The first AI model company to crack the profitability narrative may not be OpenAI.
According to a WSJ exclusive confirmed by CNBC, Anthropic expects to achieve its first quarterly profit in Q2 2026—roughly $559M in operating profit on projected quarterly revenue of $10.9B. If it delivers, this would surpass the commercialization pace of both OpenAI and xAI.
Bottom line first: the profit figure is real, but don't rush to declare the AI industry profitable.
Let's look at the structure. $10.9B quarterly revenue annualizes to ~$43.6B. Anthropic's full-year 2025 revenue was approximately $2B. In one year, that's a 20x+ increase. The growth driver: Claude enterprise API adoption, particularly Claude 3.5/3.6 surging in code generation and enterprise workflows.
But the cost side is equally staggering. Reuters reports Anthropic signed an agreement to pay SpaceX $1.25B per month for compute capacity. Do the math: $15B annually, all going to compute. This is why Anthropic—after years of losses—is finally approaching profitability. Not because it's cutting costs, but because revenue growth is finally catching up to compute procurement speed.
There's a narrative shift worth noting: model companies are converting funding into compute lock-in. Anthropic's billions in funding largely became compute commitments with SpaceX and AWS. This isn't traditional "R&D spending"—it's more like pre-buying future infrastructure使用权.
The FT reported that Anthropic's Q2 revenue will surpass OpenAI and xAI. This number needs question marks. OpenAI has never publicly disclosed quarterly revenue, nor has xAI. Both WSJ and CNBC used "source says" phrasing, indicating anonymous sourcing rather than official company confirmation.
My read: Anthropic is genuinely approaching a profitability inflection point—the API call volume growth curve supports this. But the $10.9B figure, if it includes non-GAAP adjustments (like compute contract accounting treatment), the actual口径 may differ from the headline.
One more detail: Anthropic's profit is "operating profit," not net income. Stock-based compensation and interest expenses aren't factored in yet. Still, operating profit turning positive is a meaningful signal—the core business model works.
Industry implications: if Anthropic proves first that model companies can be profitable, it puts pressure on OpenAI's IPO valuation—investors will ask, "If Anthropic can profit, why are you still losing $14B?"
But don't get too excited. A $1.25B monthly compute bill means that if revenue growth slows, the profit window closes fast. This isn't a passive income model—it's a system that must keep running at full speed on a treadmill.
Key sources: