Meta's AI Bet Meets Backlash: Stock Fell More Than 6 Percent After Raised Spending Forecast
The check was never confirmed as executed, but the disclosure does something rare: it shows how Meta's leadership actually weighed committing $100–200 billion to AI compute before the models were proven. At the same time, CFO Susan Li repeatedly warned that a failed bet would be "very, very bad in financial terms" — and investors have already shown where the limit lies.
Mark Zuckerberg and Meta's chief AI officer Alexandr Wang have given a first-hand account of the conversations from late 2025, when the company considered committing to investments in the hundred-billion-dollar range before the new models had proven themselves. They spoke in an episode of the podcast Cheeky Pint, hosted by John Collison, released Monday, October 13, 2026. The quotes are relayed through Business Insider's rendition of the episode (Business Insider).
"We hadn't seen the models yet"
Zuckerberg's own description of the situation at the end of 2025 was that the optimism was real, but the evidence was not. "At the end of last year we thought, ok, we're quite excited about the progress. We hadn't seen the models yet," Zuckerberg told Collison, according to Business Insider.
He portrayed the sheer size of the bet as a conversation with Wang, who led the technical work. "I'm, like, we're going to write this check for, like, I don't know, like $100 billion, $200 billion. I'm like, 'Alex, are you sure this is going to work?'" Zuckerberg said of Wang.
Worth noting: Zuckerberg never specified the timeframe or exact scope of the contemplated commitment, according to Business Insider. The $100–200 billion figure was something he weighed — not an amount confirmed as disbursed or finally agreed.
The CFO's repeated warnings
Wang, who described having encouraging research results at the time but success far from assured, also spoke about the counterweight inside the company. CFO Susan Li pulled him aside repeatedly during an executive gathering in Hawaii in late 2025.
"Susan pulled me aside many times and said, 'I just want to tell you this: if this doesn't work, this is going to be really, really bad. It's going to be very, very bad in financial terms,'" Wang said of Li, according to Business Insider.
The same bluntness marked Wang's own characterization of the bet's nature. In his own words, as relayed by Business Insider: "this is a fuck ton of risk."
Leadership's defense of the risk
Zuckerberg's argument for why the risk is acceptable goes beyond Meta's own fate. By his logic, the bet fails only if AI as a field fails — not merely if Meta fails.
"So I think in order for it not to work out, it's not just that we have to fail — AI overall has to fail," Zuckerberg said. "And even if that happens, we still have all these other uses to productively use compute."
This is a familiar argumentative structure among the big hyperscalers: the compute investments are framed as partially reversible, because the infrastructure can be put to other internal purposes if AI revenues fail to materialize. Such reassurance is, however, difficult to verify in advance — it assumes the alternative use value actually matches the size of the investment, and no evidence of that exists in this case.
The numbers underneath
The contemplated $100–200 billion must be read against what Meta has actually reported. The company's latest guidance puts 2026 capital expenditures, including finance lease principal payments, at between $130 and $145 billion, according to Business Insider.
For 2027, analysts at Cantor Fitzgerald, in forecasts previously seen by Business Insider, have predicted spending will reach $200 billion. It is worth emphasizing that this is an analyst forecast, not Meta's own guidance — and that the figure coinciding with Zuckerberg's contemplated $200 billion check is coincidental and concerns two different things: one is an investment amount Zuckerberg weighed internally, the other an external estimate of future capital spending.
The trend is nonetheless clear: Meta's spending curve points upward, and the question both management and markets are asking is not whether the company will spend an enormous amount — but how much it can absorb if the payoff fails to materialize.
Investors as the counterweight
It is in the encounter with the market that the risk discussion takes on concrete meaning. Investors have shown they have limits: Meta's stock fell more than 6 percent in after-hours trading after April's figures, despite strong results, after the company raised its annual spending forecast to between $125 and $145 billion.
That reaction suggests parts of the market have not shared the risk appetite leadership is arguing for. Li's repeated internal warnings point in the same direction — and Zuckerberg now openly relates that he was nevertheless willing to contemplate significantly larger amounts than the company has ultimately communicated publicly.
Open questions
Several things remain unresolved after this disclosure. The contemplated amount of $100–200 billion has never been confirmed as an executed or final commitment, and Zuckerberg gave no timeframe or exact scope. It is also not clear which concrete models or projects the check was meant to fund, or what the compute was in practice ultimately used for.
It is also worth recalling that all the quotes come via Business Insider's rendition of the podcast, not from an independent transcript or a direct recording we have verified. The date of the episode's release likewise rests on Business Insider's dating.
What nonetheless remains is a rare insight into how the decisions behind the industry's largest capital commitments are actually made: a CEO weighing checks in the hundred-billion range, a chief AI officer admitting that success is far from given, and a CFO reminding everyone involved what failed bets cost.