OpenAI warns US to close electron gap with China, build 100GW of energy capacity [pdf] (cdn.openai.com)

OpenAI sent a public memo to the White House OSTP urging the U.S. to “close the electron gap” with China by targeting 100 GW of new power capacity per year. Citing internal and commissioned analyses, OpenAI argues that energy is now the binding constraint on AI leadership: China added 429 GW of capacity in 2024 versus the U.S.’s 51 GW, and OpenAI’s own buildout (six Stargate sites totaling ~7 GW planned and $400B in near-term investment) will require large shares of skilled trades and grid capacity. The company estimates the first $1T of AI infrastructure could add >5% GDP over three years and says it’s on track to meet a $500B/10 GW commitment by end-2025 while also deploying behind‑the‑meter gas/solar and grid-connected projects. The memo frames electricity as strategic infrastructure and lays out concrete technical and policy implications: accelerate transmission construction, shorten interconnection queues, expand domestic manufacturing of transformers, HVDC converters, switchgear, cables and semiconductors, and scale workforce training (electricians, mechanics, ironworkers). OpenAI recommends expanding tax credits like AMIC, using authorities such as DPA Title III and DOE Loan Programs Office to de‑risk manufacturing, and modernizing permitting to unlock capacity. For AI/ML practitioners and policymakers, the memo signals that compute scaling is now tightly coupled to national energy policy, supply‑chain resilience, and large‑scale grid investment.