The race to win AI won’t be settled by who builds the smartest model. It’ll be settled by who controls the power to run it. That’s the throughline in a new piece from The Information, which reaches back more than a century to Thomas Edison to explain who’s positioned to win both the AI and battery races unfolding right now.
The historical hook is a good one. Edison is remembered as the inventor of the light bulb, but his real genius was building the entire system around it: the power stations, the grid, the distribution. As The Information frames it, the lesson from Edison’s era is that raw invention rarely wins on its own. The player who controls the infrastructure and can scale it wins. And in 2026, the infrastructure question for AI is energy.
Why energy is now the AI bottleneck
Compute gets the headlines, but power is the real constraint. Training and running frontier models eats electricity at a staggering rate, and data center demand is straining grids across the US. The companies racing to deploy AI at scale have run into a wall that has nothing to do with chips or algorithms. They simply can’t get enough reliable power fast enough.
That’s where batteries enter the story. Grid-scale storage smooths out the gap between when energy is generated and when data centers need it. Whoever masters cheap, dense, scalable storage gains leverage over the whole AI buildout. It’s the modern version of Edison wiring up Manhattan.
What stands out here is the convergence. The AI race and the battery race used to be separate stories. They’re now the same story.
The competitive picture
This reframes who the real players are. It’s not just OpenAI, Anthropic, and Google trading model benchmarks. It’s the hyperscalers signing nuclear deals, the utilities scrambling to add capacity, and the battery makers, many of them Chinese, who dominate manufacturing today.
That last point matters for anyone watching the geopolitics. If batteries are core infrastructure for AI, then battery supply chains become a national security question. It connects directly to the wave of protectionist moves we’re seeing, from tariffs to restrictions on foreign hardware. Control the storage, and you hold leverage over the AI economy your rivals are trying to build.
What this means for practitioners and businesses
A few practical takeaways from where things stand:
- Treat energy as a strategic input, not a line item. If your AI roadmap assumes cheap, unlimited compute, stress-test it against power availability and cost. Those assumptions are getting shakier.
- Watch the infrastructure players, not just the model labs. The companies locking in power deals and storage capacity today are buying an advantage that compounds. That’s where a lot of value will accrue.
- Factor supply chain risk into hardware planning. Battery and grid components are becoming politically sensitive. Sourcing decisions you make now could get complicated by regulation later.
- Think in systems, like Edison did. The winners won’t be the ones with the single best component. They’ll be the ones who stitch compute, power, and storage into something that scales.
The bigger pattern
The Edison comparison works because it’s a reminder that technology waves reward integrators over inventors. The steam engine, the railroad, electricity, the internet. In each case, the lasting fortunes went to whoever built and controlled the infrastructure layer, not just whoever had the first bright idea.
AI looks like it’s following the same script. The model breakthroughs are real and they matter. But the durable advantage is shaping up around energy and storage, the unglamorous plumbing that makes everything else run.
Over the next few years, expect the AI conversation to shift steadily from model capabilities toward power contracts, grid capacity, and battery output. The companies that see that early, and act on it, are the ones writing Edison’s next chapter. You can read the full argument at The Information.