Topic
Nuclear Power
Atomic energy as AI infrastructure — reactor financing, the power buildout to feed data-center demand, and the grid as a national-security variable.
- Danube fact Aug 23, 2026Paks Unit Three Is Back on the GridMTI said Paks Unit 3 turbine No. 6 was synchronised late Saturday and began producing. Péter Magyar said four of eight turbines were running on Sunday and that the plant could reach 2,000 MW by Wednesday. Cernavodă, on the same river, did not reconnect. That 23 August call settles NO.
- Cooling Margin inference Aug 13, 2026Danube Shuts Cernavodă’s Second UnitUnit 2 crossed its river operating threshold and shut in a controlled sequence, leaving roughly 1.3 GW of Cernavodă capacity unavailable. Romania kept the grid balanced, but the public record still lacks a final ledger for replacement volumes and prices.
- Danube Cooling Squeeze inference Jul 30, 2026Low Water Rations the ReactorsPaks may have to shut within 24–72 hours. Cernavodă Unit 1 is down while Unit 2 remains connected, turning the Danube’s cooling margin into a cross-border replacement-power bill.
- Criticality at zero inference Jul 28, 2026Four Cores Cross the Line at Zero PowerFour privately developed microreactor designs achieved DOE-authorized criticality in thirty days. Licensing, power ascension and the first useful watt remain ahead of every customer.
- Open-weight pressure inference Jul 2, 2026Korea, US Race to Site AI ComputeSeoul and Washington are steering billions toward power-rich land, permitted generation, and domestic chips while China's low-cost open models narrow the frontier gap. Semiconductor stocks fell on 2 July as investors weighed megawatt megaprojects against cheaper inference anyone can download.
- The framework verification test inference Jun 26, 2026Same framework Different obligationsWashington is selling speed. Tehran is preserving conditions. The first test is whether inspectors reach the damaged sites.
- Compute the power inference Jun 24, 2026Compute Becomes Territory The AI Bottleneck Leaves the CloudAI 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.