Oklo and X-energy join the US nuclear push for AI: the reactor pipeline meets the data-center pipeline

Oklo et X-energy officialisent leur adossement au push nucléaire US pour l'IA. Les réacteurs modulaires (SMR) passent d'idée de deck à ligne de capex.

In plain terms - Oklo and X-energy - two American SMR (small modular reactor) companies - have joined the US federal push to plug AI's power demand with new nuclear (Tech in Asia, 22 July 2026). This is the moment where the SMR pitch meets the data-center capex line, on paper.

Contexte

AI power demand has moved from a modelling assumption to a grid-planning reality in 12 months. New York State's Q2 moratorium on data-center construction (fil ai-power-wall, publi #1097) was the political warning shot. In parallel, DOE has been aligning federal loan guarantees and nuclear supply chains with hyperscaler capex. Oklo (Aurora reactor, sodium-cooled) and X-energy (Xe-100, high-temperature gas-cooled) are the two most-funded US SMR players.

Under the hood

• Oklo : ~15 MWe Aurora design, 20-year fuel load using HALEU (high-assay low-enriched uranium). NRC combined license path.
• X-energy : 320 MWe Xe-100 (4 × 80 MWe modules), TRISO fuel. Backed by Amazon (data-center anchor), Dow (industrial process heat), and Energy Northwest.
• Both are US-supply-chain SMRs - a key criterion for federal loan guarantees and DOE offtake mechanisms.

Analyse

The reason SMRs are back in the AI conversation is a mismatch of build timelines: hyperscaler data centres commit MW in 18-36 months, gas peakers take 3-5 years, gigawatt-class nuclear 8-12 years. SMRs promise 4-6 year builds at 15-320 MW. That fits inside a data-center campus's second phase, not its first - which means SMRs enable expansion pipelines that are otherwise grid-limited.

The catch is well-known: no US SMR has yet delivered first commercial power. Oklo's Aurora is targeted for 2028 at Idaho National Lab; X-energy's Xe-100 is aimed for 2030 at Dow Seadrift. Anything on paper today is a 2029+ MW.

For a CFO or infra investor, the immediate signal is not "power problem solved". It is that federal capital and federal siting are aligning behind US-supply-chain SMRs, which prices in for the first time. Land-cost, grid-connection cost, and long-lead-item queue positions are the trades that move now.

Scénarios

• Base case (55%) : Oklo Aurora delivers first power 2028-2029; X-energy 2030-2031; hyperscaler PPAs signed but volume small.
• Slip (30%) : NRC licensing pushes both by 12-18 months; hyperscalers backfill with gas + grid firming.
• Breakthrough (15%) : DOE offtake + federal loan guarantees compress project finance costs enough that a second wave of orders lands before first power.

Implications

For a data-center CFO: SMR PPAs are still speculative - plan around gas and grid firming until 2029. For an infra investor: the sub-tier suppliers (HALEU fuel, TRISO, forgings) are where the pinch already is.