Artificial intelligence (AI) is advancing faster than the network that powers it. As Microsoft CEO, Satya Nadella recently warnedthe real limit of AI isn’t the chips… it’s the electricity. Energy has become the main bottleneck in AI, pushing tech giants to create their own energy source and reboot all available energy sources. Whoever gains this power advantage will win the AI race.
For the United States, responding to the insatiable appetite for AI constitutes the greatest systemic risk of the next decade. America needs a massive expansion of power plants, transmission lines, and advanced hardware, all while using AI itself to advance the grid and optimize electricity distribution.
COMMENT
In 10 years, utilities, policymakers and large electricity users will be judged on a single criterion: whether we closed this gap between supply and demand on time. If America loses the global race for AI leadership, it won’t be because our chips or models were inferior, but because we couldn’t bring them online. Six AI and grid trends will decide the outcome of this race.
Trend #1: Power is the new AI bottleneck
In the AI economy, it’s no longer just about computation. Businesses need access to reliable and affordable energy.
Data centers already consume 3 to 4% of American electricityand their share of the pie is expected to reach 11-12% by 2030. AI workloads are fundamentally more power-intensive than cloud architecture utilities predicted a decade ago, with AI queries requiring more compute (and therefore power) than traditional search.
Utilities are facing a new reality: integrated resource plans, grid modernization programs, and capital expenditures will be tested against a high-density and ever-increasing AI load. Those who cannot provide reliable, scalable energy will see hyperscaler investments go elsewhere.
Trend #2: Bring Your Own Power
For data center planners, the central question is no longer “How many racks and servers do we need?” » to “How many megawatts, where and when?” » There is already a structural mismatch between demand and deliverable capacity, with interconnection queues building up for years.
If hyperscalers can’t connect quickly, they will skip the queue and create their own build behind the fence. Companies like Meta, Microsoft and Google sign long-term deals on fusion, nuclear and renewable energy electricity offersand many develop or directly finance production to secure an advantage. AI is now effectively about “bringing your own power.” For these companies, it is also a question of “transforming this energy into profit” by reselling surplus electricity on the markets.
Trend #3: Every electron wins
The old battle between “good” and “bad” electrons – renewables versus fossils, the race to net zero versus incrementalism – is collapsing in concrete energy supply decisions. Now, carbon intensity and even cost matter less than the most basic question: “When can I get it?”
As Energy Secretary Chris Wright said at the January 15 briefing State of the Energy Industry Forum“We need anything that will productively add to the energy stack, that is, increase our capacity and reduce our costs…The source doesn’t matter to me, but the impact of the source matters to the world.”
We are seeing a shift in national policy where every credible energy technology has a place at the table, and coal does not disappear. For emerging economies, this remains the fastest lever for industrial growth. Natural gas will continue to grow, perhaps more than forecasts suggest.
Nuclear fission is about to develop, double or even triple in the world by 2050 and become a sovereign wealth fund business for the US government. Fusion has been elevated by the U.S. Department of Energy (DOE) as a strategic marketing prioritysupported by new science, technology and supply chain initiatives.
Solar power and batteries will continue to dominate new construction, supported by a vast favorable terrain and modular, affordable and rapidly deployable technology.
The regulatory playbook is evolving to shorten timelines and reduce delays, with federal permitting reforms and new legislation advancing the manufacturing and upgrades of grid components. These changes will be contested and litigated, but over time, projects that can be built and integrated quickly will prevail.
Trend #4: Component supply chains are the real ceiling limiters
Even where politics, economics, and community sentiment align, the sad truth is that we lack specialized networking hardware. Turbines, power transformers, switchgear and capacitor banks now have multi-year delivery timesand major OEMs are effectively sold out until the end of the decade. Further threatening supplies, China controls the manufacturing and processing of critical materials for many of these components.
Capacitors are a telling example. They sit quietly inside almost every energy asset: substations, power electronic converters, HVDC links, SMR and fusion systems, EV chargers, etc. In the United States, capacitor films have historically been entirely imported, and many system designs rely on existing films that withstand the high temperatures and rapid switching typical of data centers, fusion, and 800-volt EV architectures.
Components such as capacitor films need improvement. Companies are bringing new models to market, like Peak Nano, which has designed a film capacitor that heats hotter, cycles faster, and stores more energy in a smaller footprint, all while being designed and sourced domestically and through allied supply chains.
Capacitor films are just one of many components whose performance and availability will determine how quickly the U.S. economy can grow. Policy begins to favor modernized equipment and relocation. Over the next decade, we will likely see more incentives for allied manufacturing, targeted sourcing, and critical material designations for network-enhancing components.
Trend #5: AI is an integral part of demand and network fusion
AI will help twist much more efficiency off our grid. It can forecast demand, optimize dispatch, reduce losses and better manage distributed energy resources. But AI systems will consume almost all of the additional capacity created by these efficiency gains.
AI will also accelerate the fusion and other advanced energy technologies, from materials discovery to machine design and operation. This creates an interesting feedback loop. We need to power AI so that it can help modernize the grid and accelerate business fusion… which, in turn, will ultimately support AI’s own electricity consumption.
The implication? Funding and planning for AI energy needs is not optional. AI is essential to solving the consumption problem it amplifies.
Trend #6: the grid becomes a strategic weapon
In the new global energy system, the grid itself is an instrument of geopolitical power. Those who produce, transport and export robust, affordable electricity will set global energy standards, shape supply chains and gain strategic leverage.
The United States increasingly views nuclear, fusion, and grid-enhancement technologies as strategic assets, beginning to support them with sovereign wealth type structures. China has long followed a parallel path, pairing an aggressive domestic deployment with exporting the equipment and financing other countries need to build their power systems.
As this dynamic intensifies in the United States, support for large grid projects will be increasingly tied to domestic and allied supply, and export-oriented utilities will face pressure to adopt interoperable standards that align with these supply chains.
Looking Ahead: Anticipating the U.S. Government’s Next Steps
The next decade of energy leadership will not be defined by ambition or capital, but by solutions to difficult constraints. Where power can be located. What equipment can be built. And how aging infrastructure can be adapted to unprecedented load growth.
In this reality, power becomes a strategy. Access to reliable, scalable electricity will determine not only business outcomes, but also national economic growth, energy security, and AI leadership. To address this, we anticipate that U.S. policy will focus on a clear set of priorities:
- Encourage modernization: Prioritize dispatchable, flexible capacity and digital grid upgrades, not just added megawatts.
- Orchestrate the request: Treat data centers as network participants, rewarding load shifting and reduction when the system is stressed.
- Secure the supply chain: Critical ground and shore components like transformers, switchgear and power electronics.
- Plan for AI load growth: Make growing demand for AI and data centers a fundamental assumption of national energy planning.
- Make the license a competitive advantage: Accelerate transportation, manufacturing, and network upgrades related to reliability, AI, and critical infrastructure.
For Secretary Wright and the DOE, the equation governing AI is simple: no power, no progress. The electricity economy of 2026 to 2036 will not demand perfection, but it will reward preparation. Countries that act quickly and decisively in the face of this new reality will determine their competitiveness in AI and the future energy economy.
—Shaun Walsh is CMO for Pic Nano.
