National Security vs. Climate Is the AI Race’s Most Convenient Lie

The easiest way to shut down a climate argument is to rename it a security argument. Once AI is framed as a race against China, almost any energy choice becomes untouchable. Gas turbines beside data centers start to look like patriotism, and anyone asking about emissions sounds like they missed the briefing.

That framing is doing real work. It is clearing political space for tech companies to expand power-hungry infrastructure without owning the climate costs. It also hides a stranger fact: the country most often invoked as the reason to ignore climate constraints has been treating clean energy as a national security project for years.

Kate Crawford has put the contradiction plainly. The dominant story, she argues, says it is AI and national security or the planet. In that story, leaders will predictably pick national security every time. The framing sounds tough-minded. In practice, it lets companies avoid the harder question, which is how to build large-scale computing without locking in a dirtier and more fragile energy system.

The frame that lets companies off the hook

The national-security story works because it turns a policy choice into a loyalty test. If a company says it needs massive new electricity supply right now to train models, serve inference, and keep pace with geopolitical rivals, hesitation becomes suspicious. The debate moves fast. Grid planning, pollution, water use, local health, and long-term energy prices get recast as secondary details.

That is a gift to the companies building the demand surge.

If the conversation were honest, firms would have to answer awkward questions. Why should communities absorb methane generators, diesel backup fleets, new gas plants, and transmission strain so a handful of companies can chase model scale? Why should utilities socialize grid upgrades while private firms capture the upside? Why are some executives suddenly free-market absolutists until they need priority access to public infrastructure?

The security frame softens all of that. It tells the public there is no time for careful sequencing. Build first, decarbonize later. Connect the load now, clean it up eventually. Ask fewer questions because history is moving.

But the atmosphere is not impressed by urgency theater. A gas plant built for today’s data center demand can run for decades. Transmission lines, utility resource plans, substation expansions, and local air quality burdens all outlast today’s benchmark charts. Model rankings can flip in months. Energy infrastructure tends to sit there, billing and polluting, long after the press release has been forgotten.

This is why Crawford’s point lands so hard. The supposed tradeoff is not neutral analysis. It is a political narrative with beneficiaries.

China treated energy as strategy

The easiest mistake in this discussion is to imagine China as a climate saint. It is not. China still burns a huge amount of coal. It still approves coal capacity. Its total emissions remain immense. If you flatten the picture into “China went green,” you miss the real lesson.

The lesson is that China treated energy dependence as a strategic vulnerability and moved accordingly.

According to Crawford, China now has roughly 900 gigawatts of renewable power connected to its grid, and it added around 300 gigawatts of solar in 2024 alone. Those numbers are staggering even if you know the country’s scale. They tell you something deeper than environmental ambition. They show a state trying to reduce exposure to imported fuel, global price shocks, and geopolitical chokepoints.

That matters for AI because electricity is not an input you can hand-wave away. Training runs are costly, but the larger long-term issue is inference: millions or billions of routine queries, always-on serving infrastructure, and the compounding demand of enterprise use. If your electric system can produce a growing share of marginal power from wind and solar, the fuel cost on additional demand can get very low. Not zero in the magical sense—grids still need storage, transmission, balancing, maintenance, and backup—but low enough to change the economics.

China appears to understand that advanced computing and energy sovereignty belong in the same sentence. It made that connection more than a decade ago, partly because being one of the world’s largest energy importers was an obvious strategic weakness. If your industrial future depends on electricity, and your electricity depends on fuel you do not control, you do not have much autonomy. Clean generation was not only a climate move. It was industrial policy with a security rationale.

That is the part American debate keeps missing. China is not choosing between strategic competition and energy transition. It is using energy transition as part of strategic competition.

The American buildout is choosing the expensive path

In the United States, the direction of travel has looked very different. Utilities and developers are increasingly planning new gas capacity to serve data centers. Some of the most visible AI projects have leaned on on-site methane turbines because the grid could not deliver enough power quickly enough. xAI’s Colossus facility in Memphis became notorious for exactly this reason: the fastest path to computing capacity ran straight through fossil generation.

That is not an isolated quirk. It is a preview.

Developers are proposing campuses in the multi-gigawatt range. A 10-gigawatt data center complex is so large it starts sounding fictional, yet projects at that scale are now discussed with a straight face. Depending on reactor size, that is in the neighborhood of ten nuclear units’ worth of electricity demand. There is no painless way to meet load like that, and certainly no fast way if the surrounding grid is already constrained.

So the system reaches for what it knows. Gas. Temporary turbines that become semi-permanent. New pipeline logic. Permitting shortcuts. Deferred climate commitments. Local communities get told this is unavoidable because the future cannot wait.

The irony is that this is not the cheap path, and it is not the secure path. Fossil-heavy power is exposed to fuel-price volatility, supply constraints, physical delivery risks, and regulatory conflict. It also adds pollution burdens near facilities that are often located close to communities with little leverage over billion-dollar infrastructure decisions. Calling that arrangement strategic does not make it stable.

Even many industry leaders seem to know this. Privately and sometimes publicly, executives admit that clean energy is the only scalable long-term answer. They do not say that because they suddenly discovered ecological virtue. They say it because free fuel is hard to beat. Sun and wind are attractive for the same reason software likes automation: once the system is built, the operating economics improve.

The American problem is not a lack of technical awareness. It is the persistence of an ideological reflex that treats renewables as optional, slow, or politically decorative while treating fossil expansion as pragmatic realism. That might have passed for seriousness fifteen years ago. It looks stranger every year the economics improve and the grid reality becomes more obvious.

There is also a timing mismatch that rarely gets enough attention. Data center developers want power now. Building a robust clean grid requires transmission, storage, market reforms, interconnection queues that function, and regional planning with an actual horizon beyond next quarter. When policymakers refuse to speed that work and then point to fossil fuel as the only available option, they are solving a delay they helped create.

Climate damage already weakens military power

The national-security argument gets even thinner once you remember that climate change is itself a security threat. Military planners have been saying this for years, usually in language drier than the situation deserves. Rising heat, flooding, wildfire smoke, water stress, and severe storms degrade readiness, damage bases, complicate logistics, and increase instability in fragile regions.

This is not abstract. Runways buckle in extreme heat. Coastal installations face chronic flooding. Equipment fails sooner under harsher environmental stress. Human performance drops. Humanitarian crises multiply. Migration pressures grow. Supply chains become more brittle. A military can buy more autonomous systems and still be weakened by the climatic conditions those systems must operate in.

So when AI advocates imply that climate concerns are a luxury compared with security needs, they are sawing through the branch they are standing on. The same atmosphere warming your data center cooling loads is also reshaping strategic risk.

There is a narrower point here too. Fuel dependence has always been a military concern. Armies understand logistics with a kind of intimacy civilians usually encounter only during power outages. If your digital industrial base depends on abundant electricity, then the resilience, domestic availability, and price stability of that electricity are security issues in the plainest possible sense. You do not need a poetic definition of security to see it.

This is why the current American posture feels so upside down. It treats clean energy as a drag on strategic competition even though climate damage and fuel dependence are already degrading resilience. China may not be decarbonizing fast enough for the planet, but it has grasped a basic strategic truth: importing vulnerability and calling it strength is still vulnerability.

Efficiency is a competitive advantage

There is another sleight of hand in the race narrative. It assumes the winner is whoever burns the most electricity fastest. That flatters infrastructure spending, but it may be a poor guide to durable advantage.

If one country can build models that perform well, run cheaply, and demand less power per useful output, that country gains room to maneuver. Lower energy intensity means lower operating cost, less grid stress, easier deployment, broader adoption, and fewer political fights around each new facility. It also means your AI industry is less hostage to a massive buildout of generation that may arrive late and cost more than planned.

This is partly about algorithms and hardware. Better model architectures, smarter training techniques, improved chips, and more efficient inference stacks all matter. It is also about where those systems live. Software efficiency and energy infrastructure reinforce each other. Cheap clean power makes scale more affordable. Efficient systems make clean power easier to stretch across the economy.

That is why Crawford’s formulation matters: a country that builds models which work better, impose less damage on communities, and use less energy has an advantage in the market. Environmental performance is not a moral side quest attached to the main story. It is increasingly part of the product and part of the business model.

The same goes for public legitimacy. AI firms are asking society for enormous physical concessions: land, power, water, subsidies, fast permitting, grid upgrades, and tolerance for local disruption. If the visible return is higher bills, dirtier air, and a new wave of fossil infrastructure, public consent will not stay cheap. A cleaner path is not just better optics. It is a way of making the expansion politically durable.

The real choice is infrastructure, not patriotism

The phrase “national security versus climate” survives because it feels urgent and emotionally clean. It gives leaders a binary. It gives companies cover. It gives everyone a way to pretend that physical infrastructure can be wished into place without tradeoffs.

The actual decision is messier and more concrete. It is about what kind of electric system will underwrite advanced computing over the next twenty years. One path leans on gas, temporary fossil fixes, and the belief that climate consequences can be managed later. The other path treats clean generation, storage, transmission, and efficiency as core strategic capacity.

China’s example does not prove that everything is easy. Its grid still carries deep contradictions. Coal remains embedded. Regional imbalances are real. Renewable abundance does not automatically solve transmission bottlenecks or curtailment. But it does prove something the American debate keeps pretending is impossible: a state can pursue technological power and energy transition at the same time, for reasons that are as strategic as they are environmental.

That is what makes the current U.S. framing so self-defeating. By presenting clean energy as a drag on the AI race, it pushes the country toward an infrastructure base that is more expensive, more polluting, and less resilient. Tech companies benefit from the shortcut because it gets them capacity faster and shifts the long-term costs outward. The country does not.

A nation that powers its AI expansion with volatile fuel, strained grids, and local pollution is not choosing security over climate. It is building dependence into the foundation of its most important industrial project.