Chamath Palihapitiya wonders whether the money being pumped into AI is paying off for anyone outside the small group of companies already amassing it.
The Social Capital founder, in two posts on X on July 17 and 18, called out AI labs on how they train their own models versus how they deal with others copying their work. Chamath then pointed out what he’s observed about the buzzwords dominating SEC filings and how they tend to diminish over time, a reference to the current hype for artificial intelligence and proxy technologies.
books“Right now, everyone is looking to AI like a life raft. But those same people have yet to demonstrate a repeatable, auditable, and verifiable ROI, even as their capital and operating expenditures grow with nominal costs on all things AI.”
Uber, Microsoft, and Meta already exist Rein in AI budgets In response to the findings of A McKinsey poll Which stated that most companies see no impact on profits from generative AI. Other critics do not see a future in which the entire market is not affected by one major default.
Do AI labs play by different rules?
Palihapitiya delivered a jab at Anthropic, and by extension, other players in the frontier modeling space, like OpenAI, on Friday when he published an evaluation of Anthropic’s Fable model on distillation, the practice of using the output of one model to train a cheaper competitor.
Palihapitiya mentioned distillation As an ethical problem it is a real bone of contention. He pointed out that the laboratories themselves built their systems on the open Internet, including copyrighted books, articles, codes, and more.
Now, the same companies that relied on these resources gathered around the world to train their frontier models are now pitted against other companies doing the same with them.
Have big tech budgets finally been reined in?
And the froth is what big tech companies are now trying to drain. Uber’s AI budget The year 2026 was exhausted in roughly four months. The company had to cap programming tools at $1,500 per employee per tool, which is reportedly tracked on an internal dashboard.
Microsoft is phasing out Claude Code licenses in its Experiences and Devices division. It now encourages its engineers to use the GitHub Copilot CLI, a move referred to in an internal memo as intentional scaling.
In an April note, Andrew Bosworth, Meta’s chief technology officer, said: “All movement is not progress and token use alone is not a measure of impact of any kind.”
Nanda Initiative of MIT I looked at leading AI companies across industries and found that 95% of them produced no measurable financial return at all.
Is the AI industry too focused to fail gracefully?
Technology critic Ed Zitron warned this week that OpenAI has become “one of the biggest commitments in modern economic history,” suggesting that its failure would mark a Lehman Brothers moment in the age of artificial intelligence.
By its calculations, OpenAI plans to spend more than $50 billion on computing this year, has assumed about $748 billion in liabilities to Microsoft, Amazon and Oracle, and posted a net loss of $38.5 billion in 2025 on revenue of $13.07 billion.
oraclewhich has pledged more than $340 billion to build capabilities for OpenAI, saw its credit rating cut to the lowest investment grade by S&P Global, with OpenAI named as a major risk.
This is the setting that Palihapitiya urges. A few companies are at the center of trillions of committed spending; However, returns remain unproven for most buyers.
Palihapitiya acknowledged that artificial intelligence is real, adding that it is “the defining change in our lives.” What he wants to change, however, is that returns need to start going to more companies rather than just a few. “We are at the early stage where a few companies are making all the money from our largesse,” he wrote. “This has to be reset so everyone wins.”
Don’t just read cryptocurrency news. Understand that. Subscribe to our newsletter. It’s free.
