Scores by kind of claim
Each kind of claim is scored on its own, out of 100: the share that held up, with the range it likely falls in. Fewer resolved claims give a wider range.
-
Factual claims
Statements about the present or the past, checked against the record: did they hold up?
83 78 88 -
Predictions
Statements about events outside the person's control, checked once they fell due: did they come true?
Insufficient record: 2 predictions resolved, 5 needed for a score.
-
Promises
Commitments to the person's or organisation's own future action, with something to deliver: was it delivered?
Insufficient record: 0 promises resolved, 5 needed for a score.
All types combined: 83 (likely 78 to 88, high confidence), from 86 claims: 86 factual claims. It mixes kinds of claim in different proportions for different people, so compare people kind by kind.
Daniela Amodei's claims
43 claims
“Training these models is extremely expensive. It takes a massive amount of compute power, which is limited in supply.”
Daniela Amodei, Anthropic president, interviewee 9 May 2026 Daniela Amodei, Co-Founder and President of Anthropic: Building AI the Right Way (transcript) (The Transcript Desk)
Training AI models is extremely expensive and requires a massive amount of compute, which is limited in supply.
Resolved true The record shows it happened or was so.
What would count as true or false
Judged as of 2026-05-09. True: credible sources (industry reports, company disclosures, research such as Epoch AI) show that training frontier AI models costs very large sums (tens or hundreds of millions of dollars or more), needs massive compute, and that the supply of AI compute (GPUs, accelerators, data-centre capacity) was constrained around that time. Mostly true: cost and compute needs are confirmed, but supply was only somewhat tight rather than clearly limited. Mixed: one of the two parts (expense and compute needs; limited supply) is supported and the other is not. Mostly false: the claim applies only to the largest models and is badly overstated for the models meant, or compute was plentiful. False: evidence shows training is not expensive and compute is not constrained.
Evidence
- Training compute costs are doubling every eight months for the largest AI models Epoch AI · 19 June 2024
“The most advanced models now cost hundreds of millions of dollars, with expenses measured by amortizing cluster costs over the training period.”
- American AI Companies Can’t Get Enough Chips Center for a New American Security (CNAS) · 7 May 2026
“In 2026, artificial intelligence (AI) chip production has become a binding constraint on the pace of the AI compute buildout.”
- Is AI running out of compute? J.P. Morgan Asset Management · 17 April 2026
“Orders for Nvidia GPUs have grown to $1 trillion through 2027, double a year ago, with lead times across GPU and custom silicon stretching to nearly a year.”
- Is AI running out of compute? J.P. Morgan Asset Management · 17 April 2026
“We are unlikely to “run out” of compute, but market conditions will increasingly reflect a supply-constrained environment.”
Reasoning
Epoch AI reports that the most advanced models cost hundreds of millions of dollars to train, measured by amortising the cost of the compute cluster over the training run [E1]. That supports both the expense and the large compute needs. Two days before the claim date, CNAS stated that in 2026 AI chip production had become a binding constraint on the compute buildout [E2]. J.P. Morgan reported $1 trillion in Nvidia GPU orders through 2027 and lead times of nearly a year, describing an increasingly supply-constrained market [E3, E4]. J.P. Morgan's comment that compute is unlikely to "run out" [E4] does not undercut the claim that supply is limited.
Cite this claim
Daniela Amodei, “Training AI models is extremely expensive and requires a massive amount of compute, which is limited in supply.” (said 9 May 2026). Verdict: resolved true. Reliability API, claim c_179b0d1b24a6, method 0.1.0+28da144b. https://epistacklabs.com/claims/c_179b0d1b24a6
Claims without a verdict
Is a verdict here wrong, or do you speak for Daniela Amodei? Reply on the claim's own page, with the form at its foot (how replies work). Verdicts are made automatically; how verdicts and scores are made.