The Real AI Race Isn't About Algorithms. It's About Electricity.
By Abhilash Mishra
There's a saying in Silicon Valley that artificial intelligence is the new electricity. It's meant to capture AI's transformative potential, but it might be more literal than that. The key constraint on AI development isn't turning out to be algorithmic breakthroughs or talent or even regulation – it's actual electricity. And that reality is forcing a profound rethinking of how we approach both AI development and climate policy.
This is the context for understanding President Biden's latest executive order on AI infrastructure, which might be his most consequential AI policy legacy. Quietly released in the closing days of his first term, it represents a fundamental shift in how we think about AI policy – and perhaps, more importantly, how we think about the relationship between technological progress and environmental sustainability.
The numbers are staggering: AI data centers currently consume about 2% of US electricity production. By 2030, that could rise to 10%. To put that in perspective, that's roughly equivalent to the current electricity consumption of a medium-sized state. This isn't just a technical problem; it's a civilization-scale challenge.
What makes Biden's order fascinating is how it reframes the AI race. For the past few years, we've been trapped in a debate between AI accelerationists who want to push development as fast as possible and AI safety advocates who want to slow things down. But this debate misses the most practical and immediate-scale challenges to AI development and misses the real bottleneck: we literally might not have enough power to run the AI systems we're capable of building.
The order takes three approaches to this problem, balancing private sector incentives, incentives for clean energy deployment, and consumer protection from high energy costs. First, it opens federal lands for data center construction. This isn't just about real estate – it's about the government accepting a direct role in physical AI infrastructure, something that would have been unthinkable just a few years ago. It's a recognition that some problems are too big for the private sector to solve alone.
Second, it mandates clean energy integration for these data centers. This is where things get interesting. The administration is essentially betting that the urgency of AI development can be harnessed to accelerate the green energy transition. It's a bit like using a fever to fight an infection – turning one challenge into a solution for another.
Third, it includes provisions to protect consumer energy prices. This might seem like a minor technical detail, but it reveals something profound about the political economy of technological change. The administration has recognized that the sustainability of AI development isn't just about environmental sustainability – it's about political sustainability. If AI development drives up energy prices for ordinary Americans, the political backlash could derail both AI progress and climate action.
But there's something deeper happening here that deserves our attention. This order represents a shift in how we think about the relationship between technological progress and environmental protection. For decades, we've treated these as competing priorities: you could have rapid technological development, or you could have environmental sustainability, but you couldn't have both.
This order suggests a different possibility: that the physical constraints of AI development might force us to solve our energy problems faster than we otherwise would. It's a version of what the economist Albert Hirschman called "binding constraints" – sometimes, having less flexibility forces better solutions.
Consider what this means in practice. Every major AI company is now essentially forced to become a renewable energy company as well. They have no choice – the economics of running massive data centers on fossil fuels simply won't work in the long run. This creates powerful new constituencies for clean energy development and grid modernization.
There's a historical parallel worth considering. In the early 20th century, the rise of the automobile industry created enormous environmental problems. But it also created constituencies and technologies that eventually enabled environmental solutions. The companies that learned to make efficient engines for cars ended up making efficient engines for everything else too.
Could we be seeing something similar with AI? The technologies developed to make AI data centers more energy efficient might end up making everything more energy efficient. The political pressure to build clean power for AI might end up accelerating the clean power transition for everyone.
This brings us to the broader question this order raises: What does effective governance look like in an age of exponential technological change? The traditional model of regulation – where government tries to constrain private sector activity – clearly isn't sufficient. But neither is the libertarian fantasy of letting technology develop without any guardrails.
What Biden's order suggests is a third way: using the government's convening power and physical resources to shape the direction of technological development, without directly controlling it. It's not about saying "no" to AI development or even "slow down" – it's about saying "this way."
The clever part is how it aligns incentives. Instead of trying to force AI companies to care about climate change, it makes solving climate change a precondition for their own growth. Instead of treating environmental protection as a constraint on technological development, it makes it an enabler.
This approach isn't perfect. There are legitimate questions about whether we can build clean energy capacity fast enough to meet AI's growing demands. There are concerns about land use and about whether this level of AI development is desirable even if we can power it sustainably.
But what makes this order important is how it reframes these challenges. It suggests that the key to solving our biggest problems might not be choosing between competing priorities, but finding ways to make them reinforce each other. It suggests that the path to sustainable AI development might be the same as the path to a sustainable planet.
That's a profound shift in how we think about both technology policy and environmental policy. And it might be Biden's most important legacy in both areas. Sometimes, the most important policy innovations aren't about solving problems directly – they're about changing how we think about what the problems really are.
