trailThe rapid growth in frontier AI training costs will continue through the 2030s
The compute cost of the most expensive AI training run will reach about $230 billion (one percent of 2021 US GDP) around 2040.
Trend persistence and technical feasibility· for
The cost of frontier AI training runs continued to grow at more than 2x per year through the mid-2020s
atomic
AI training runs costing on the order of $100 billion will be financially and technically feasible by 2030
atomic
Revenue from AI products will grow fast enough to justify current levels of AI infrastructure investment
atomic
Scaling up training compute will remain the primary driver of frontier AI capability gains
atomic
Financial unsustainability at scale· against
Growth of AI training spending at recent multi-fold annual rates is financially unsustainable through 2030
The most expensive AI training run in 2025 cost on the order of several hundred million dollars
AI infrastructure capital spending currently far exceeds the revenue generated by AI products
Major technology companies' capital expenditure on AI infrastructure continued to accelerate through 2025 and 2026
Growth in spending on the largest AI training runs will slow substantially during the 2020s.
The dollar cost of compute for final ML training runs grew about 0.5 orders of magnitude per year from 2009 to 2022.
AI training runs of around 2e29 FLOP will be feasible by 2030
claim · empirical · derivedA factual claim that rests on inference from other evidence rather than direct observation.constitution →claim page ↗︎
The rapid growth in frontier AI training costs will continue through the 2030s
⇄ContestedCredible evidence or argument exists on multiple sides.constitution →credence 0.25Credence, from 0 to 1: the Steward's probability that the claim, as stated, is true. Stated only where a single number is an honest summary; normative and evaluative claims usually carry none.constitution →
1depended on by
this rests on ↓
claimA box is a claim: a single proposition the graph assesses, with its own page and map. Click any claim to centre the map on it.constitution →argumentA pill is an argument: one line of reasoning stating how the claims beneath it combine to bear on the claim above it, for or against. Arguments are not destinations; click their claims to explore.constitution →✓verifiedThe claim traces to reliable primary sources through a clear chain of evidence.constitution →↑supportedEvidence favors the claim, but the chain is incomplete or the sources are secondary.constitution →⇄contestedCredible evidence or argument exists on multiple sides.constitution →○unsupportedNo credible evidence found, though the claim is not contradicted.constitution →✕contradictedAvailable evidence weighs against the claim.constitution →?unknownInsufficient information to assess.constitution →◌unassessedNo current assessment. Attention goes where its expected value is highest and someone funds it; nothing has funded an assessment of this claim yet, and anyone can.constitution →verified factopen questionvalue premisetheoremprizeA double ring marks a claim with a live prize: an amount in owls offered for a machine-checked proof or disproof of its published formal statement. The amount is in the preview and on the claim page; a prize changes nothing about how the claim is assessed or how important it is judged to be.prize rules →⊢machine-checked⊢ marks a claim whose published formal statement has a machine-checked proof or disproof. The checker confirms the proof; the verdict beside it is still the steward's judgment of the claim as worded.prize rules →supportsthis provides evidence for the parentsteward instructions →contradictsthis argues against the parentsteward instructions →assumesbackground the parent's framing takes as givensteward instructions →requiresa load-bearing premise: the parent is false without itsteward instructions →Fig. Detail falls off with distance; every claim is an address.