AI-Native Enterprises Rethink Long-Term PPAs As Power Becomes A Strategic Cost

With data center demand rising and power usage under growing scrutiny, securing energy has become a central concern for companies in the AI market. To lock in supply, many are signing 20- to 30-year PPAs. But those agreements are still tied to a legacy grid, where the physical limits of transmitting electricity over long distances mean roughly 5% of power is lost as heat along the way. Across the lifespan of a multi-decade contract, that loss becomes a significant and largely invisible cost. As the industry tests new approaches to AI infrastructure and power delivery, locking into an inflexible agreement today could leave companies unable to capture the savings of next-generation energy systems tomorrow.

Allan Duarte has seen how big capital moves. Now a GenAI Implementation and Adoption Lead at a global management consulting firm and a Stanford GSB Sloan Fellow, he previously spent years advising the President of Brazil's Federal Senate on more than $200 billion in economic and technology reforms. In a recent newsletter post, he outlined the financial drag of this transmission loss. Speaking strictly in his personal capacity as an investor and strategist, Duarte notes that many procurement teams need to rethink how they structure energy deals to preserve flexibility.

"I think the idea is 'optionality.' The major players in AI are finding ways to build that into their contracts. They secure current energy, and those contracts will also be bulletproof for new and emerging sources of sustainable power," Duarte says.

The physics tax hiding in every PPA and AI timeline compression

New advances in renewables, superconducting materials, and next-gen nuclear power generation are reshaping how power is produced and delivered to customers, and how power-intensive enterprises factor consumption into their budgets. It's still the case, however, that many of these organizations are signing multi-decade deals without adjustment mechanisms, which inadvertently amount to a bet against these future grid improvements.

The pressure is already showing up across AI infrastructure planning. Everpure and MIT Technology Review's recent energy intelligence report shows 51% of leaders ranking rising energy costs as the greatest energy-related threat to their digital and AI initiatives, while 68% report energy cost increases of 10% or more from AI and data workloads. That makes energy visibility more than an efficiency exercise. It becomes part of how companies decide whether today’s power contracts can adapt to tomorrow’s infrastructure economics.

"Optionality" is a different approach to PPAs, in which enterprises can get the energy they need now using today's standards while building in protocols to move to more cost-effective technologies and pricing tiers as they emerge. Duarte suggests that forward-thinking leaders should treat optionality not just as a legal detail, but as the core strategy for how they buy power. "When they are building contracts right now, I think that they can lock in the energy that they need, but also they can have the option to make a transition," he says. "Right now, and for the next ten years, they will still have to deal with this 5% loss. But if they have specific clauses and sections in their contracts that foresee this possible change, it can save you much more in the long term."

Part of the challenge lies in gauging the timeframes within which these innovations could occur. What might take several decades of research and testing can shrink to mere years with the fast pace of AI research in materials science. When accelerating scientific progress meets the capital pouring into infrastructure, multi-decade PPAs suddenly overlap with what could be a much shorter window of technological change. "Superconductor materials will likely be improved and developed because of new AI and research. This is compounding," Duarte says. "We can think right now that it's going to take twenty years, but maybe it's going to take ten or less because one breakthrough builds upon another. I don't think this will be different because there is so much money in play here; everyone is looking for new solutions."

Stress-testing the contracts you already have

Standard, "set-it-and-forget-it" procurement habits simply weren't built for the current speed of AI. Duarte points out that companies blindly auto-renewing their power deals are missing a crucial window to update their terms. He advises organizations to take a closer look at their long-term energy stack to understand exactly what they are committed to. "The main stress test you can do is look into your current PPAs and see how long your lock-in is. If you are locked in for twenty or thirty years, you definitely need to start building some kind of flexibility right away instead of just rolling over using the same terms," he says.

For contracts still being negotiated, the window to act is now. "If you are getting new PPAs and closing new contracts, that's the time to insert these kinds of options. Make sure the contract considers these new possibilities and doesn't lock you in for thirty years," Duarte says. "By doing this, you secure your current energy, but your contracts give you flexibility and resilience as new energy scenarios become reality."

A portfolio approach to emerging energy

PPA negotiation isn't typically an issue for enterprises with large energy needs. For mid-sized enterprises that lack the leverage of hyperscalers, however, securing favorable terms isn't always easy. That's why Duarte sees a powerful path forward for both types of organizations in treating emerging energy technologies as an evolving portfolio. That includes taking early positions in startups trying to move power more efficiently over distance. We can already see that playbook in action among major AI players, backed by massive infrastructure commitments and government contracts for AI-assisted discovery of new materials.

"If you look at the big players like OpenAI, they have a lot of investments in small startups that build energy solutions. They are investors in modular nuclear companies, ensuring they can access this kind of energy when it becomes available. There is Oklo, a developer of cleaner fission power plants, and Exowatt, which is a small modular solar company that Sam Altman has invested in," Duarte says.

The point isn't to bet on technology, but to help support it and get a better understanding of the innovations coming down the pipeline. "I would try to figure out all these different possibilities, whether in terms of superconductors, solar, or new ways to harness energy," he says. "If even one of them really works out well, that represents millions or even billions in savings. Depending on the company, they can secure cheaper, better energy."

Buying power like a venture investor

For AI companies signing energy deals that will outlive the current generation of hardware, the calculus is shifting. The companies that win the next decade of compute won't just be the ones with the most power. They will be the ones whose contracts let them swap into the cheapest, cleanest electrons the moment those become available.  In Duarte's view, companies that treat energy as a flexible asset are often better positioned to capitalize on the next wave of infrastructure. It is a perspective that requires leaders to think more like early-stage investors than passive buyers of power.

"We are definitely in a new era of tech development. It is nothing like any other cycle before. This cycle will move much faster and be much more impactful," Duarte says. "I would definitely try to make sure I'm involved in all of these new possibilities before they become mainstream. You can do that by seeking equity or securing contracts in advance. It's a small bet you can make right now to ensure you are positioned."

The views and opinions expressed are those of Allan Duarte and do not represent the official policy or position of any organization.