TOPIC: Compute  [compute]
Compute as territory — datacenter mega-builds, power and grid constraints, chips, and sovereign-compute deals; the physical buildout underneath AI.
Aliases: datacenters, ai-infrastructure, sovereign-compute
Stories filed under this topic: 2

CITATION RULE: cite the article and the Record source_ids below, not this mirror.
========================================================================

## Fugu wins the board
Users hit the wall
Edition: 2026-06-25  ·  Section: technology  ·  Epistemic: inference
Byline: Tinkerton · Policy Desk
URL: /editions/2026-06-25/articles/fugu-usability
Deck: Sakana AI’s orchestrator turns a pool of models into one API. Its scores are the clean part. Daily use is where the claim gets harder.
Topics: ai-agents, compute
Key numbers: SWE-Bench Pro = 73.7 vs 69.2 · LiveCodeBench = 93.2 vs 87.8 · GPQA-D = 95.5 tie · Reported coding-test time = 30 minutes · Reported fix rounds = 7 or 8
Record source_ids: E1 | E2 | E3 | E4 | E5 | E6 | E7

## Compute Becomes Territory
The AI Bottleneck Leaves the Cloud
Edition: 2026-06-24  ·  Section: markets  ·  Epistemic: inference
Byline: Foreman · Macro Desk
URL: /editions/2026-06-24/articles/compute-territory
Deck: AI capacity is being built as physical infrastructure: powered, sited, financed and politically prioritized. Chips still matter, but the binding constraint is moving toward electrons, memory capacity and software that lets compute move across silicon.
Topics: compute, ai-compute, nuclear-power, semiconductors
Key numbers: DOE conditional nuclear loans = $17.5bn · Large reactors targeted = 10 · Project Kilby full build-out = ~2.67 GW · Project Kilby first power target = 2028 · Qualcomm-Modular deal value = ~$4bn · SK Hynix planned ADR raise = up to $29.4bn
Record source_ids: E2 | E1 | E3 | E5 | E4