The AI Race Is Becoming A Sprint For Power

Does the U.S. need to do more than build AI Data Centers to compete globally?

Deposit Photos

Is America engaged in a new arms race?

Leading experts say yes, only this time it’s a high-stakes competition with China. And as Financial Times reports, it’s not clear if America will prevail. “US confidence in its technological lead has been rattled by China’s DeepSeek, an AI model apparently developed more cheaply and with far less computing power than US counterparts.”

That article appeared in February. Since then, the new administration has pressed “go” on its technological ambitions, accelerated in part by the DeepSeek debacle that wreaked havoc on the American stock exchange.

The need to build more AI Data Centers is central to America’s strategy. “Data centers from the pre-AI era are increasingly ill-suited for the demands of today’s AI workloads…” according to GoldmaSachs.com. “… In 2022, a cutting-edge AI system integrated eight GPUs into a single server. By 2027, the leading system will likely have 576 GPUs in a filing-cabinet-size rack, requiring a whopping 600 kilowatts (kW), enough to power 500 US homes.”

But are there costs to producing so many AI Data Centers? That answer is a decisive yes, depending on who you ask. In September, Bloomberg.com ran a story detailing how the rush for AI global hegemony is harming local residents, reporting on how wholesale electricity costs are up as much as 267% in some areas. “About two-thirds of the power consumed in the US runs off of a state or regional grid, where the system operator manages the trading of energy. These wholesale commodity costs are passed on to households and businesses on their utility bills, which then include other charges to maintain and expand the network. That can affect even customers who aren’t in close proximity to a data center, since their energy relies on the same grid.”

Why Does AI Use So Much Energy?

When it comes to understanding the AI arms race, it’s important to ground any discussion with context about AI power needs. PennState Institute of Energy and the Environment offers just one dimension of such herculean demands. “AI, particularly large language models (LLMs), requires enormous computational resources. Training these models involves thousands of graphics processing units (GPUs) running continuously for months, leading to high electricity consumption. By 2030–2035, data centers could account for 20% of global electricity use, putting an immense strain on power grids.”

Importantly, LLMs do not represent the only application of AI. Public conversations can center on LLMs because so many of us use them, both in our personal and professional lives but they represent only a fraction of AI energy requirements.

Here are five more critical categories:

1. AI Inference at Scale

Right now, AI is responding to questions all day, every day, like “Why is the sky blue?” This digital Socratic Dialogue requires ever more massive computer centers running 24/7 to supply curious people like you and me with dependable answers. Without such constant power streams, the Q&A system we’ve come to depend on the last few years could fail.

2. Multimodal AI

Humanity’s endless questions, needs, requirements, prompts and creative artifacts don’t just deal with mere text inputs alone. Users increasingly turn to AI for images, audio, and video, not just text. As just one example, OpenAI’s Sora depends on such computing. Likewise, AR/VR runs continuously on gargantuan power sources.

3. AI-Driven Research

Just because less people are seeking answers via “Googling something” doesn’t mean search has gone away. Instead, society is busy replacing keyword search via generative answers, the kind you find with LMO. But that only increases the compute required by so much endless querying. Bottom line: search isn’t going away. It’s becoming more energy-intensive. Now try imagine what happens if/when every search becomes “AI search.”

4. Autonomous Vehicles and Robotics

This month, Waymo announced its fleet of self-driving cars is highway-ready. While that may be a boon for those wishing to ditch their cars for a hands-off commute, it means even more strain on power grids. If anything, we can expect such energy demands to explode in the coming years as fully self-driving truck fleets emerge, more warehouse robots appear, and even autonomous military applications skyrocket. This leads us to our next category.

5. Security and Defense Systems

Last month, Defense One reported how just one of the Armed Services branches, the Air Force, has signaled extraordinary infrastructure needs to implement AI readiness. “Air Force officials want private companies to build artificial intelligence data centers on more than 3,000 acres of land on five of its military bases, raising questions about security, ethics, and land use.” This is but one example of the many uses of AI we can expect for national defense in the coming years, especially as widespread military applications come further online, including drone warfare, satellite sensing, and global threat detection.

The Space Race, Revisited

Growing up in the 1980s, it was common to see bumper stickers pushing peace messages like “Arms Are For Hugging.” The notion of Mutually Assured Destruction (MAD) was meant to be comforting, that the U.S. and the U.S.S.R. would decide not to nuke the other for fear of enemy reprisal.

This time around, the global stakes appear to be much, much higher. Consider what may happen if one country achieves artificial superintelligence (ASI) before its rival. That nation would possess intellectual supremacy eclipsing its competitor. Or to use an analogy it would be like employing modern machine guns against an army armed only with wooden shields.

Frederick Kempe of The Atlantic Council doesn’t mince his words in his own chilling estimation of this moment’s importance: “The world has entered the most consequential tech race since the dawn of the nuclear age, but this time the weapons are algorithms instead of atoms. Rather than a race to obtain a single superweapon, this is one to determine how societies think, work, and make decisions. AI is transforming not only the distribution of power around the globe but also the very nature of that power and how it will be exercised.”

Again, as discussed, the costs to build and sustain the power required to operate America’s artificial intelligence boggle the mind. And they’re growing daily. But this doesn’t mean such expenditures aren’t essential for our nation’s long-term competitiveness, not just militarily, but economically. Historical antecedents like the waning of Great Britain, reveal all too starkly what can happen when once great nations cede their technological ambitions.

“The sun never sets on the British empire,” was a truism until it wasn’t.

If the United States wishes to compete on an ever-crowding world stage, we mustn’t succumb to despair as to what’s required of us, both energetically, and as a matter of national character. Massive investments in AI Data Centers represent just the tip of what is needed for America to compete. But not all.

More than ever, we need a story to rally behind, one that inspires Americans not to accept the status quo as the best we can do, but instead to draw on our past to envision a future that’s bigger and brighter. For if we meet the coming days with the same ingenuity that carried us through the last global race, we won’t just achieve our power requirements; we’ll hand down a world worthy of our children and all those yet to come.