OpenAI, Anthropic Promise Both Riches and Restraint

OpenAI and Anthropic have identified the ultimate growth-stage pitch: convince Wall Street their businesses are sound and fast-growing while simultaneously assuring governments and the world that their models don't pose unacceptable risks. It's the kind of delicate balance that has never been attempted before in the history of capital markets, which is to say it has been attempted roughly 40,000 times and failed spectacularly each time. Both firms are apparently aiming for "potentially r"—a phrase that trails off mid-sentence in the Axios reporting, much like this entire strategic narrative trails off mid-credibility.

The premise here is delightfully circular: these companies need to demonstrate explosive growth potential to justify pre-IPO valuations, which requires proving their AI models are powerful enough to reshape entire industries and generate enormous returns. Simultaneously, they must convince regulators and the public that these same models are sufficiently constrained, safety-tested, and aligned with human values that we shouldn't be concerned about existential outcomes. The contradiction isn't subtle. It's like a pharmaceutical company telling the FDA a drug is dangerously potent while telling investors it's perfectly safe—just with higher stakes and less regulatory teeth.

OpenAI has form on this particular gymnastics routine. The company has pivoted from nonprofit to capped-profit to Microsoft subsidiary-but-also-independent roughly as many times as it has released product updates. Anthropic, founded by former OpenAI executives, was explicitly positioned as the "safety-first" alternative—a positioning that conveniently converts into a premium valuation multiplier once you reframe safety compliance as a market differentiator. Both companies understand that the regulatory scrutiny around AI has become real enough to be a business risk, but not quite real enough to actually constrain their fund-raising narratives.

The magic words here are "balance" and "delicate." What these actually mean: "We will say whatever each audience wants to hear." To venture capitalists, OpenAI and Anthropic are growth rockets with AI-shaped fuel tanks and infinite addressable markets. To governments, they're responsible stewards implementing robust guardrails. To the public, they're simultaneously both things—a neat trick achieved through the simple expedient of never committing to specific, enforceable claims in any direction.

History suggests this strategy has a shelf life measured in fiscal quarters. Every company going public makes grand promises about balancing competing interests: growth and profitability, disruption and responsibility, innovation and risk management. Some follow through. Most don't, pivoting toward whichever metric drives the stock price on any given Tuesday. AI firms face an additional complexity: if the risk narrative they're downplaying actually materializes, the downside isn't a missed earnings target or a regulatory fine—it's the kind of tail risk that makes a bad acquisition look like a rounding error.

What's fascinating is that neither company appears to have considered a third option: going public on an honest pitch. "We're building powerful technology with genuinely uncertain implications, and we're making money while doing it anyway," lacks the appeal of a dual-messaging strategy, which explains why we'll probably never hear it from a banker's PowerPoint. Instead, we'll get carefully worded S-1 filings that allow both readings simultaneously, regulatory presentations thick with compliance theater, and earnings calls where "AI safety" and "margin expansion" occupy separate verbal dimensions.

The IPO market has always rewarded companies that can convince different audiences of contradictory things. What's new here is the stakes: past iterations of this trick involved overstatement of unit economics or aggressive TAM expansion. This version involves the small matter of whether your core product might pose existential risks. The investors will win either way—if AI is as safe as promised, returns will be enormous; if it's as risky as everyone fears, they'll have gotten liquid before the reckoning. Governments will politely request additional assurances. And OpenAI and Anthropic will continue doing what they do best: telling everyone exactly what they want to hear.