In 2020, it seemed unlikely that nuclear would make a comeback in the West. Nuclear energy's window was closed, Beijing was making moves, and Gates's TerraPower was blocked. America was in a 20-year holding pattern, and Germany actively planned decommissioning its nuclear facilities.
nuclear stagnation
In 2023, GPT-4 polarized tech and policy, with radical lunatics calling for the bombing of data centers. Ideological capture migrated from boardrooms to policy drafts, leading to the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence EO the day before Halloween.
Now, in 2025, the doomer narrative vacillates between energy and water consumption misinformation to massive job losses. But, America's AI Action Plan marks a 180º policy reversal and is a public blueprint for the global competition of this decade. It acknowledges a mindset shift from a world of constraints to one of abundance through expansion of energy, education, and infra.
It is up to Congress to allocate and the agencies to approve, but the Washington stance swaps risk management for a global industrial-competition frame, betting that speed plus targeted export controls beats sweeping regulation.
Canada is a rounding error, UK is behind the curve, the EU and Israel have no viable market competitor (unless, you include SSI, which should not be underestimated), the Middle East embraced American innovation after an attempt, and China continues its love of open source technology to avoid tech transfer barriers. Without China, BRICS is absent and, most surprisingly, India has no serious infrastructure or strategy.
gpqa by country
Perhaps the UK Parliament should consider unwinding the DeepMind acquisition as their hope for competition (note: sarcasm). China's progress is seemingly proof that the model collapse hypothesis from training on synthetic data is false.
It will be unsurprising to see emerging economies once delivering human-labor arbitrage (Phillipines, Nigeria, Pakistan, Vietnam) swapping for energy or infrastructure arbitrage in this intelligence era.
Five years ago, championing nuclear was heretical and LLMs could only complete sentences. In 2025, we can't bring enough nuclear online quickly enough and LLMs compete against world-class mathematicians. In 2030, we will look at this Action Plan as the political starting line for global realignment around on-demand intelligence, the infrastructure required to serve it, and the foot-gun regulations attempting to save its citizens from Judgement Day.