Escape Meridian

Leopold Aschenbrenner · May 29, 2024

Leopold Aschenbrenner argues in SITUATIONAL AWARENESS for an American AI-compute buildout

Markdown · JSON · Original source

Leopold Aschenbrenner’s piece was published on May 29, 2024. He argues that the race to is also an industrial race: training clusters require chips, datacenters and power, and the infrastructure is already being built. His rough calculations project enormous investment and electricity needs if current trends continue. He sees power, permitting and datacenter construction—not willingness to spend—as immediate obstacles, and argues that US natural gas and regulatory changes could make the buildout feasible. He warns that building clusters in the Middle East could put infrastructure for training and running AGI under autocratic control, which he frames as an American national-security risk. His conclusion is forceful: “The power constraint can, must, and will be solved,” and the clusters should be built in America or close democratic allies. How much of the projected buildout can be supported depends on uncertain assumptions about AI revenue, compute growth and the pace of power infrastructure construction.

Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.

Leopold AschenbrennerArticle · situational-awareness.aiOpen original source

Original source

Leopold Aschenbrenner · Article

Source details

Original source date: May 29, 2024 (2024-05-29T22:21:37+00:00).

Date basis: Date stated on the publisher’s page.

Added to Escape Meridian: 2026-11-18T12:43:10.989Z. First observed: 2026-10-30T09:13:59.699Z. Retrieved: 2026-10-30T09:15:44.415Z.

Referenced by

Recorded citations in public readings. A reference alone does not establish agreement or influence.

No public references recorded here yet.

Reference data (JSON)

Limits

Text as retrieved on October 30, 2026; originally published May 29, 2024. The projected investment, cluster costs and power requirements are described by Aschenbrenner as rough, back-of-the-envelope calculations; they depend on trends and assumptions that may not hold.

Open question

How much of the projected buildout depends on assumptions about AI revenue, compute growth and the ability to build power infrastructure at the required pace?

Source-stated

Aschenbrenner argues that the race to AGI requires industrial mobilization, including investment in chips, datacenters and power.

The race to AGI won’t just play out in code and behind laptops—it’ll be a race to mobilize America’s industrial might.

Source-stated

Aschenbrenner projects that total annual AI investment could exceed $1 trillion by 2027, while presenting this as a rough estimate.

As revenue from AI products grows rapidly—plausibly hitting a $100B annual run rate for companies like Google or Microsoft by ~2026, with powerful but pre-AGI systems—that will motivate ever-greater capital mobilization, and total AI investment could be north of $1T annually by 2027.

Source-stated

Aschenbrenner argues that power, land, permitting and datacenter construction are more immediate constraints than spending for training clusters.

Perhaps the wildest part is that willingness-to-spend doesn’t even seem to be the binding constraint at the moment, at least for training clusters. It’s finding the infrastructure itself

Source-stated

Aschenbrenner argues that US power constraints can be overcome and that AGI datacenters should be built in America or close democratic allies for national security.

The power constraint can, must, and will be solved.

Source-stated

Aschenbrenner argues that locating AGI datacenters abroad could create security risks and leave AGI infrastructure subject to dictatorships.

Anything else creates an irreversible security risk: it risks the AGI weights getting stolen24 (and perhaps be shipped to China)

Source and contributor context

Leopold Aschenbrenner · Publisher

The publisher is identified here. No individual author has been inferred.

Source media

Media retrieval status: unchecked.

Terms and references

Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.

AI compute

The computing resources, meaning chips, data centers and electricity, used to train or run AI models.

Artificial general intelligence

The idea of AI with broad problem-solving, creativity and adaptability across different tasks.

Editorial provenance

Version 1. Publication ID: 25cfa822-2d3a-daec-3a69-286af095660d.

Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.

Source check: passed by anthropic/claude-opus-5.5 on 2026-11-18T12:39:12.098Z. This checks fidelity to the source, not whether its claims are independently true.

Editorial selection: accept; reasons: bridge_admitted_by_model.

Observation: e0736fcb7ebdf652eae9623bbc10ab7befa929fbc9362cee5ec1dcb9c9812b1b. Source text hash: acbdd456e73221b27e5d38f4494ad7d52e86241530701118fc91fd685308571b.

Revision and source-check history

Version 1 · 2026-11-18T12:43:10.989Z

Leopold Aschenbrenner argues in SITUATIONAL AWARENESS for an American AI-compute buildout

Leopold Aschenbrenner’s piece was published on May 29, 2024. He argues that the race to is also an industrial race: training clusters require chips, datacenters and power, and the infrastructure is already being built. His rough calculations project enormous investment and electricity needs if current trends continue. He sees power, permitting and datacenter construction—not willingness to spend—as immediate obstacles, and argues that US natural gas and regulatory changes could make the buildout feasible. He warns that building clusters in the Middle East could put infrastructure for training and running AGI under autocratic control, which he frames as an American national-security risk. His conclusion is forceful: “The power constraint can, must, and will be solved,” and the clusters should be built in America or close democratic allies. How much of the projected buildout can be supported depends on uncertain assumptions about AI revenue, compute growth and the pace of power infrastructure construction.

Original source

Leopold Aschenbrenner · Article

Source details

Original source date: May 29, 2024 (2024-05-29T22:21:37+00:00).

Date basis: Date stated on the publisher’s page.

Added to Escape Meridian: 2026-11-18T12:43:10.989Z. First observed: 2026-10-30T09:13:59.699Z. Retrieved: 2026-10-30T09:15:44.415Z.

Limits

Text as retrieved on October 30, 2026; originally published May 29, 2024. The projected investment, cluster costs and power requirements are described by Aschenbrenner as rough, back-of-the-envelope calculations; they depend on trends and assumptions that may not hold.

Open question

How much of the projected buildout depends on assumptions about AI revenue, compute growth and the ability to build power infrastructure at the required pace?

Source-stated

Aschenbrenner argues that the race to AGI requires industrial mobilization, including investment in chips, datacenters and power.

The race to AGI won’t just play out in code and behind laptops—it’ll be a race to mobilize America’s industrial might.

Source-stated

Aschenbrenner projects that total annual AI investment could exceed $1 trillion by 2027, while presenting this as a rough estimate.

As revenue from AI products grows rapidly—plausibly hitting a $100B annual run rate for companies like Google or Microsoft by ~2026, with powerful but pre-AGI systems—that will motivate ever-greater capital mobilization, and total AI investment could be north of $1T annually by 2027.

Source-stated

Aschenbrenner argues that power, land, permitting and datacenter construction are more immediate constraints than spending for training clusters.

Perhaps the wildest part is that willingness-to-spend doesn’t even seem to be the binding constraint at the moment, at least for training clusters. It’s finding the infrastructure itself

Source-stated

Aschenbrenner argues that US power constraints can be overcome and that AGI datacenters should be built in America or close democratic allies for national security.

The power constraint can, must, and will be solved.

Source-stated

Aschenbrenner argues that locating AGI datacenters abroad could create security risks and leave AGI infrastructure subject to dictatorships.

Anything else creates an irreversible security risk: it risks the AGI weights getting stolen24 (and perhaps be shipped to China)

Source and contributor context

Leopold Aschenbrenner · Publisher

The publisher is identified here. No individual author has been inferred.

Source media

Media retrieval status: unchecked.

Terms and references

Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.

AI compute

The computing resources, meaning chips, data centers and electricity, used to train or run AI models.

Artificial general intelligence

The idea of AI with broad problem-solving, creativity and adaptability across different tasks.

Editorial provenance

Version 1. Publication ID: 25cfa822-2d3a-daec-3a69-286af095660d.

Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.

Source check: passed by anthropic/claude-opus-5.5 on 2026-11-18T12:39:12.098Z. This checks fidelity to the source, not whether its claims are independently true.

Editorial selection: accept; reasons: bridge_admitted_by_model.

Observation: e0736fcb7ebdf652eae9623bbc10ab7befa929fbc9362cee5ec1dcb9c9812b1b. Source text hash: acbdd456e73221b27e5d38f4494ad7d52e86241530701118fc91fd685308571b.

passed · 2026-11-18T12:39:12.098Z · anthropic/claude-opus-5.5

Publication: 25cfa822-2d3a-daec-3a69-286af095660d

attribution — pass: The reading consistently attributes the arguments and projections to Leopold Aschenbrenner, the publisher and author of this single essay in the SITUATIONAL AWARENESS series. Projections are framed as his arguments, not established facts.

In this post, I’ll walk you through numbers to give you a sense of what this will mean

summary — pass: The reading leads with the author's central thesis that the AGI race is an industrial race. It covers his rough projections, the power and permitting constraints, his natural-gas and deregulation proposals, and his Middle East security warning. It preserves his forceful conclusion with an exact quote. AGI and compute are defined on the page, and training clusters are explained inline as needing chips, datacenters and power. The question follows from the source's own caveats about rough trend extrapolation and does not lead.

The race to AGI won’t just play out in code and behind laptops—it’ll be a race to mobilize America’s industrial might.

date — pass: The reading gives May 29, 2024, which matches the UTC calendar day of evidence.originalPublished (2024-05-29T22:21:37Z). The evidence is a publisher page, so no archive disclosure is needed. The reading has no hindsight framing, and the projections for 2026–2030 are presented as the author's forecasts, not outcomes.

support — pass: Each claim is supported by its excerpt: industrial mobilization, $1T+ annual investment by 2027, infrastructure rather than spending as the binding constraint, the power constraint to be solved, and security risks abroad. The America-or-allies framing matches the source. The closing caveat about uncertain assumptions reflects the source's own description of its estimates as rough and conditional on trends continuing.

The national interest demands that these are built in America (or close democratic allies).

media — pass: No media was selected. No verified media candidates were supplied, so selecting none is appropriate.

Challenge or correct this reading

Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.