The part where “valuation” meets “actually trying to buy it”

Primary markets are where everyone agrees you’re brilliant. Secondary markets are where someone has to wire money. That’s why the OpenAI vs Anthropic chatter keeps snapping back to secondaries: they’re the nearest thing the AI industry has to a lie detector that speaks in discounts.

According to TechCrunch, which points to reporting in the Financial Times, OpenAI’s current valuation is around $852 billion. To justify that in a normal “eventually it’s public” universe, at least one investor told the FT you’re implicitly assuming an IPO valuation north of $1.2 trillion. That is not “we’re optimistic.” That’s “we’re buying the sequel before the first film has shipped.”

Why Anthropic suddenly looks like the sensible option in the casino

On the other side of the trade: TechCrunch says the FT reported Anthropic’s annualized revenue jumping from $9 billion at end-2025 to $30 billion by end-March, driven largely by demand for coding tools. (Yes, “coding tools” are now a macroeconomic force. Please update your priors accordingly.)

If you’re a secondary buyer, that growth curve matters because it gives you something a hype narrative can’t: a plausible path to “this becomes a giant, boring enterprise software cash machine.” OpenAI can absolutely become that too, but it currently comes with more plot twists, more dependency on consumer distribution, and more governance lore per dollar of revenue.

The Mechanism Test: what secondaries are pricing (and what they’re ignoring)

Secondaries aren’t predicting “who has the best model.” They’re pricing governance confidence, sales execution, and the odds your cap table won’t turn into a hostage negotiation.

• Governance confidence: in secondaries you’re not buying a demo, you’re buying years of board decisions. If the story feels unstable, buyers demand a discount.
• Distribution durability: consumer virality is great until platforms change their minds, regulators get curious, or your CAC turns into a small nation’s GDP. Enterprise stickiness is dull, and dull is investable.
• Revenue quality: “annualized revenue” can hide a lot of sins, but it’s still a stronger anchor than “everyone loves us.” Coding tools with real usage are an especially clean signal.

So is OpenAI “the Netscape of AI”? (a useful insult, even if it’s wrong)

TechCrunch notes Sapphire Ventures president Jai Das (no stake in either company) told the FT he sees OpenAI as “the Netscape of AI.” That’s not a prophecy, it’s a warning label: a reminder that being first and being the default are different jobs, and the second job is usually won by whoever controls the distribution channel (or writes the rules for it).

OpenAI can still win. But if the market starts treating Anthropic as the “safer” bet at today’s prices, the competitive battlefield shifts. It stops being purely about capability and becomes about how quickly each lab can turn capability into durable cashflows without exploding the trust surface (security incidents, policy blowback, procurement hesitation, and the occasional governance soap opera).