The US is winning the AI race where it matters most: commercialization. Since DeepSeek R1 shocked the market in January 2025, American companies have moved faster. OpenAI pushed harder into agents and Codex. Anthropic turned Claude Code into a business. China has contenders, but the US is clearly ahead in revenue, adoption, tools, and reach.

Trump fits this moment well. He is a salesman at core, and Larry Ellison is too. That helps explain why AI infrastructure is an easy political product. Selling AI today is easier than selling Oracle databases in the 1980s. This time the oracle talks.

DeepSeek matters for a different reason. Its strategic value for China is not mainly commercial. It helps China reduce dependence on Nvidia and push inference toward domestic stacks such as Huawei Ascend. That supports supply chain autonomy. It is not the same as profitable AI leadership.

Christian Klein of SAP has argued that Europe does not need more data centers and that large language models alone are not enough. He is right that models alone are not enough. Europe also spent about $58.8 billion on Indian software services in FY 2023 to 2024 and about $67.1 billion the next year. AI only becomes valuable when it is tied to real data, real workflows, and real products. But his broader view misses the main fact. The United States is winning because it is building every major layer at once: chips, power, data centers, cloud platforms, developer tools, consumer platforms, and enterprise software.

Many people use the wrong scorecard. Papers and engineer counts do not prove AI leadership. The test is who can finance infrastructure, train and serve models at scale, and apply AI across the economy.

Energy is part of that lead. Modern GPU and TPU systems turn electricity into compute. Cheap power lowers model costs. That is why electricity prices matter.

The US is cheaper than the big Western European economies. Canada is cheaper still. China and Russia are lower cost than the US in this comparison. So power matters. But power is not the most important layer.

The decisive layer is cloud infrastructure and data. The US owns the global hyperscalers. AWS, Azure, and Google Cloud give American firms the main channels through which models reach the world. It also owns platforms that generate and organize the data of the AI age. YouTube is a video corpus. Google Drive and Microsoft 365 sit inside daily office work. GitHub sits inside software development. These are distribution systems and data platforms. New models can be pushed into products people already use every day.

That is why electricity alone does not decide the race. A country can have cheap power and still lose if it does not have cloud scale, platform reach, developer ecosystems, and access to large flows of useful data. The US has all of that at once. China has much of it in its large domestic market. Europe does not.

Europe has long had strong engineering talent. But talent is not enough. US hyperscalers already dominate the market, and catching up is slow. Even if Europe decided today to finance real cloud champions, building the infrastructure would only be the first step. Europe would then need to move banks, manufacturers, and public agencies onto those platforms. That process would take most of a decade. By then AWS, Azure, and Google Cloud would be even further ahead in scale, software, and data.

There is one exception. Arkady Volozh is trying to build Nebius into a European AI infrastructure company. But that confirms the rule. Europe is still at the start.

So Klein is right that LLMs alone are not enough. But the lesson is not that data centers matter less. It is that data centers matter inside a much larger system. The US is winning because it has power, capital, cloud infrastructure, and data platforms all working together. Energy is important. Cloud and data are even more important. That is where the American lead is strongest.

There is another frontier: weaponized AI. The next phase may be Country X AI versus other countries' AI in bot networks, cyber campaigns, and autonomous weapons. A provider does not need magic to do this. It is disturbingly easy to tune systems to dehumanize rivals, justify violence, or target entire populations. Once models are embedded into media, networks, and weapons, bias becomes force. The AI race is also a security race.

Models like Anthropic's Mythos point to another shift. The old Linux instinct was many eyes on open code. Frontier cyber models may push states and defense firms toward the opposite logic: security by obscurity, with closed software, closed tooling, closed firmware, and closed chips. If a model cannot train on the code and architecture of a target stack, it will usually have less context and less speed. That does not make systems safe, but it does raise the value of proprietary stacks all the way down to hardware.