Morning Edition · Monday, October 19, 2026Published at 2:14 AM EDT · New York
Clark's figure of $44.4 trillion in US output by 2030 is far above the middle path in Anthropic's own September economic model, and the debt terms show what lenders charge to fund the compute behind that forecast.
Jack Clark, a co-founder of Anthropic, used the International Monetary Fund (IMF) Per Jacobsson lecture in Bangkok to tell finance ministers and central bankers that the singularity has started. "I believe that this year, we've gained convincing evidence that it has now begun," he said, according to a clip posted by AI Post. He argued that AI will take the world from roughly 10 million scientists, a figure he attributed to the United Nations Educational, Scientific and Cultural Organization (UNESCO), to about 10 billion, "almost all virtual".
He then attached a specific figure to the claim. CNBC reported that Clark told policymakers AI could greatly increase US gross domestic product. 36Kr's account of the speech says he gave a path to $44.4 trillion by 2030, against current output of just over $30 trillion. In that account, the path depends on AI autonomously running complex scientific exploration, powered by recursive self-improvement.
Anthropic's own published model does not support that figure as a central case. In September its economics team released the Econ Scenario Explorer, which breaks every US job into Department of Labor task bundles. Its "modest" scenario puts 2030 output at $34.1 trillion and its "substantial" scenario at $36.3 trillion. Only the "extreme" scenario comes close to Clark's figure. That scenario assumes annual growth of 15.4% by 2030, an economy-wide unemployment rate of 11.9%, and a 21.5% drop in cognitive employment. The authors assign no probabilities to any scenario. Clark's lecture therefore treated the extreme scenario in his company's own model as the expected outcome.
The financing side shows how lenders price the same future. Anthropic offered Blackstone a 10% coupon on $9 billion of junior debt in a $60 billion chip financing called Project North Star, the Financial Times reported, as relayed by US News. Similarly rated borrowers pay about 7% in public bond markets. FourWeekMBA's summary of the prospectus says the debt underwrites a five-year lease of Google tensor processing units (TPUs). A $42 billion senior tranche covers about one-third of the $125.2 billion Anthropic has committed under that lease. The prospectus also flags a potential conflict of interest because Broadcom is both a chip supplier and a lender in the deal.
Anthropic, which carries one of the industry's largest fixed compute commitments, is also making the most aggressive public growth forecast, and its co-founder delivered that forecast to the IMF, the institution whose analysis shapes official macroeconomic forecasts worldwide. Neither fact disproves the other. The roughly 300 basis-point premium over similarly rated public debt is the clearest market estimate of the risk that Anthropic's revenue arrives later than the lecture implies.
Part of a tracked trend
Frontier Labs Publish the Macro Case for Their Own Technology
AI labs increasingly publish formal macroeconomic scenarios — GDP paths, labor-share trajectories, displacement estimates — that become reference inputs to fiscal and regulatory debate, so expect recurring lab-authored economic modeling, contested replication by academic and official forecasters, and policy anchored on numbers the vendors themselves produced.
Start a discussion in Townsquare.
More from this edition
If lenders and officials accept Anthropic's growth framing, Anthropic can fund compute more cheaply. Alphabet and Broadcom get contracted demand for tensor processing units (TPUs). Blackstone gets a coupon above 10% on $9 billion of junior debt that, per the Financial Times as relayed by others, can convert into Anthropic shares.
The speech and the debt terms check out, but reports drawn from Reuters coverage say Clark called the $44 trillion figure Anthropic's "extreme" scenario, so the article's claim that he treated it as the expected outcome is not supported. The option to convert into shares also makes the 10% coupon a weak comparison with the roughly 7% paid on ordinary public bonds.
An open-source-intelligence read of how likely this story is true with its real nuance, not a judgment of any outlet. It assesses the claim, weighing independent and adversarial reporting. How we label confidence.
What this means
The mechanism links narrative to the cost of capital. Lab-authored growth scenarios support the revenue assumptions that justify leasing tens of billions of dollars of TPUs, and the junior coupon shows lenders want high-yield compensation for that risk. Alphabet and Broadcom gain contracted TPU demand. Blackstone and the private-credit funds that take the syndicated junior paper carry the first loss if Anthropic's revenue growth slows. The well-documented precedent is the 1998 to 2001 fiber buildout, financed with high-yield debt against traffic-growth forecasts that carriers themselves published, which ended in the defaults of 2001 and 2002. The consensus treats compute demand as a near-certainty. The pricing here says lenders do not. The desk's read is that over the next several months more frontier-lab compute will be financed through structured, lease-backed private credit priced near high-yield levels. That depends on lab revenue continuing to grow fast enough to service the senior tranches. Official forecasters such as IMF staff will cite the lab scenarios but publish materially lower central paths. The read would be wrong if the junior tranche clears close to public investment-grade spreads, or if labs start funding these leases from operating cash flow.
What to watch
Observations to monitor, not financial advice.
Synthesized from: Polylog editors · Anthropic Research · CNBC · 36Kr · US News / Reuters (FT report) · FourWeekMBA
Comments
0No comments yet.