rates, training throughput and evaluator comparability.
A49More cycles / higher token volume
B49More cycles / higher token volumeWritten change order before exceeding the approved cycle count or compute allowance.
A50New workflow, data regime or grader
B50New workflow, data regime or graderRe-scope data rights, acceptance benchmarks, engineering and deadline.
A51Different region, isolation or GPU topology
B51Different region, isolation or GPU topologyReassess availability, security posture, commitments and price.
A52Production launch or SLA upgrade
B52Production launch or SLA upgradeSeparate serving order with p95/p99 target, traffic envelope, availability, and minimums.
A53Production demand exceeds contracted envelope
B53Production demand exceeds contracted envelopeReview at 80% capacity; amendment before adding replicas, GPUs or >20% contracted traffic / concurrency.
A54Customer pause / termination
B54Customer pause / terminationCustomer covers approved work and uncancelable new supplier commitments, net of reasonable redeployment.
04 Guide Assumptions · 532 cells, 0 formulas
CellLabelSaved valueFormula
A1USER GUIDE + ASSUMPTION REGISTER
A2The spreadsheet is an editable prototype of a deal-underwriting product; assumptions are explicitly illustrative, not disclosed company rates.
A4HOW TO USE (UNDER 5 MINUTES)
A51 · Start
B51 · Start01 Deal Setup
C51 · StartEdit orange/blue inputs for customer, model, cycles, demand and stop rules.
A62 · Capacity
B62 · Capacity01 Deal Setup rows 78-111
C62 · CapacityEnter usable GPUs per contract, valid dates (week numbers), and acquisition choice.
A73 · Actual evals
B73 · Actual evals01 Deal Setup score log
C73 · Actual evalsEnter scores in order, then record the customer's Stop/Continue decision. This updates guidance, NOT the signed commercial commitment.
A84 · Economics
B84 · Economics02 Economics Capacity
C84 · EconomicsReview GPU-hours, weekly shortage, cash/allocated/opportunity cost and margin.
A95 · Client documents
B95 · Client documents03 Offer and SOW
C95 · Client documentsUse the offer + SOW as review drafts. If the customer stops early, any contractual credit or re-scope requires mutual approval.
A106 · Decision
B106 · Decision01 Deal Setup summary
C106 · DecisionDo not send pricing if HOLD, supplier unconfirmed, or quality/throughput unbenchmarked.
A12KEY MODEL EQUATIONS (SIMPLIFIED)
A13RL rollout work
B13RL rollout workAttempts x (input tokens x input weight + output tokens)
D13RL rollout workEquivalent tokens, not raw token cost
A14GPU-hours
B14GPU-hours(training + rollout + eval work) / separate measured GPU throughput
C14GPU-hoursTraining, rollouts and each-cycle evaluations use separate benchmark throughput.
D14GPU-hoursModel-size performance affects hours
A15Required capacity
B15Required capacityPeak GPUs x week availability; sequential jobs reduce peaks
D15Required capacityGPU-hours alone do not prove feasibility
A16Incremental supplier cost
B16Incremental supplier costCheapest eligible PayGo / extension / reservation on uncovered capacity
D16Incremental supplier costReserved contracts charge for idle time
A17Risk-adjusted delivery COGS
B17Risk-adjusted delivery COGSLabor + external graders + storage + platform + allocated GPU + new commitment + expected rework
D17Risk-adjusted delivery COGSOpportunity cost shown separately
A18Suggested customer quote
B18Suggested customer quoteMAX(itemized fees, forecast COGS / (1-target margin))
C18Suggested customer quoteUplift is a proposed revised fee; client must approve before billing.
D18Suggested customer quoteClient offer row 12 explicitly requires written approval.
A19Production COGS
B19Production COGSMAX(token work / SLO throughput, warm-replica floor) x GPU rate + direct support
D19Production COGSRequires benchmark; separate service agreement
A22ASSUMPTIONS - DEFAULT, JUSTIFICATION AND REQUIRED EVIDENCE
F22ASSUMPTIONS - DEFAULT, JUSTIFICATION AND REQUIRED EVIDENCEDOCUMENTED MODEL OPTIONS | 2026-10-18 KB SNAPSHOT + PUBLIC MODEL SPECS (NOT LIVE AC2 INVENTORY)
A23Inputs / cells
B23Inputs / cellsDefault / key assumption
C23Inputs / cellsWhy used
D23Inputs / cellsReplace with / source
A24Model + sizes · B7/B8
B24Model + sizes · B7/B8Named models, total/active B, eligibility
C24Model + sizes · B7/B8Parameters are factual when sourced; class factor is not a performance benchmark.
D24Model + sizes · B7/B8Guide F:Q has model-by-model provenance; AC2 sources list named models, not exhaustive live support.
F24Model + sizes · B7/B8Model selection
G24Model + sizes · B7/B8Total params (B)
H24Model + sizes · B7/B8Active params (B)
I24Model + sizes · B7/B8Architecture
J24Model + sizes · B7/B8RL training
K24Model + sizes · B7/B8Serving
L24Model + sizes · B7/B8Evidence
M24Model + sizes · B7/B8Proxy class
N24Model + sizes · B7/B8HW to qualify
O24Model + sizes · B7/B8Parameters source / URL
P24Model + sizes · B7/B8AC2 catalog / URL
Q24Model + sizes · B7/B8Qualification note
A25Method · B9
B25Method · B9GRPO (other methods selectable; require revised COGS before approval)
C25Method · B9The included rollout/policy/eval formulas describe GRPO only; non-GRPO cannot be quoted from unchanged assumptions.
D25Method · B9Document the selected method, optimizer, workloads and benchmarked GPU efficiency; then approve revised cost model.
A26Cycles/cadence · B10:B14
B26Cycles/cadence · B10:B143 weekly; 2-week readiness; 1-week handoff
C26Cycles/cadence · B10:B14Bounded change-control program; weekly units
D26Cycles/cadence · B10:B14Approved customer timeline and env/runtime plan
F26Cycles/cadence · B10:B14Qwen3.6-35B-A3B
G26Cycles/cadence · B10:B1435
H26Cycles/cadence · B10:B143
I26Cycles/cadence · B10:B14MoE
J26Cycles/cadence · B10:B14Listed
K26Cycles/cadence · B10:B14Listed
L26Cycles/cadence · B10:B14Model-named
M26Cycles/cadence · B10:B142
N26Cycles/cadence · B10:B14Validate
O26Cycles/cadence · B10:B14https://huggingface.co/Qwen/Qwen3.6-35B-A3B
P26Cycles/cadence · B10:B14https://docs.appliedcompute.com/platform/training/models
Q26Cycles/cadence · B10:B14Training & serving named; illustrative benchmark only
A27Stop/go · B15:B21
B27Stop/go · B15:B21Normalized 0–1, higher-is-better score; 60% baseline / 80% target / 1 pp plateau
C27Stop/go · B15:B21Permits multiple metric labels without changing stop rules; lower-is-better metrics need a model redesign.
D27Stop/go · B15:B21Frozen held-out dataset, signed rubric, human decision
F27Stop/go · B15:B21Qwen3.8-27B
G27Stop/go · B15:B2127
H27Stop/go · B15:B2127
I27Stop/go · B15:B21Dense
J27Stop/go · B15:B21Listed
K27Stop/go · B15:B21Listed
L27Stop/go · B15:B21Model-named
M27Stop/go · B15:B212
N27Stop/go · B15:B21Validate
O27Stop/go · B15:B21https://www.alibabacloud.com/help/en/model-studio/newly-released-models
P27Stop/go · B15:B21https://docs.appliedcompute.com/platform/training/models
Q27Stop/go · B15:B21Text model; compare vision model license separately
A28Deal cap · B22
B28Deal cap · B22$260k
C28Deal cap · B22Illustrative customer approval budget
D28Deal cap · B22Confirmed purchasing authority
F28Deal cap · B22Qwen3.5-4B
G28Deal cap · B224
H28Deal cap · B224
I28Deal cap · B22Dense
J28Deal cap · B22Family - verify
K28Deal cap · B22Not listed
L28Deal cap · B22Family-only
M28Deal cap · B221
N28Deal cap · B22Validate
O28Deal cap · B22https://huggingface.co/Qwen
P28Deal cap · B22https://docs.appliedcompute.com/platform/training/models
Q28Deal cap · B22Exact training catalog entry not established
A29Production · B23:B29
B29Production · B23:B29Optional; 20M in/10M out; 2 warm replicas
C29Production · B23:B29Shows fixed availability cost even with low tokens
D29Production · B23:B29Traffic traces, latency SLO, load test
F29Production · B23:B29Qwen3.5-9B
G29Production · B23:B299
H29Production · B23:B299
I29Production · B23:B29Dense
J29Production · B23:B29Family - verify
K29Production · B23:B29Not listed
L29Production · B23:B29Family-only
M29Production · B23:B291
N29Production · B23:B29Validate
O29Production · B23:B29https://huggingface.co/Qwen
P29Production · B23:B29https://docs.appliedcompute.com/platform/training/models
Q29Production · B23:B29Exact training catalog entry not established
A30Policy tokens · B35
B30Policy tokens · B3542.67M / cycle
C30Policy tokens · B35Illustrative 128M total at 3 cycles
D30Policy tokens · B35Measured training profiler and step configuration
F30Policy tokens · B35Qwen3.5-27B
G30Policy tokens · B3527
H30Policy tokens · B3527
I30Policy tokens · B35Dense
J30Policy tokens · B35Family - verify
K30Policy tokens · B35Not listed
L30Policy tokens · B35Family-only
M30Policy tokens · B352
N30Policy tokens · B35Validate
O30Policy tokens · B35https://huggingface.co/Qwen
P30Policy tokens · B35https://docs.appliedcompute.com/platform/training/models
Q30Policy tokens · B35Exact training catalog entry not established
A31Rollout demand · B36:B39
B31Rollout demand · B36:B394k attempts; 12k in/10k out; prefill 0.5
C31Rollout demand · B36:B39Covers agentic token shape and asynchronous rollout work
D31Rollout demand · B36:B39Representative task traces; benchmark engine throughput
F31Rollout demand · B36:B39Qwen3.5-35B-A3B
G31Rollout demand · B36:B3935
H31Rollout demand · B36:B393
I31Rollout demand · B36:B39MoE
J31Rollout demand · B36:B39Family - verify
K31Rollout demand · B36:B39Not listed
L31Rollout demand · B36:B39Family-only
M31Rollout demand · B36:B392
N31Rollout demand · B36:B39Validate
O31Rollout demand · B36:B39https://huggingface.co/Qwen
P31Rollout demand · B36:B39https://docs.appliedcompute.com/platform/training/models
Q31Rollout demand · B36:B39Exact training catalog entry not established
A32Eval demand · B40:B42
B32Eval demand · B40:B42200 cases/cycle + 200 baseline; 20k equiv tokens
C32Eval demand · B40:B42Recurring tests run after each RL cycle; baseline test precedes training.
D32Eval demand · B40:B42Grader time, API tokens, eval profile
F32Eval demand · B40:B42Qwen3.5-122B-A10B
G32Eval demand · B40:B42122
H32Eval demand · B40:B4210
I32Eval demand · B40:B42MoE
J32Eval demand · B40:B42Family - verify
K32Eval demand · B40:B42Not listed
L32Eval demand · B40:B42Family-only
M32Eval demand · B40:B423
N32Eval demand · B40:B42Multi-node
O32Eval demand · B40:B42https://huggingface.co/Qwen
P32Eval demand · B40:B42https://docs.appliedcompute.com/platform/training/models
Q32Eval demand · B40:B42Large model: topology/throughput must be benchmarked
A33Throughput · B43:B48
B33Throughput · B43:B48100k work tokens/GPU-hour; 8/8/4 GPUs
C33Throughput · B43:B48Backsolves an illustrative 3,360 GPUh; NOT an AC measured rate
D33Throughput · B43:B48AC model+SKU-specific benchmark, cluster topology
F33Throughput · B43:B48Qwen3.5-397B-A17B
G33Throughput · B43:B48397
H33Throughput · B43:B4817
I33Throughput · B43:B48MoE
J33Throughput · B43:B48Family - verify
K33Throughput · B43:B48Not listed
L33Throughput · B43:B48Family-only
M33Throughput · B43:B483
N33Throughput · B43:B48Multi-node
O33Throughput · B43:B48https://huggingface.co/Qwen
P33Throughput · B43:B48https://docs.appliedcompute.com/platform/training/models
Q33Throughput · B43:B48Large model: topology/throughput must be benchmarked
A34Engineer effort/rates · B50:B61
B34Engineer effort/rates · B50:B61FDE/ML/Infra/PM hours; $150-$200/hr
C34Engineer effort/rates · B50:B61Fully burdened, directly attributable delivery cost
D34Engineer effort/rates · B50:B61Staffing plan, finance labor rates
F34Engineer effort/rates · B50:B61Qwen3.5-9B-VL
G34Engineer effort/rates · B50:B619
H34Engineer effort/rates · B50:B619
I34Engineer effort/rates · B50:B61Dense
J34Engineer effort/rates · B50:B61Listed
K34Engineer effort/rates · B50:B61Not listed
L34Engineer effort/rates · B50:B61Model-named
M34Engineer effort/rates · B50:B611
N34Engineer effort/rates · B50:B61Validate
O34Engineer effort/rates · B50:B61https://docs.appliedcompute.com/platform/training/vlm
P34Engineer effort/rates · B50:B61https://docs.appliedcompute.com/platform/training/vlm
Q34Engineer effort/rates · B50:B61Vision inputs add image/encoder cost not captured by token-only scenario
A35External costs · B62:B66
B35External costs · B62:B66$1.5k fixed + $1.5k/cycle; $1.5k storage; $4k support; 10% rework
C35External costs · B62:B66Avoids GPU-only undercount and prices uncertain rework
D35External costs · B62:B66Vendor API invoices, object storage, historical delivery
F35External costs · B62:B66Qwen3.6-35B-A3B-VL
G35External costs · B62:B6635
H35External costs · B62:B663
I35External costs · B62:B66MoE
J35External costs · B62:B66Listed
K35External costs · B62:B66Not listed
L35External costs · B62:B66Model-named
M35External costs · B62:B662
N35External costs · B62:B66Validate
O35External costs · B62:B66https://docs.appliedcompute.com/platform/training/vlm
P35External costs · B62:B66https://docs.appliedcompute.com/platform/training/vlm
Q35External costs · B62:B66Vision workload must be separately benchmarked
A36Quote rate card · B68:B75
B36Quote rate card · B68:B75$30k readiness; $78k eng; $24k/cycle; $30k handoff
C36Quote rate card · B68:B75Illustrative itemized fee totaling $210k for base 3 cycles
D36Quote rate card · B68:B75Approved commercial rate card + customer proposal
F36Quote rate card · B68:B75Qwen3.8-27B-VL
G36Quote rate card · B68:B7527
H36Quote rate card · B68:B7527
I36Quote rate card · B68:B75Dense
J36Quote rate card · B68:B75Listed
K36Quote rate card · B68:B75Not listed
L36Quote rate card · B68:B75Model-named
M36Quote rate card · B68:B752
N36Quote rate card · B68:B75Validate
O36Quote rate card · B68:B75https://docs.appliedcompute.com/platform/training/vlm
P36Quote rate card · B68:B75https://docs.appliedcompute.com/platform/training/vlm
Q36Quote rate card · B68:B75Vision workload must be separately benchmarked
A37Target margin · B76:B77
B37Target margin · B76:B7745% target; 35% warning floor
C37Target margin · B76:B77Explicit quote guardrail, not company-published policy
D37Target margin · B76:B77Internal policy, win/loss and expansion economics
F37Target margin · B76:B77Qwen3.8-Flash-Next
G37Target margin · B76:B77125
H37Target margin · B76:B776
I37Target margin · B76:B77MoE
J37Target margin · B76:B77Not listed
K37Target margin · B76:B77Listed
L37Target margin · B76:B77Model-named
M37Target margin · B76:B773
N37Target margin · B76:B77Multi-node
O37Target margin · B76:B77https://www.chels.ai/models
P37Target margin · B76:B77https://docs.appliedcompute.com/platform/serving/models
Q37Target margin · B76:B77Serving named, training NOT confirmed; parameters provisional secondary
A38Existing supply · B79,B109:D111
B38Existing supply · B79,B109:D1118 compatible spare GPUs weeks 1-8
C38Existing supply · B79,B109:D111Shows allocation economics instead of assuming all PayGo
D38Existing supply · B79,B109:D111Actual inventory/region/expiry/other demand
F38Existing supply · B79,B109:D111Qwen3-4B
G38Existing supply · B79,B109:D1114
H38Existing supply · B79,B109:D1114
I38Existing supply · B79,B109:D111Dense
J38Existing supply · B79,B109:D111Research only
K38Existing supply · B79,B109:D111Not listed
L38Existing supply · B79,B109:D111Research-only
M38Existing supply · B79,B109:D1111
N38Existing supply · B79,B109:D111Validate
O38Existing supply · B79,B109:D111https://www.appliedcompute.com/research/neural-cheat-sheets-learning-to-summarize-with-reinforcement-learning
P38Existing supply · B79,B109:D111See primary source
Q38Existing supply · B79,B109:D111AC research use does not establish current self-serve listing
A39Supply availability · B81
B39Supply availability · B81No (unverified)
C39Supply availability · B81Allows scenario calculations; not an actual quota or reservation
D39Supply availability · B81Cloud provider availability confirmation
F39Supply availability · B81Qwen3-8B
G39Supply availability · B818
H39Supply availability · B818
I39Supply availability · B81Dense
J39Supply availability · B81Research only
K39Supply availability · B81Not listed
L39Supply availability · B81Research-only
M39Supply availability · B811
N39Supply availability · B81Validate
O39Supply availability · B81https://www.appliedcompute.com/research/staleness-in-fully-async-rl
P39Supply availability · B81See primary source
Q39Supply availability · B81Research only; requires availability/access check
A40GPU prices · B82:B85
B40GPU prices · B82:B85$3.99 PayGo; $2.75 extend; $2.50 reserve; 4 weeks
C40GPU prices · B82:B85Contrasts flexibility vs stranded commitment
D40GPU prices · B82:B85Provider quote; https://lambda.ai/pricing is proxy only
F40GPU prices · B82:B85Kimi K3
G40GPU prices · B82:B852,800
H40GPU prices · B82:B85104
I40GPU prices · B82:B85MoE
J40GPU prices · B82:B85Listed
K40GPU prices · B82:B85Listed
L40GPU prices · B82:B85Model-named
M40GPU prices · B82:B853
N40GPU prices · B82:B85B300
O40GPU prices · B82:B85https://arxiv.org/abs/2607.24653
P40GPU prices · B82:B85https://docs.appliedcompute.com/platform/training/models
Q40GPU prices · B82:B85Serving catalog requires BYO weights; AC research uses B300 multi-node
A41Allocation/opportunity · B87:B88
B41Allocation/opportunity · B87:B88$2.50 allocation; $0 opp
C41Allocation/opportunity · B87:B88Keeps accounting cost separate from scarce capacity decisions
D41Allocation/opportunity · B87:B88Finance allocation + real alternative deal contribution
F41Allocation/opportunity · B87:B88Kimi K2.7 Code
G41Allocation/opportunity · B87:B881,000
H41Allocation/opportunity · B87:B8832
I41Allocation/opportunity · B87:B88MoE
J41Allocation/opportunity · B87:B88Listed
K41Allocation/opportunity · B87:B88Not listed
L41Allocation/opportunity · B87:B88Model-named
M41Allocation/opportunity · B87:B883
N41Allocation/opportunity · B87:B88Multi-node
O41Allocation/opportunity · B87:B88https://www.kimi.ai/resources/kimi-k2-7-code
P41Allocation/opportunity · B87:B88https://docs.appliedcompute.com/platform/training/models
Q41Allocation/opportunity · B87:B88Large multi-node model; verify license and serving eligibility
A42Serving charges · B91:B99
B42Serving charges · B91:B99$15k min; $6k warm replica; per-token overage
C42Serving charges · B91:B99Monthly price also has a visible target-margin safeguard; not official AC rates
D42Serving charges · B91:B99Load tests, infra billing, SLA and commercial approval
F42Serving charges · B91:B99Kimi K2.6
I42Serving charges · B91:B99MoE
J42Serving charges · B91:B99Listed
K42Serving charges · B91:B99Not listed
L42Serving charges · B91:B99Model-named
M42Serving charges · B91:B993
N42Serving charges · B91:B99Multi-node
O42Serving charges · B91:B99https://docs.appliedcompute.com/platform/training/models
P42Serving charges · B91:B99https://docs.appliedcompute.com/platform/training/models
Q42Serving charges · B91:B99Precise weights/active count unverified in supplied material
A43Existing pool compatibility · F109:F111
B43Existing pool compatibility · F109:F111Compatible
C43Existing pool compatibility · F109:F111Only models qualified capacity; no mixed/incompatible clusters
D43Existing pool compatibility · F109:F111SKU, interconnect, region, security verification
F43Existing pool compatibility · F109:F111GLM-5.3
G43Existing pool compatibility · F109:F111744
H43Existing pool compatibility · F109:F11140
I43Existing pool compatibility · F109:F111MoE
J43Existing pool compatibility · F109:F111Listed
K43Existing pool compatibility · F109:F111Listed
L43Existing pool compatibility · F109:F111Model-named
M43Existing pool compatibility · F109:F1113
N43Existing pool compatibility · F109:F111Multi-node
O43Existing pool compatibility · F109:F111https://rohitghumare.com/guides/glm-5-3/
P43Existing pool compatibility · F109:F111https://docs.appliedcompute.com/platform/training/models
Q43Existing pool compatibility · F109:F111Verify Z.ai first-party card before binding quote
A44Model + hardware qualification
B44Model + hardware qualificationB7 model; B8 auto params; B101:B103 approval gates; G109:G111 actual GPU family
C44Model + hardware qualificationAvoid pricing giant MoE/vision models as 32B dense hardware. No unverified quote can be READY.
D44Model + hardware qualificationRequire live AC2 access, license review, actual GPU topology and workload benchmarks.
F44Model + hardware qualificationGLM-5.2
G44Model + hardware qualification744
H44Model + hardware qualification40
I44Model + hardware qualificationMoE
J44Model + hardware qualificationListed
K44Model + hardware qualificationNot listed
L44Model + hardware qualificationModel-named
M44Model + hardware qualification3
N44Model + hardware qualificationMulti-node
O44Model + hardware qualificationhttps://www.chels.ai/models
P44Model + hardware qualificationhttps://docs.appliedcompute.com/platform/training/models
Q44Model + hardware qualificationParameter count secondary; confirm model card
F45GLM-5.3-Flash
I45GLM-5.3-FlashUnknown
J45GLM-5.3-FlashNot listed
K45GLM-5.3-FlashListed
L45GLM-5.3-FlashModel-named
M45GLM-5.3-Flash3
N45GLM-5.3-FlashValidate
O45GLM-5.3-Flashhttps://docs.appliedcompute.com/platform/serving/models
P45GLM-5.3-Flashhttps://docs.appliedcompute.com/platform/serving/models
Q45GLM-5.3-FlashNo reliable parameter count in supplied KB; do not fill with guess
A46WHAT THIS MVP DOES NOT AUTOMATE (IMPORTANT)
F46WHAT THIS MVP DOES NOT AUTOMATE (IMPORTANT)GLM-5.3-Flash-VL
I46WHAT THIS MVP DOES NOT AUTOMATE (IMPORTANT)Unknown
J46WHAT THIS MVP DOES NOT AUTOMATE (IMPORTANT)Listed
K46WHAT THIS MVP DOES NOT AUTOMATE (IMPORTANT)Not listed
L46WHAT THIS MVP DOES NOT AUTOMATE (IMPORTANT)Model-named
M46WHAT THIS MVP DOES NOT AUTOMATE (IMPORTANT)3
N46WHAT THIS MVP DOES NOT AUTOMATE (IMPORTANT)Validate
O46WHAT THIS MVP DOES NOT AUTOMATE (IMPORTANT)https://docs.appliedcompute.com/platform/training/vlm
P46WHAT THIS MVP DOES NOT AUTOMATE (IMPORTANT)https://docs.appliedcompute.com/platform/training/vlm
Q46WHAT THIS MVP DOES NOT AUTOMATE (IMPORTANT)Vision throughput and weights require explicit benchmark
A47Model catalog coverage
B47Model catalog coverage30 documented examples; live AC2 catalog is NOT exhaustively verified
C47Model catalog coverageTraining and serving lists differ; family names and research examples aren't confirmed product access.
D47Model catalog coverageReconcile guide F:Q with live AC2 catalog and customer license before quoting.
F47Model catalog coverageGLM-5.1
I47Model catalog coverageUnknown
J47Model catalog coverageResearch only
K47Model catalog coverageNot listed
L47Model catalog coverageResearch-only
M47Model catalog coverage3
N47Model catalog coverageValidate
O47Model catalog coveragehttps://www.appliedcompute.com/case-studies/harvey
P47Model catalog coverageSee primary source
Q47Model catalog coverageHarvey case study, not verified current catalog entry
A48Parameter interpretation
B48Parameter interpretationTotal B != active B != throughput; MoE and VLM need special care
C48Parameter interpretationTotal B approximates weight storage; active B affects per-token ops; neither guarantees training GPU cost.
D48Parameter interpretationModel supplier cards for parameters, AC2 benchmark for costs, validated GPU topology.
F48Parameter interpretationNemotron 3 Ultra
G48Parameter interpretation550
H48Parameter interpretation55
I48Parameter interpretationMoE
J48Parameter interpretationListed
K48Parameter interpretationListed
L48Parameter interpretationModel-named
M48Parameter interpretation3
N48Parameter interpretationMulti-node
O48Parameter interpretationhttps://github.com/NVIDIA-NeMo/Nemotron/blob/main/usage-cookbook/Nemotron-3-Ultra-Base/README.md
P48Parameter interpretationhttps://docs.appliedcompute.com/platform/training/models
Q48Parameter interpretationLarge model; topology, support, and licensing review
A49Model pricing proxies
B49Model pricing proxiesPlanning class 1/2/3 with 2.5x/1x/0.5x throughput
C49Model pricing proxiesUnverified demo scenario only; not derived from true MoE FLOPs, activation memory, or GPU kernel efficiency.
D49Model pricing proxiesBlock deal approval until B101/B102/B104 and GPU family B103 reviewed.
F49Model pricing proxiesNemotron 3.5 Lightning
G49Model pricing proxies30
H49Model pricing proxies3
I49Model pricing proxiesMoE
J49Model pricing proxiesNot listed
K49Model pricing proxiesListed
L49Model pricing proxiesModel-named
M49Model pricing proxies2
N49Model pricing proxiesValidate
O49Model pricing proxieshttps://github.com/NVIDIA-NeMo/Nemotron/blob/main/docs/nemotron/lightning35/README.md
P49Model pricing proxieshttps://docs.appliedcompute.com/platform/serving/models
Q49Model pricing proxiesServing only in the KB snapshot
A50Capacity by GPU SKU
B50Capacity by GPU SKUPool G109:G111 must match planned GPU type B103
C50Capacity by GPU SKUOld workbook counted any manually compatible pool; now checks region/fit and GPU family.
D50Capacity by GPU SKUCheck topology/interconnect and provider quota; week-level counts still approximate.
F50Capacity by GPU SKUNemotron 3 Super
I50Capacity by GPU SKUUnknown
J50Capacity by GPU SKUNot listed
K50Capacity by GPU SKULegacy mention
L50Capacity by GPU SKULegacy-only
M50Capacity by GPU SKU3
N50Capacity by GPU SKUValidate
O50Capacity by GPU SKUhttps://www.appliedcompute.com/old-home
P50Capacity by GPU SKUSee primary source
Q50Capacity by GPU SKUMentioned in older homepage only; not in current list
A51Production model availability
B51Production model availabilityServing option only when selected model listed, benchmarked
C51Production model availabilitySome named models have serving-only support; AC2 training catalog differs.
D51Production model availabilitySeparate serving order, validated load test and SLO commitment.
F51Production model availabilityMiniMax M2.7
G51Production model availability230
H51Production model availability10
I51Production model availabilityMoE
J51Production model availabilityListed
K51Production model availabilityNot listed
L51Production model availabilityModel-named
M51Production model availability3
N51Production model availabilityMulti-node
O51Production model availabilityhttps://github.com/vllm-project/recipes/blob/main/models/MiniMaxAI/MiniMax-M2.7.yaml
P51Production model availabilityhttps://docs.appliedcompute.com/platform/training/models
Q51Production model availabilitySupplier serving hardware may not match training
F52DeepSeek-V4.1-Flash
G52DeepSeek-V4.1-Flash552
H52DeepSeek-V4.1-Flash8 in / 16 out
I52DeepSeek-V4.1-FlashMoE
J52DeepSeek-V4.1-FlashNot listed
K52DeepSeek-V4.1-FlashListed
L52DeepSeek-V4.1-FlashModel-named
M52DeepSeek-V4.1-Flash3
N52DeepSeek-V4.1-FlashMulti-node
O52DeepSeek-V4.1-Flashhttps://deepseek.com/en/news/deepseek-v4-1-flash/
P52DeepSeek-V4.1-Flashhttps://docs.appliedcompute.com/platform/serving/models
Q52DeepSeek-V4.1-FlashInput/output activate different parameter counts
F53GPT-OSS 120B
G53GPT-OSS 120B116.8
H53GPT-OSS 120B5.1
I53GPT-OSS 120BMoE
J53GPT-OSS 120BNot listed
K53GPT-OSS 120BListed
L53GPT-OSS 120BModel-named
M53GPT-OSS 120B3
N53GPT-OSS 120BValidate
O53GPT-OSS 120Bhttps://openai.com/index/introducing-gpt-oss/
P53GPT-OSS 120Bhttps://docs.appliedcompute.com/platform/serving/models
Q53GPT-OSS 120BInference support documented; RL training not documented
A54DROPDOWN INPUTS | CHOOSE PREDEFINED OPTIONS
F54DROPDOWN INPUTS | CHOOSE PREDEFINED OPTIONSpplx-decider-v1-27b
G54DROPDOWN INPUTS | CHOOSE PREDEFINED OPTIONS27
I54DROPDOWN INPUTS | CHOOSE PREDEFINED OPTIONSDecision head
J54DROPDOWN INPUTS | CHOOSE PREDEFINED OPTIONSAccess required
K54DROPDOWN INPUTS | CHOOSE PREDEFINED OPTIONSNot listed
L54DROPDOWN INPUTS | CHOOSE PREDEFINED OPTIONSAccess-gated
M54DROPDOWN INPUTS | CHOOSE PREDEFINED OPTIONS2
N54DROPDOWN INPUTS | CHOOSE PREDEFINED OPTIONSValidate
O54DROPDOWN INPUTS | CHOOSE PREDEFINED OPTIONShttps://docs.appliedcompute.com/platform/training/decisions
P54DROPDOWN INPUTS | CHOOSE PREDEFINED OPTIONShttps://docs.appliedcompute.com/platform/training/models
Q54DROPDOWN INPUTS | CHOOSE PREDEFINED OPTIONSDecision model training requires AC access
A55Training method · B9
B55Training method · B9GRPO / SFT / OPSD / RMSD / Decision / Custom
C55Training method · B9Only GRPO has current cost logic; other choices HOLD the quote.
D55Training method · B9Confirm method-specific capacity and price before approval.
F55Training method · B9Other open-weight model (qualify)
I55Training method · B9Unknown
J55Training method · B9Not listed
K55Training method · B9Not listed
L55Training method · B9Custom / unverified
M55Training method · B93
N55Training method · B9Validate
O55Training method · B9https://docs.appliedcompute.com/platform/training/configuration
P55Training method · B9See primary source
Q55Training method · B9May be viable via BYO weights, but no implied support or price
A56Eval metric · B15
B56Eval metric · B15Five normalized score types (higher = better)
C56Eval metric · B15Stop logic still compares actual scores to target and plateau.
D56Eval metric · B15For latency, loss, or cost metrics, redesign stop logic.
A57Data readiness · B33
B57Data readiness · B33Validated / due kickoff / pending / unavailable
C57Data readiness · B33Pending or unavailable prevents customer-ready approval.
D57Data readiness · B33Confirm customer evidence and data-access date.
A58GPU pool fit · F109:F111
B58GPU pool fit · F109:F111Compatible / needs verification / incompatible
C58GPU pool fit · F109:F111Only Compatible pools count toward existing capacity.
D58GPU pool fit · F109:F111Validate GPU family, interconnect, region and permissions.