AI infra company taken from chaos to S-1 ready in 9 months
An AI inference platform doubling every six months, with finance held together by a controller, two spreadsheets, and prayer. We rebuilt the entire finance stack and walked the company to audit-grade close discipline.
- Three different revenue recognition policies across consumption, committed, and BYOC contracts.
- GPU depreciation schedules treated identically to commodity hardware, inflating gross margin by ~620 bps.
- No SOX-readiness framework, no segregation of duties, controller doing wire approvals personally.
- Forecast accuracy at ±28% on a quarterly basis, public-market unforgivable.
- Audit firm flagged five potential material weaknesses in the pre-engagement walkthrough.
Revenue policy unification
Wrote a defensible ASC 606 policy covering consumption, committed, BYOC, and overage tiers. Restated trailing 18 months of revenue under the unified framework.
GPU economics overhaul
Rebuilt depreciation schedules by chip class (H100, H200, B200, custom ASIC) with realistic useful-life curves, surfaced 620 bps of overstated margin, then recovered most of it through utilization improvements.
Controls & SOX scaffolding
Designed segregation of duties, key controls catalog, and quarterly testing protocol. Hired three accountants and a controller into the structure, not around it.
FP&A institutional rebuild
Driver-based model anchored on token volume, GPU utilization, and contract mix. Brought forecast accuracy to ±4% within two quarters.
A US-based AI inference platform serving Fortune 500 enterprises with sub-50ms model serving. $184M ARR growing 11% MoM, having tripled headcount in 18 months. The board had quietly green-lit an IPO window 12 months out, but the audit firm refused to sign without a top-to-bottom finance rebuild. Revenue recognition was being handled in three different ways across consumption, committed, and bring-your-own-compute contracts. The CFO had resigned two weeks before we were brought in.
- Month 1
Diagnostic & triage
Three-week deep audit. 47 issues catalogued; 12 marked filing-blocking and queued ahead of everything else.
- Month 2–3
Revenue policy & restatement
ASC 606 policy ratified. Trailing 18-month restatement completed and tied to the GL within 6 cents.
- Month 4–5
GPU economics & cost of revenue
Useful-life curves and chip-class depreciation live. Gross margin restated; reconciled to the new utilization dashboard.
- Month 6–7
Controls & hiring
SOX-readiness framework implemented. Controller and three accountants hired. First clean quarterly close under the new framework.
- Month 8
FP&A cadence
Driver-based model live. Forecast accuracy improved from ±28% to ±6% in one quarter.
- Month 9
Audit sign-off
Big-four audit firm signed clean. Zero material weaknesses. S-1 narrative drafted alongside legal counsel.
- Audit sign-off achieved on the first attempt, no material weaknesses raised.
- S-1 financial sections drafted and stress-tested by counsel before the IPO window opened.
- Gross margin restated honestly, then recovered 480 bps via GPU utilization improvements.
- Finance team scaled from 4 to 9 with a written org design, no hero hires.
- Board reporting moved to weekly KPI cadence, monthly variance reviews, quarterly board pack.
- 01
GPU depreciation is the single most-abused line on every AI infra P&L. Get it right before someone else does it for you in S-1 comments.
- 02
Three revenue recognition policies means you have zero. Pick one, restate, and live with the lower number for a quarter.
- 03
Audit readiness is a 9-month exercise that companies start at month -2. The math never works.
"We were running a public-company business on a Series A finance stack. CapMaven told us the truth in week one and then spent nine months making it untrue."
Did you replace the CFO during this work?+
No. We ran as fractional CFO + finance build team while the company recruited a permanent public-company CFO. We led the search alongside the board and handed off in week 38.
How did you handle the gross margin restatement with the board?+
Pre-meeting, in writing, with a recovery plan attached. The 620 bps haircut was paired with a credible 480 bps recovery path. The board approved the restatement in the same session.
Is this only for IPO-bound companies?+
No. Every AI infra company over $50M ARR benefits from the same revenue policy and GPU economics discipline, the IPO timeline just forces the conversation.
Your case starts here.
Two weeks. Three scenarios. One honest read on your numbers.
The CFO Diagnostic gives you the same artifact every case above started with, a cash model, a unit-economics teardown, and a capital-readiness scorecard.
Start here,
Stop guessing. Start knowing.
Book a free 20-minute discovery call, or go straight to a $400 CFO Diagnostic. The Diagnostic delivers a working read of your cash position, runway, and top 3 financial risks within two weeks. Whether you engage further or not, it's the clearest financial picture most founders have ever seen.