Anthropic's Jack Clark Tells the IMF AI Could Push US GDP to $44 Trillion

Close to half the US economy's current size, added within four years. That's the number Anthropic's Jack Clark put in front of the IMF while Wall Street was still fixated on CoreWeave's balance sheet.

Jack Clark, Anthropic's co-founder and head of policy, delivered the Per Jacobsson Foundation Lecture at the IMF-World Bank Annual Meetings in Bangkok on October 17. His message to the room full of central bankers and finance ministers was blunt: AI is improving fast enough that it could expand the US economy by close to half its current size within four years. Policymakers need to start planning for that now, not after it happens.

Clark pointed to Anthropic's own Economic Scenario Explorer, an internal modeling tool the company published in September. It lays out three paths for the US economy through 2030. In the modest case, where AI diffuses the way the internet did, gross domestic product reaches $34.1 trillion, barely 1.6% above where the economy would land without AI. In the substantial case, where AI autonomously handles half of knowledge work, GDP hits $36.3 trillion, an 8.3% lift. In the extreme case, where AI outperforms humans on nearly every knowledge-work task and leaves little for people to do instead, GDP reaches $44.4 trillion, a jump of more than 30%.

"If you end up with this level of GDP growth, you have moves available to you as a policymaker that are unimaginable today," Clark told the audience, according to Reuters. He went further. His message to the IMF crowd was plain: "Policymakers should get ready to spend."

That is a strange thing to hear from the policy chief of a company that just spent much of September defending itself against a researcher asking exactly how solid those numbers are. Gary Marcus, the cognitive scientist and longtime AI skeptic, has called Anthropic's underlying $30 trillion-plus revenue framing a fantasy. His argument: extrapolating from models that still hallucinate and make basic errors to tens of trillions in restructured economic output skips past evidence that does not yet exist. Clark's own framing hedges in the other direction. He told the IMF that AI capability will keep advancing at a fast, sustained rate, but that diffusion across the actual economy will likely be slower and messier than enthusiasts expect. Two very different numbers can both be true at once: the technology compounds quickly, the plumbing that turns it into GDP does not.

Moody's warns that AI spending is turning the world's most profitable companies into capital-intensive ones

Moody's issued a sector-wide credit warning on July 24, flagging that the combined AI capital spending of six major tech companies will hit $785 billion in 2026 and approach $1 trillion by 2027. The report highlights how $1.2 trillion in lease commitments creates hidden leverage that traditional P&L analysis misses, and how Alphabet posted its... - AI spending threatens tech company credit - hyperscaler capital expenditure impact on profitability

Clark's Bangkok remarks land in the middle of the most uncomfortable week Wall Street's AI trade has had all year. CoreWeave shares fell 16% intraday on October 14, and yields on its unsecured 9.625% bonds due 2032 climbed toward 15%, a level bond markets typically reserve for companies they suspect might not make it. The notes traded below 80 cents on the dollar, according to reporting on the selloff. CoreWeave's own CEO, Michael Intrator, called the market reaction irrational and pointed to a $104.2 billion second-quarter revenue backlog plus more than $25 billion in fresh customer commitments signed in the third quarter. Backlog is a promise, though, not cash in the bank, and the company is still carrying roughly $52 billion in combined debt and lease obligations against operating cash flow that has not caught up.

That is the exact tension Clark's GDP pitch is aimed at. If AI genuinely compounds the way Anthropic's substantial or extreme scenarios describe, the debt financing the current buildout, CoreWeave's bonds, Marvell's chip bets, Microsoft and Meta's capital spending, eventually gets justified by real output. If it doesn't, bond traders pricing CoreWeave like distressed debt turn out to have been the realists all along. Clark isn't a neutral party in that argument. He runs policy for a company that needs two things at once: continued government support, and continued investor conviction that frontier AI pays for itself. The IMF stage was always going to be a venue for making that case to the people who set the rules and the capital flows that let it keep going.

Clark did pair the growth pitch with a warning, telling the IMF that the pace of change ahead needs "unusual and serious" international rules, not just optimism about GDP charts. That is a smaller point than the $44 trillion headline, but a telling one. Even Anthropic's own policy chief is framing this as a sudden-change event that could outrun the institutions meant to manage it, not a steady glide path anyone can simply underwrite.

None of Anthropic's scenarios are forecasts, and the company says as much. They're a model of what could happen under specific assumptions about adoption speed and task automation, not a prediction that any of the three outcomes will occur. But put that next to a week where credit markets started treating one of AI's biggest infrastructure landlords like a company that might default. Clark's testimony reads less like neutral economic modeling and more like an opening argument in the case Anthropic needs Washington, and investors, to believe.

Wall Street is finally asking whether the AI boom can pay for itself • DeepSeek Ties Its New V4.1-Flash-1017 Model Tighter to Its Own Harness • Putin envoy says US and Russia are discussing joint AI data center deals

This article is posted in AI News, check it out for more related stories.