AI Company Valuation: Emerging Frameworks

Introduction

AI companies represent the most challenging valuation problem in TMT today. Traditional frameworks struggle with companies that are growing at triple-digit rates, burning billions in compute infrastructure, and operating in a market where the competitive landscape shifts quarterly. AI captured nearly 50% of all global venture funding in 2025 (up from 34% in 2024), with $202.3 billion invested in the sector, a 75% year-over-year increase. Foundation model companies alone captured $80 billion in 2025, representing 40% of all global AI funding. The valuations are staggering: OpenAI raised at an $840 billion valuation on $14.2 billion in revenue, Anthropic at $380 billion on a $2.5 billion run rate. For TMT investment bankers, developing frameworks for AI valuation is essential as AI-driven deal activity accelerates across M&A, growth equity, and capital markets.

The AI ecosystem separates into three layers, each requiring a different valuation approach.

Foundation Models, Infrastructure, and Applications

Foundation model companies (OpenAI, Anthropic, Google DeepMind, Mistral) build large language models that serve as the infrastructure for AI applications. These companies are valued on a combination of compute capacity (GPU fleet size, cloud infrastructure commitments), model performance (benchmark scores, capability breadth), and revenue growth (API revenue, consumer subscription revenue). OpenAI's more than 900 million weekly active users and 50+ million consumer subscribers represent unprecedented engagement metrics for an AI company. Infrastructure providers supply the compute, data, and tools that AI companies need: semiconductor companies (NVIDIA, AMD), cloud platforms (AWS, Azure, Google Cloud), and data infrastructure companies. These are valued using traditional TMT frameworks with AI-specific growth premiums. Public AI infrastructure companies traded at 23.4-35x revenue in 2025. AI-native applications build products that use AI models (whether proprietary or third-party) to solve specific business problems. These companies are valued most similarly to SaaS businesses but with premiums for defensible data moats, proprietary model capabilities, and domain-specific training data.

Beyond standard SaaS metrics (ARR, NRR, gross margin), AI companies introduce sector-specific metrics that affect valuation.

Gross margin is particularly important because AI companies face significant compute costs that traditional SaaS companies do not. A SaaS company with 80% gross margins and an AI company with 55% gross margins may have similar revenue, but the AI company's lower margins mean less of each revenue dollar flows to the bottom line, which should compress revenue multiples. OpenAI's gross margins have been estimated in the 50-60% range due to GPU inference costs, compared to 75-85% for traditional SaaS. As model efficiency improves and inference costs decline, AI gross margins are expected to expand toward SaaS-like levels, and this margin trajectory is a key valuation input.

Compute commitments provide visibility into infrastructure investment and future capacity. OpenAI's cumulative cloud commitments exceed $500 billion (including $250 billion with Microsoft Azure, $38 billion with AWS, and approximately $300 billion with Oracle), representing some of the largest infrastructure deals in technology history. These commitments are both an asset (guaranteed capacity to serve growing demand) and a liability (fixed costs that must be absorbed even if revenue growth decelerates).