How AI Is Rewriting America’s Energy Future
A Key Driver of Long-Term Success for AI and Non-AI Startups
From Stability to Strain: The New Energy Reality
After more than a decade of relative stability, U.S. electricity demand is now climbing sharply. In 2024, national electricity use rose by 3%—the largest annual increase since the Great Recession—with projections calling for 3% growth in 2025 and 5% in 2026.
This dramatic shift is driven by a surge in U.S. industrial growth, rising commercial demand, and, most significantly, the escalating energy requirements and costs associated with AI and data centers. Regional grid planners are sounding the alarm, warning that peak electricity demand could grow by 10% to 54% by 2050, depending on geography.
The most intense pressure is coming from the AI sector itself. In 2023, data centers accounted for 4.4% of U.S. electricity use. By 2028, that number could rise to between 6.7% and 12%. With data center electricity consumption projected to grow at a compound annual rate of 23%, the sector could soon surpass manufacturing in total energy use—potentially driving nearly half of all electricity demand growth by 2030.
How Cloud Providers Are Securing the AI Future
To get ahead of this trend, major cloud providers are racing to secure reliable, long-term sources of power for their AI infrastructure. Their strategies reveal a decisive shift toward nuclear and renewable energy investments:
Amazon acquired a data center powered by Talen Energy’s nuclear station (March 2024) and invested $500 million in X-energy to develop over 5 GW of small modular nuclear reactors (SMRs) by 2039.
Alphabet has taken a diversified approach. It is backing a 300 MW solar pipeline in Taiwan (July 2024), partnering with Energix on a 1.5 GW solar supply (August 2024), and collaborating with Kairos Power to secure 500 MW of nuclear capacity by 2030. Additional investments include $208 million with Intersect Power for renewables (December 2024), geothermal PPAs in Taiwan with Baseload Capital (April 2025), and a partnership with Elementl for advanced nuclear development (May 2025).
Microsoft is doubling down on nuclear innovation. It partnered with Helion to pursue fusion power by 2028 (May 2023), signed a $10 billion renewable deal with Brookfield (May 2024), and secured a 20-year agreement to restart the Three Mile Island nuclear plant to support its AI workloads (September 2024).
These aggressive energy procurement efforts underscore a broader concern: America’s energy infrastructure is under growing strain. Grid operators and regulators, including NERC, have warned of elevated blackout risks, especially during peak summer months. The combination of surging demand, aging transmission systems, and the accelerated retirement of coal and gas plants is putting serious pressure on reliability. Adding new capacity is no easy fix—interconnection delays, permitting bottlenecks, and multi-year lead times for critical equipment like transformers are slowing progress.
Impact on Startups
The surge in data center demand and rising energy consumption—driven largely but not exclusively by AI—are reshaping the landscape for startups and beyond. As cloud providers prioritize large-scale customers, startups face mounting challenges: higher compute costs, limited GPU access, and longer provisioning times. These pressures are pushing teams to optimize for energy efficiency and cost, while also prompting investors to scrutinize infrastructure strategy alongside go-to-market plans. Beyond financial strain, startups are increasingly expected to demonstrate sustainable AI practices, including clean energy sourcing and carbon transparency.
Yet these challenges also present opportunities: startups that improve energy efficiency, innovate in green AI, or enable more cost-effective deployment models can stand out. Decisions about geography, infrastructure, and partnerships—once backend concerns—are now strategic imperatives in a power-constrained digital economy that affects all computing, not just AI workloads.
(Interestingly, one cloud services executive shared anecdotal evidence from both the U.S. and EMEA suggesting that startups are beginning to push back—albeit gently—on customer inquiries about green AI and broader sustainability efforts.)
Conclusion
The AI era is ushering in a new kind of arms race—one not just for talent, models, or funding—but for electricity itself. (The cloud services executive mentioned above said that, so far, he hasn't seen startups struggling to access compute for their workloads—with the exception of some Oracle Cloud Infrastructure customers. It may be something worth watching going forward.)
As power-hungry data centers multiply and strain a grid already hampered by aging infrastructure and slow capacity expansion, the ability to secure clean, reliable energy is becoming a defining factor for success. While tech giants are making bold energy investments, startups are grappling with rising cloud costs and funding pressures—driven in part by emerging constraints on compute and energy availability and pricing. Whether you're building the next foundation model or enabling its deployment, one thing is clear: in tomorrow’s innovation economy, energy is the new currency.
If you know a tech company or startup grappling with rising costs for cloud, compute, or related opex, feel free to pass this post along—it may offer helpful perspective.