Compute allocated to customers, usage-based revenue recognized properly, and a burn forecast that respects how compute costs actually step.
For an AI company, gross margin is a compute question. Inference cost per customer determines whether the business model works, and most AI companies cannot state gross margin by customer with any confidence.
We run the accounting and CFO function for AI companies where compute is the dominant cost and pricing is still moving.
GPU spend, reserved capacity, and inference cost sit in one line item. Without allocation to customers and products, gross margin is an assertion.
Tokens, seats, credits, and overages mixed in one contract make recognition genuinely difficult, and MRR reporting inconsistent.
Training runs, data acquisition, and development work sit awkwardly across ASC 350-40 and Section 174. The position needs documenting before someone asks.
Compute costs step with committed capacity rather than scaling smoothly, so an averaged burn rate misleads exactly when it matters.
We build a unit economics model with compute cost allocated to customers and products, recognize usage-based revenue properly under ASC 606, document a defensible capitalization position for model development, and forecast burn against stepped compute commitments instead of smooth averages.
By customer and by product, from cloud and usage data.
ASC 606 schedules for consumption pricing.
Consistent with your Section 174 position.
Reserved capacity modeled as the step cost it is.
Unit economics an institutional investor will engage with.
Domestic research must be capitalized and amortized over five years, and since most AI spend is development, this can create taxable income at a company with no profit. The R&D credit substantially overlaps and can offset payroll tax for qualifying small businesses. Reserved compute commitments and prepaid capacity raise prepaid-versus-expense timing questions. Data licensing across jurisdictions can create withholding and nexus exposure.
Yes, from cloud cost data and usage telemetry. It's the difference between knowing your gross margin and guessing at it.
It depends on the stage and purpose of the work, and the answer has to be consistent with your Section 174 position. We document both together.
Runway discipline, compute cost control, clean books for the next round, and the R&D credit study. That's most of the value at that stage.
Tell us where you are today. We will follow up within one business day with a view on what is worth fixing first.
Get a free consultation