Speed-to-Power: An Energy Policy Agenda for a Thriving AI Market

What energy policy agenda will enable startups to compete in AI?

That question is becoming more urgent as AI’s exponential growth drives an unprecedented demand for electricity. The Department of Energy forecasts that data center electricity use will nearly triple to 12 percent of US supply by 2028, up from 4.4 percent in 2023. Yet this surge is colliding with an analog-era grid: as of late 2023, an astounding 2,600 gigawatts (GW) of new energy projects—over 95 percent of which are zero-carbon—were waiting in grid interconnection queues, a backlog more than double the entire installed capacity of the US power plant fleet.

This multi-year logjam creates more than just scarcity; it creates a two-tiered market that structurally disadvantages startups. Large, established hyperscalers can lock in long-term Power Purchase Agreements (PPAs), securing billions of dollars in predictable energy and effectively buying out entire regions’ future power capacity. Startups don’t have this luxury. Lacking the scale and balance sheets to make such long-term commitments, startups cannot negotiate these priority positions and are left to face volatile spot markets and higher rates. Moreover, startups don’t have the legal teams or lobbying muscle to develop a bespoke AI energy policy agenda or to advance that agenda on Capitol Hill. Because of their comparative disadvantages in business and policy negotiations, the next generation of AI innovators may be denied access to the energy they need to compete with larger platforms; energy scarcity could become a durable competitive moat.

In AI, energy policy is not a peripheral concern; the challenge of “speed-to-power”—how fast a new data center can access energy—is key to competing in the AI industry. To ensure a level playing field, we need an AI energy policy agenda that works for startups.

This agenda should focus on three core concepts:

1. Generating energy abundance
2. Ensuring fair energy access
3. Innovating in energy supply and demand

The Abundance Agenda: More Supply, Lower Costs

AI’s energy demands require more supply. Without enough energy, costs increase and both large and small developers will face obstacles to building new AI tools. As the administration’s AI Action Plan recognizes, the foundational pillar of a forward-thinking energy policy must therefore focus on dramatically increasing the total supply of energy and data center capacity to lower costs for everyone.

Momentum for a pro-abundance energy agenda is increasingly bipartisan, with leaders from the Trump administration to the Democratic mayor of Silicon Valley’s largest city, championing robust reforms to expand energy supply and accelerate infrastructure buildout. Journalists Derek Thompson and Ezra Klein have provided an intellectual framework for abundance policy, articulating how supply-side solutions can underpin America’s competitiveness and innovation. Organizations such as the Abundance Institute and the Niskanen Center are researching and advocating for policy to realize a future of plentiful, affordable energy, bridging ideological divides and shaping a common path forward. This emerging consensus reflects a rare moment in contemporary policy where growth, innovation, and economic opportunity are rallying points across parties and institutions.

Policymakers are starting to realize the importance of expanding energy access. For example, the Federal Energy Regulatory Commission issued Order No. 2023, which aims to reform the grid interconnection process and clear backlogs by shifting to a “first-ready, first-served” cluster study model. More recently, the Trump administration’s AI Action Plan calls for fast-tracking energy infrastructure construction, and Congress has held hearings on the energy needs of the AI sector. States are also heavily engaged in energy policy—some, fearing strain on the energy grid, are reconsidering existing data-center incentives or considering proposals that would require more transparency about data center energy usage, while others are doubling down on encouraging more AI activity within their borders.

But policymakers should take additional steps to reduce burdens to energy production and access:

• Accelerate energy permitting. Speed up federal and state permitting for new energy generation and transmission infrastructure to allow supply to more rapidly meet demand from the AI sector. This should be coupled with a concerted effort to streamline and standardize the state and local permitting processes for data center siting.
• Design smart transparency rules. Step cautiously in enacting data center transparency mandates—proposals that would require operators to report on metrics like energy consumption and water usage. There were more than 20 such bills introduced in 2025 alone. Any new transparency rules should be carefully designed to avoid slowing the construction of new capacity or imposing burdensome compliance costs that will disproportionately affect startups, given their small teams and limited budgets.
• Unlock existing grid capacity. Reform outdated utility regulations that discourage the use of innovative grid-enhancing technologies. Incentivizing the deployment of these modern optimization tools could unlock as much as 100 GW of capacity on existing power lines.
• Invest in the workforce. We must invest in the skilled workforce needed to build this new infrastructure. Addressing critical labor shortages, particularly for electricians, is not only essential to removing a key bottleneck for deployment but also represents a major opportunity to create a new generation of high-paying jobs for skilled trade workers across the country.

While boosting supply is paramount, we must also continue to innovate in efficiency. It’s crucial to increase public R&D for more energy-efficient AI, spanning both hardware and algorithms. However, we must also be clear-eyed about the limits of this approach. Efficiency gains alone are insufficient to address the demand problem, as improvements in efficiency frequently trigger increased consumption, a phenomenon known as the Jevons Paradox.

The Pro-Competitive Agenda: Ensuring Fair Access

While an abundance of energy is critical for both startups and large hyperscalers, it’s insufficient on its own to foster a competitive AI ecosystem. Policies that create more supply must be accompanied by policies that distribute that supply fairly. Preferential energy access would create new barriers to AI development, favoring large players over startups. To avoid this outcome, energy policy must address the structural disadvantages startups face in the energy and compute markets.

The cornerstone of fair, balanced access is ensuring that startups do not pay for the energy needs of hyperscalers. Large-scale AI deployment could increase energy costs for everyone else: a single hyperscale data center can require grid upgrades costing billions of dollars. If policy permits these costs to be “socialized” across all ratepayers rather than being borne directly by the entity causing them, startups will be disadvantaged.

For instance, a large hyperscaler could negotiate a custom contract with a state utility to power a new data center, a project requiring costly, large-scale upgrades to transmission lines and substations. If state regulators, eager to attract the investment, allow the utility to recover those upgrade costs across its entire customer base—overall electricity rates for all other customers will rise. Simultaneously, this type of project could consume all planned grid capacity in the region for years to come, leaving startups unable to secure power for their own smaller facilities. If costs rise due to hyperscaler demand, but hyperscalers don’t pay for this increased price, then residential customers and smaller businesses—including startups—would effectively subsidize the energy infrastructure of their largest competitors.

To prevent this outcome, policymakers could implement a “causer pays” principle that requires large-load users to directly fund the grid upgrades they require, rather than socializing these costs. This ensures that the entities driving increased demand are the ones who bear the financial responsibility for the necessary infrastructure improvements.

Fairness should also extend to energy procurement. To correct the imbalance caused by large companies securing favorable long-term PPA terms, policymakers and energy regulators should establish aggregated purchasing programs that allow startups to pool their demand, giving them the collective bargaining power needed to secure the same competitive energy contracts.

Energy policy should also reduce barriers to entry in AI by ensuring startups have access to the foundational resources needed for innovation. The immense cost of high-performance computing can prevent promising ideas from ever being developed. As we have written previously, the federal government can address this challenge by establishing a National AI Competitiveness Institute (NAICI), which would provide startups, researchers, and nonprofits with access to the infrastructure they need, including compute power, data, and model evaluation resources. By providing direct access to these energy-intensive computing resources, the NAICI would help startups manage the high infrastructure costs that can cut off their ability to innovate, making it easier for Little Tech to compete with larger platforms in building and testing new AI models.

The Innovation Agenda: Unleashing New Approaches to Energy Supply and Demand

Startups cannot compete with hyperscalers on sheer capital expenditure; their strength lies in their innovation and agility. Policy must create opportunities for them to leverage these unique advantages. A forward-looking strategy should focus on enabling startups to capitalize on their operational flexibility and participate in the development of next-generation energy technologies.

A key area for innovation is demand-side flexibility: the ability of energy consumers to actively manage when and how much electricity they use in response to the needs of the grid. State utility commissions, for instance, could direct utilities in their jurisdictions to offer innovative tariffs, like real-time pricing, which would allow startups to significantly reduce costs by shifting their computational workloads to off-peak hours when power is cheaper.

Furthermore, state and federal policymakers can increase startup access to demand response programs, which pay companies to temporarily reduce their energy use during periods of high grid stress. Lowering participation barriers for these programs would enable agile startups to generate revenue simply by pausing non-critical workloads, turning their operational agility into a direct economic advantage.

At the same time, policy should promote the development of next-generation power sources that can provide the 24/7 clean energy that AI data centers require, such as Small Modular Reactors (SMRs). To accelerate SMR deployment, policymakers must build on the important progress of the bipartisan 2024 ADVANCE Act, which began to address the high cost of licensing through reducing certain fees and authorizing prize competitions to cover regulatory costs for innovators. Policymakers should ensure this new framework succeeds by conducting oversight to confirm the law’s regulatory reforms are being implemented effectively.

Likewise, advanced geothermal energy is a promising source of the 24/7 clean power that AI data centers require, but its deployment is severely hampered by complex permitting processes that can take seven to ten years on federal land. To accelerate this technology, policy should focus on speeding up federal leasing and drilling permits to reduce these costly development delays.

The Department of Energy’s recent funding for emerging technologies like AI and nuclear energy is a welcome step. Policymakers can build on this initiative by supporting the commercialization of technologies that could help address AI startups’ energy challenges.

For any new technology that expands energy supply or mitigates energy demand, startups must have equal access to the opportunity. To prevent large platforms from securing exclusive access to these new energy resources, legislators and regulators could mandate that utility development partnerships include opportunities for startup participation, either as energy off-takers, who purchase shares of the generated power, or as co-locators, who physically site their facilities at the new power source.

Securing America’s Energy Advantage in AI

The race for AI leadership could be won or lost on the electrical grid. The policy landscape, designed for a pre-AI era and predisposed to favor large providers over startups, now risks creating insurmountable barriers for the next generation of innovators. The path forward is not to simply power the nascent AI sector, but to ensure that energy policy does not leave small AI developers behind.

By embracing a comprehensive strategy built on abundance, pro-competitive market access, and a commitment to innovation, we can move beyond the zero-sum mindset of scarcity and toward a future where startups have affordable access to the energy they need. Let’s ensure the best ideas—not just the biggest balance sheets—have the power they need to win.