Chamas Palihapitiya asked: Who are the real winners of the AI craze?

Chamas Parihapitiya is asking: Has the huge amount of money invested in artificial intelligence really brought returns to other companies except a few companies that have already benefited from it?

The founder of Social Capital posted two posts on the X platform on July 17 and 18, pointing out that AI laboratories adopt double standards when training their own models and treating other peers who replicate their results. Later, he pointed out a phenomenon he noticed: the popular words that dominate SEC documents often lose popularity over time-a clear allusion to the current boom in AI and agent technology.

He wrote: "Right now, everyone is pouncing on AI like grasping at a lifeline. But these people have so far failed to demonstrate a repeatable, audited, verifiable return on investment, even as their capital expenditures and operating expenses continue to climb with the cost of all AI-related tokens."

Uber, Microsoft and Meta have begun cutting AI budgets after a McKinsey survey showed that most companies do not see any revenue impact from generative AI. Other critics cannot foresee a future without a major mistake dragging down the entire market.

Does AI laboratories follow different rules?

On Friday, Palihapitiya launched a barrage of Anthropic (and other players in cutting-edge modeling areas that extend to OpenAI). He cited Anthropic's Fable model's assessment of distillation-using the output of a model to train a cheaper competitive model. Palihapitiya pointed out that distillation as a moral issue is indeed controversial. He emphasized that these laboratories themselves are based on the open Internet, including copyrighted books, articles and code. Today, companies that once relied on leveraging resources from around the world to train cutting-edge models are opposed to others treating them the same way.

Are the budgets of big technology companies finally coming under control?

The bubble is exactly what big technology companies are trying to squeeze out now. Uber's 2026 AI budget will be exhausted in about four months. The company had to cap the use of coding tools at $1500 per employee per tool, which is understood to be monitored through internal dashboards. Microsoft is phasing out Claude Code licenses in its Experience and Devices division and instead encouraging engineers to use GitHub Copilot CLI-a move an internal memo called a "deliberate benchmark." Andrew Bosworth, Meta's chief technology officer, said in an April memo: "All actions do not amount to progress, and token usage alone cannot measure any form of actual impact." The MIT NANDA initiative studied cross-industry enterprise AI pilot projects and found that 95% of them did not produce any measurable financial return at all.

Is the AI industry too concentrated to fail safely?

Technology critic Ed Zitron warned this week that OpenAI has become "one of the largest liabilities in recent economic history" and pointed out that its failure will amount to a Lehman Brothers moment in the AI era. According to his calculations, OpenAI plans to invest more than US$50 billion in computing this year, has assumed approximately US$748 billion in debt to Microsoft, Amazon and Oracle, and recorded a net loss of US$38.5 billion in revenue in 2025. Oracle has committed to investing more than $340 billion in infrastructure for OpenAI, but its credit rating was downgraded to the lowest investment grade by Standard & Poor's Worldwide, and OpenAI is listed as a key risk factor.

This is exactly what Palihapitiya questions: A few companies are at the center of trillions of dollars in promised spending, yet for most buyers, the returns remain unconfirmed. Palihapitiya admitted that AI is real, adding that "this is the most decisive change in our lifetimes." But what he hopes to change is that the rewards should start benefiting more companies, not just a few. He wrote: "We are in the early stages of a few companies making all the profits from our generous investment. This must be readjusted to achieve a win-win situation for everyone."