Sectors

AI and Machine Learning

AI and machine learning companies face a distinctive financial modeling challenge: compute costs appear in the cost of goods sold, compressing gross margins in ways that require explicit modeling and investor disclosure.

Financial Infrastructure Profile

An AI company's financial model has a structural characteristic that distinguishes it from most other software businesses: the cost of running inference is a cost of goods sold, not an operating expense. This single classification decision affects every gross margin calculation in the financial model and in all investor materials. A Series A investor evaluating an AI company will apply a gross margin benchmark from the AI sector, not from the broader software sector. That benchmark is lower. A financial model that has classified inference compute as an operating expense will report a gross margin that looks strong against software benchmarks and weak against AI benchmarks. Investors with sector experience will identify the misclassification immediately.

The revenue model diversity of AI companies adds a second layer of financial modeling complexity. A company that earns revenue from API consumption, a SaaS subscription, and professional services implementation engagements has three revenue streams with different recognition methodologies, different gross margins, and different growth characteristics. The financial model must separate them, apply the correct treatment to each, and produce a consolidated picture that reflects the blended economics of the business as it scales. This is not optional. It is what an investor's financial team will reconstruct during diligence if the company has not done it first.

Sector-Specific Financial Challenges

  • Compute cost classification: inference compute costs are a cost of goods sold and must be classified accordingly in the chart of accounts, the management accounts, and the financial model, with the methodology documented for investor review
  • Gross margin disclosure at product level: the gross margin for API consumption revenue differs materially from SaaS subscription revenue and professional services, and must be reported separately to allow investors to evaluate the economics of each revenue stream independently
  • Model development cost treatment: the accounting treatment for the cost of developing and training the underlying model, including the capitalisation question under IAS 38, must be resolved and applied consistently before the first audit
  • Revenue model selection and its financial modeling implications: API consumption, SaaS subscription, and professional services each produce different revenue recognition timelines, different gross margin profiles, and different growth model structures
  • GPU and infrastructure cost modeling at scale: the financial model must project compute costs as a function of usage volume with explicit assumptions about the cost per unit of inference at different scales
  • Valuation methodology selection: AI companies do not fit standard software multiples cleanly, and the valuation analysis must select methodologies that are defensible in the context of the company's specific revenue model and growth stage
  • Competitive moat quantification: the financial model should reflect the cost and time required for a competitor to replicate the model's capabilities, as this is a factor institutional investors will assess when evaluating the durability of the company's gross margin position

Relevant Service Layers

Relevant Models

Selected Outcome

Structure Layer · AI and Machine Learning · Pre Seed Stage

The company had issued equity to four co-founders and a strategic advisor under arrangements that had not been documented in a formal shareholders agreement or reflected in a maintained cap table. One co-founder had departed six months earlier and the equity position had not been formally addressed. Three SAFE notes had been issued to angel investors, none of which had been modeled at their conversion mechanics. Oakworth documented the existing equity positions, prepared the formal legal instructions required to address the departed co-founder's position, rebuilt the fully diluted cap table with all three SAFEs modeled at cap-based and discount-based conversion prices, and produced a waterfall analysis across ten exit scenarios. The cap table was used without amendment in the company's Seed round legal process, with the lead investor's legal team noting the quality of the documentation at completion.