Anthropic AI GDP projections moved from thought experiment to front-page number this week. Jack Clark, co-founder of the AI lab behind Claude, said one of Anthropic’s internal models forecasts that rapidly improving artificial intelligence could lift US economic output by nearly 50% by 2030 – to roughly $44 trillion, against a current level of just over $30 trillion.
“The estimated potential US GDP of $44 trillion, which compares to the current level of just over $30 trillion, sounds very outlandish, I know,” Clark told policymakers in Bangkok, in remarks reported by Reuters. The candor was half the point. The other half: Clark paired the number with an urgent call for international regulation to keep pace with capabilities he says are accelerating across scientific discovery and entirely new industries.
The Anthropic AI GDP forecast is three scenarios, not one number
The $44 trillion headline is the aggressive end of a scenario set, not a single-point prediction. Anthropic’s research outlines three pathways; the most aggressive projects US GDP of about $44.4 trillion by 2030, while a more modest AI-adoption path lands near $34.1 trillion – still a roughly 32% lift over today’s economy. That spread is the honest part of the story: where AI actually lands depends on adoption speed, compute buildout, regulation, and whether productivity gains show up in GDP stats or just in slide decks.
Clark’s track record gives the claim unusual weight. Anthropic’s researchers have lately shown a taste for turning gnarly puzzles into headlines – including the cryptographic mystery of the NSA’s leaked curve seeds, which the company says Claude cracked in under 30 minutes. A GDP forecast is a very different kind of puzzle, and Clark knows the audience for it is skeptical by default.
Why markets should care about the AI GDP numbers
A 30-to-50% jump in US output within four years would be an economic event without modern peacetime precedent – faster than the post-war boom, and concentrated in software and science rather than steel and suburbs. That is exactly why the International Monetary Fund used its Global Financial Stability Report this month to warn that an AI-market correction could trigger global financial turmoil if leverage and expectations run ahead of returns. The bull case and the tail risk are now sourced from adjacent spreadsheets.
The regulation ask
Clark’s Bangkok remarks leaned hard on governance: accelerating capabilities, he argued, mean international rules cannot wait for perfect information. The political weather is already shifting – Washington has rebranded its AI posture around “superintelligence” language, and China has proposed export controls on 23 AI technologies including model weights. Clark’s warning lands in a policy environment that is racing to catch up with the same acceleration his forecast depends on.
The honest caveats
Three, plainly. First, this is one internal model’s scenario range, not consensus economics – no forecaster outside Anthropic has put their name to $44 trillion. Second, the aggressive case is explicitly conditional; drop the conditions and the number evaporates. Third, AI forecasts have a long history of arriving right on time for the wrong decade. Bitcoiners, notably, have heard this movie before: the same lab’s quantum research that Google’s Willow results put in perspective is a useful reminder that capability headlines and capability timelines are different animals.
Anthropic, for its part, is pricing its own story aggressively – valuation talk around its IPO has eased from near $2 trillion to roughly $1.5 trillion after a $16.8 billion quarter. If Clark’s GDP scenario is even half right, those numbers look conservative. If it is wrong, they look like the top of the cycle. Either way, the lab just put a number on the bet – and made everyone else defend theirs.
