Dear Readers,

For years, the AI industry had a simple narrative: OpenAI was the frontrunner, everyone else was catching up. That story no longer holds. In the span of roughly twelve months, Anthropic went from a well-funded but revenue-modest challenger to a company that, by several reported metrics, generates more annualized revenue than OpenAI. The shift has been so dramatic that it forces a harder, more uncomfortable question than most tech coverage is willing to ask: When both companies eventually go public, which one actually deserves the higher valuation, and why?

The coming IPOs of OpenAI and Anthropic will be the first time public investors get to place real bets on the economics of frontier AI. Not on hype, not on demo reels, but on audited financials, margin trajectories, and the sustainability of revenue that is, for now, largely driven by raw compute consumption. The numbers that emerge will shape how capital flows into AI for the next decade. The central question: When you weigh revenue growth, enterprise traction, valuation multiples, cost structures, and compute dependency against each other, which company presents the more convincing business model ahead of a potential IPO?

All the best,

Kim Isenberg

Let’s start with the topline, because that is where the surprise lives. OpenAI reported roughly $2 billion in monthly revenue, translating to approximately $24 billion annualized (OpenAI, 05/05/2026). A Reuters and The Information report from early March put OpenAI's annualized revenue at over $25 billion (Reuters, 03/05/2026). By any historical standard, that is extraordinary growth for a company that generated around $2 billion in annual recurring revenue as recently as 2023.

Anthropic's trajectory is even steeper. The company reported a run rate of around $1 billion at the start of 2025, which surged past $5 billion by mid-year, hit $14 billion by February 2026, and crossed $30 billion in early April (Anthropic, 04/2026). More recent estimates from SemiAnalysis and Financial Times-adjacent reporting suggest Anthropic may already be approaching $44 to $45 billion in annualized revenue (SemiAnalysis, 05/2026; Investing.com / FT, 05/2026).

If those numbers are even approximately right, Anthropic has overtaken OpenAI on a run-rate basis. But "approximately right" is doing a lot of heavy lifting in that sentence.

Here is the analytical problem that most quick takes ignore. Reuters pointed out that Anthropic calculates its run rate by taking the last 28 days of consumption-based revenue, multiplying by 13, then adding subscription revenue multiplied by 12 (Reuters Breakingviews, 03/10/2026).

That formula means any short-term spike in token consumption, say from a surge in coding workloads or a major enterprise deployment, gets annualized into an eye-catching headline number.

It is a legitimate accounting approach, but it can overstate what a stable, repeatable annual revenue figure would look like.

Additionally, OpenAI supporters have argued that the two companies may not account for revenue the same way. Anthropic could be reporting gross revenue, while OpenAI may be netting out partner shares differently. If OpenAI applied the same methodology, its numbers might look higher (Reuters, 04/08/2026). None of this means Anthropic's growth is fictional. It means the direct numerical comparison requires more caution than a simple "Anthropic: 44 billion, OpenAI: 24 billion, game over" framing suggests.