Henry Blodget on the Software Selloff Hysteria and the Problem for OpenAI | Odd Lots

OpenAI's valuation has climbed from $300 billion to $800 billion in less than a year, even as the company loses money on every power user and faces fiercer competition than any supposed «winner» in tech history. Blodget recalls that when Google's Gemini surpassed ChatGPT in capability, it marked a watershed moment: less than two years after ChatGPT's launch, the incumbent had already caught up. This stands in stark contrast to Amazon in the 1990s, which maintained an insurmountable lead even as it teetered on the brink of bankruptcy.

The bull case rests on projections of $100 billion in revenue within three years and the assumption that compute costs will plummet, allowing the unit economics to flip positive. But OpenAI faces a unique disadvantage: unlike Google, Facebook, or Microsoft — which generate tens of billions in free cash flow annually to fund chip purchases — OpenAI must raise capital in the open market to keep pace. If the economics don't improve soon, the company will simply run out of money, no matter how visionary its mission to build AGI.

Blodget draws a parallel to the dotcom era, when out of hundreds of companies that went public in the 1990s, only Amazon went on to make investors serious money. Cisco took 25 years to recover its peak stock price; eBay sputtered. The lesson: being early, well-funded, and hyped is no guarantee of dominance, especially when the technology lacks the network effects and switching costs that cemented Google's monopoly two decades ago.

Blodget argues that AI is actually good news for established media brands like Bloomberg, The New York Times, and The Wall Street Journal. In an era of AI-generated slop, deepfakes, and synthetic content, readers need trusted intermediaries who can verify facts, provide context, and hold sources accountable. A million people read The New York Times at its print peak; today, billions of pieces of content circulate daily, but most are unverified and unreliable.

The industry faces a different existential challenge: there is simply too much media. Distribution has collapsed as Google and Facebook have pulled back, forcing publishers back into a direct-subscriber model that resembles the 1990s magazine era. The companies that survive will be those that serve passionate, loyal audiences with differentiated content — scoops, expertise, and editorial judgment that AI cannot replicate.

Blodget also defends his own AI newsroom experiment, in which he used ChatGPT to create five AI staffers with distinct personalities and headshots. The backlash taught him about the deep anxiety younger workers feel about job security. But he remains optimistic: media has always been about knowing what's happening and what it means, and humans will continue to want that — especially when they trust the source.