Chamath Says VC Cash Funds 70-75% of AI Spending Now

Investor Chamath Palihapitiya wrote on X on 21 October 2026 that “roughly, $0.70-75 of every dollar of AI spending right now comes from AI companies spending VC investment dollars,” and concluded that “the revenue quality in AI is currently poor.” He gave no source for the figures behind that estimate.

His post quoted a summary of remarks by Citadel founder Ken Griffin, who said investors are “almost certainly” in the midst of an artificial intelligence bubble, Bloomberg reported on 20 October 2026. Palihapitiya called his own post “the extrapolation of Ken’s point.”

Business Pill · FINANCIAL CONDITIONS

A short explainer on financial conditions and how they shape funding. It teaches the general idea only and says nothing about any company in this story.

Update, 23 October: Apollo Puts Numbers on Who Pays

Apollo’s chief economist Torsten Slok published the data behind the argument in the firm’s Daily Spark on 22 October: “Ramp data show that 80% of the top 1% of AI spenders are VC-backed, compared with only 13% of the median spender, suggesting that the circularity seen among AI hyperscalers and model providers also extends to token spending, as the startups buying the most compute are paying for it with venture capital rather than organic earnings.” Apollo’s chart, sourced to the Ramp AI Index, puts the shares at 80.4% and 13.1%.

Slok also set out the risk: “If the AI payoff is slow to materialize or never arrives, VC returns would come under pressure, since the capital funding today’s token spending would have little to show for it, and a pullback in venture funding could in turn weaken demand for compute.” As we read it, Apollo’s figures describe who the heaviest buyers are, which is a narrower claim than Palihapitiya’s estimate of how much of every AI dollar comes from venture money.

What Palihapitiya Wrote

• The chain of estimates. He wrote that “1% of companies represent 90% of AI spending,” and that of that 90%, “80% of this spend are from unprofitable VC backed companies.” Multiplying 90% by 80% gives 72%, by our arithmetic, which is consistent with his range of $0.70 to $0.75 per dollar.
• The AWS comparison. He wrote that this “happened in previous cycles as well, like cloud when early users of AWS were mostly venture backed startups,” and that before AWS and others became “legitimate cash machines, there were a few clearing events to sort out these brittle revenues.”
• What he says has to change. “We will need to transition from speculative AI spending to consistent, reliable spend that shows up in public market earnings reports,” he wrote, adding that this “hasn’t happened yet because most of the breathtaking revenue growth comes from LPs investing in VC funds who then send it to startups.”

In plain terms, he describes a chain: limited partners invest in venture funds, the funds send the money to AI start-ups, and that spending shows up as the “breathtaking revenue growth” he describes. As we read it, his argument is that this revenue rests on investors, not yet on spending that shows up in public companies’ earnings reports.

What Was Posted on X

Investor Chamath Palihapitiya’s post of 21 October 2026, quoting a summary of Ken Griffin’s bubble remarks. He estimates that $0.70 to $0.75 of every dollar of AI spending comes from AI companies spending venture capital, gives no source for his figures, and compares the moment to the early days of AWS. This piece sets his estimates beside reported figures.

Posting on 22 October, @shanumathew93 quoted Apollo’s Daily Spark: Ramp data show that 80% of the top 1% of AI spenders are VC-backed, compared with only 13% of the median spender, which Apollo says suggests the circularity seen among hyperscalers and model providers extends to token spending. The post quotes Palihapitiya’s estimate underneath, which is why it sits here; the Update above quotes Apollo directly.

What the Reported Numbers Show

Palihapitiya’s percentages are his own estimates. The reported figures from the past two days sit on both sides of his argument.

Concentration. Around one-fifth of Anthropic’s $16.3 billion in first-half revenue came from just two customers, the Wall Street Journal reported from Anthropic’s draft IPO prospectus, and Anthropic expects a “significant portion” of its revenue to come from a “limited number of customers for the foreseeable future.”

Commitments. The same draft shows Anthropic on the hook for around $450 billion in cloud computing agreements over the next seven to ten years, the bulk of them noncancelable, the Journal reported.

Hardware demand. Supermicro estimated revenue of $18.5 billion to $19.5 billion for its quarter ended 30 September, against its own “guidance of $14.5 billion to $15.5 billion,” and put the difference “primarily due to increased demand for AI platforms,” the company said on 20 October 2026.

Suppliers financing customers. The Information reported that Nvidia is in talks to invest hundreds of millions of dollars in Firmus, the Australian AI cloud provider that pulled its IPO earlier this month, according to Techmeme’s summary. TD Cowen analyst Michael Elias said his checks suggest Nvidia could be guaranteeing some lease payments for Oracle, TipRanks reported.

Credit markets. More than $9bn worth of credit default swaps linked to bonds issued by Oracle, SpaceX, Alphabet, Amazon, Meta, Broadcom and Nvidia traded on average each week in September, up from less than $2bn in July, the FT reported, citing Morgan Stanley Research.

Connecting the Coverage

Our Anthropic prospectus piece set out the draft filing’s revenue, loss and customer concentration, and our piece on Anthropic’s cloud partners covered the compute commitments the Journal described. Our Supermicro piece reported the preliminary quarter from a company that sells AI hardware.

Our Firmus piece reported the Nvidia talks, and our credit-stress piece tracked the default-swap data and the TD Cowen note on Oracle’s leases. Our neocloud shakeout piece covered executives who expect only seven to 10 of hundreds of neoclouds to survive, as The Information reported on 18 October, per Stocktwits, and our piece on Asia-Pacific listings followed DayOne’s weighing of a smaller IPO, as Bloomberg reported, citing people familiar with the matter.

As we read it, Palihapitiya’s chain describes where AI spending money comes from, while the reported figures describe where it goes: a few large buyers, multi-year commitments, and suppliers reported to be discussing or possibly providing financing to some of those buyers.

The Structural Read

As we read it, his chain runs from limited partners to venture funds to start-ups, and he says the revenue growth it produces has not yet turned into spending that shows up in public companies’ earnings reports.

As we read it, Supermicro’s preliminary quarter is a counter-signal on demand: a company selling AI hardware said its revenue estimate beat its own guidance, which Palihapitiya’s post does not address.

As we read it, the reports of Nvidia discussing an investment in Firmus and possibly backing Oracle’s leases describe a supplier stepping closer to its customers’ financing, which is the kind of revenue question his post raises.

Three Implications

CONCENTRATION Around one-fifth of Anthropic’s $16.3 billion in first-half revenue came from just two customers, the Wall Street Journal reported.

DEMAND Supermicro put its estimate above its own guidance “primarily due to increased demand for AI platforms,” the company said.

CREDIT More than $9bn of CDS linked to bonds of large AI-exposed companies traded on average each week in September, the FT reported, citing Morgan Stanley Research.

The Business Engineer Lens

This story maps onto the Business Engineer framework The AI Capex Map.

The framework puts it this way: “It is a sovereign-scale financing problem dressed in tech-company clothing.”

As we read it, Palihapitiya’s post makes a related point from the funding side: he estimates that most AI spending traces back to venture capital rather than to profits. The framework was published in May 2026 and does not discuss his post.

What Is Not Established

Palihapitiya gave no source for the 1%, 90% or 80% figures, and we have not found one; the 72% is our arithmetic on his numbers. We read only the opening of Bloomberg’s Griffin report. Nothing here establishes how much AI spending is funded by venture capital, or whether any clearing event lies ahead.

The Bottom Line

Chamath Palihapitiya estimated on 21 October 2026 that $0.70 to $0.75 of every dollar of AI spending comes from venture-backed AI companies, and called the sector’s revenue quality poor. As we read it, the week’s reported figures, from Anthropic’s customer concentration to Nvidia’s talks with Firmus, show why the question of who ultimately pays for the build-out is now being asked out loud.