Powering Artificial Intelligence: Data Centers, Finding the Power, and De-Risking

As artificial intelligence surges ahead, it’s bringing an unprecedented appetite for electricity. Massive AI models and machine learning workloads are driving a dramatic rise in data center development and related electricity and energy demands. This growth is forcing a rethink of how we power the digital economy. Traditional energy strategies are straining at the seams, and attention is turning to innovative solutions, including a resurgence of nuclear energy.

In this MSLLC ThinkTank post, we explore a few concepts related to meeting the rapidly growing electricity demand from Data Centers:

• Project Investment Risk: Key technology and regulatory risks and how investors, lenders, and developers can mitigate these project risks.

• Nuclear Energy’s Future Role: How nuclear power is a key player in meeting data centers energy needs and what challenges lay ahead.

• MSLLC’s Differentiators: How MSLLC brings deep experience to advise investors, lenders, and developers with project risk and nuclear power generation.

  
  

The AI Data Center Power Surge

The scale of power needed for cutting-edge AI is hard to overstate. Data centers running AI training clusters and large-scale models consume vast amounts of electricity, far more than conventional IT facilities. Industry analyses forecast that power demand from AI-driven data centers could grow by an order of magnitude in the next decade. In the U.S. alone, some estimates predict that AI data centers will require over 100 gigawatts of power by the early 2030s, up from only a few gigawatts today. To put that in perspective, a single sprawling AI campus with thousands of high-performance processors might draw hundreds of megawatts, on par with the output of a utility-scale power plant. Globally, data center energy usage is on track to double over the next several years, potentially accounting for a significant percentage of total electricity consumption.

Training and running AI models, especially the latest generation of large language models and generative AI, is extremely energy intensive. These systems rely on racks of specialized hardware (GPUs, TPUs, and other accelerators) running continuously. For example, processing one AI-driven query or task uses a lot more energy than a standard web search. Multiply that by billions of queries and continuous training runs and the energy profile skyrockets.

The result: today’s hyperscale data centers (and the emerging AI supercomputing hubs) are rapidly becoming one of the biggest energy users in the economy.

This AI power surge is already placing enormous stress on power infrastructure and sustainability goals. For data center operators and planners, the question is increasingly urgent: Where will all this electricity come from? Tech giants are responding with record investment in a rush to build capacity, yet this rapid growth comes with equally massive technical and financial challenges.

Explosive Growth by the Numbers

AI workloads are obliterating traditional data center limits. Top-tier AI chips now draw more than 700 watts each, often clustered four or eight per server. Thus, a single server can draw 5‑10 kilowatts, meaning a typical full rack of AI gear can exceed 50 kilowatts of load; next-gen AI systems are pushing rack power to 100-120 kilowatts and are expected to go even higher. Cooling of this equipment adds an even greater energy demand, which has also become a make-or-break challenge.

Goldman Sachs projects AI could trigger up to a 160% spike in data center power demand. The International Energy Agency reports data centers globally consumed 460 terawatt-hours in 2022 (approximately 2% of all electricity usage), a figure expected to double by 2026. In the U.S. specifically, data centers drew approximately 4.4% of the nation’s electricity in 2023 and the U.S. Department of Energy (DOE) predicts this share could reach 6.7%-12% by 2028.

AI’s impact on infrastructure has only begun. Over the next 12-18 months, we expect data center investment to accelerate sharply, but not without growing pains. Supply chain delays, permitting issues, workforce capacity, and environmental constraints are already testing delivery teams.

Grid Limitations Spark a Search for Alternative Approaches

Traditionally, data centers have drawn power from the electric grid, often supplemented by backup generators and renewable energy purchases; however, this status quo power strategy is being stretched thin. Many regions simply can’t deliver enough power fast enough to meet the sudden jump in demand from AI-centric facilities. In several tech hubs, clusters of new data centers are bumping up against grid capacity limits. Utilities are scrambling to plan new generation and transmission, yet these upgrades can take years. Meanwhile, data center projects face delays or curbs because there isn’t reliable power available at the capacity and reliability required.

Even when power is available, the around-the-clock load of AI data centers presents challenges. Unlike many industries and commercial loads that have variable electricity usage, AI data centers typically draw a flat, continuous baseload of tens or hundreds of megawatts. This unrelenting demand can stress the local grid while requiring continuous power generation sources. With the added dynamics resulting from increased intermittent sources of power, such as solar and wind, the challenges of finding baseload power to support these data centers has become complex in some regions.

Large data centers must navigate local zoning, environmental reviews, noise ordinances, and even public opposition. Land use is a real concern as zoning regulations could stall data center industry expansion if not addressed properly. Projects often need robust environmental and resiliency plans to get approved, such as demonstrating zero-water cooling designs or agreeing to noise mitigation and traffic management during construction. All of these factors can affect project timelines and costs significantly.

From the perspective of investors and lenders, all of the above raises the bar for due diligence. Whether it’s a private equity firm financing a new colocation facility or a consortium investing in a hyperscaler’s expansion, there is a heightened focus on risk mitigation and execution certainty. Capital will flow to those who can demonstrate:

• Robust power strategies

• Permitting and site viability

• Cost and schedule control

• Resilience and ESG compliance

  
  

These pressures are prompting data center developers and power generation providers to think outside the box. One emerging approach is to develop dedicated data center power generation, either inside-the-fence or within close proximity. In essence, data center developers are exploring providing their own power generation and other energy supply. Onsite natural gas turbines, internal combustion engines, and fuel cells are all in play at various data centers; yet each has limitations, particularly with air emissions or fuel logistics. Similarly, some data centers have also been relying on nearby solar photovoltaics (PV) and wind battery energy storage systems (BESS). Regardless of the power generation method, most data centers still depend on grid backup. That growing urgency is driving interest toward a more ambitious solution—dedicated nuclear energy for data centers.

Partnering with experienced technical advisors becomes crucial at this stage. As we’ll discuss, having the right expertise in your corner can make the difference between an AI data center project that is high-density and high-return versus one that is high-risk and headache-prone.

Clean, Scalable, Reliable – Nuclear Energy’s Potential New Role

After a decades-long lull, nuclear power is re-entering the conversation as a compelling option to power the AI era. This nuclear renaissance is being driven by the need for clean, scalable, and reliable electricity. Advanced nuclear technologies, especially small modular reactors (SMRs) and other next-generation designs, are gaining traction as potential game-changers for powering data infrastructure. A single SMR can supply tens or hundreds of megawatts continuously. This high energy density means a reactor installation could directly feed a large data center campus without sprawling fuel depots or daily fuel deliveries—a significant advantage over engine generation.

The data center industry is taking notice. In the past year, major cloud and AI companies have announced partnerships to secure nuclear power for their future data centers. In the U.S., several high-profile deals signal that nuclear is moving from idea to reality in this sector. For example, Google has signed an agreement with Kairos Power to develop multiple small modular reactors—targeting about 500 megawatts of capacity dedicated to its data centers by 2035. Another major hyperscaler, Microsoft, is reviving a long-shuttered reactor; under a 20‑year power purchase agreement with Constellation Energy, it will reopen Three Mile Island Unit 1 solely to power its expanding data center footprint in the Mid‑Atlantic. At the same time, high-profile e‑commerce and cloud provider Amazon Web Services (AWS) has secured up to 960 megawatts of nuclear power via a Power Purchase Agreement (PPA) with Talen Energy’s Susquehanna nuclear plant and is investing in X‑energy to help build several gigawatts of SMR capacity into the late 2030s. These moves underscore a broader trend: nuclear energy (both conventional and advanced modular) is being actively explored as a strategic power source for AI infrastructure.

Public policy is reinforcing this momentum. Governments see an opportunity to meet climate targets and economic goals by marrying nuclear energy with the digital boom. In the U.S., federal incentives and loan programs are helping to finance the next wave of reactors, and several states have lifted moratoriums and introduced bills to promote SMR deployment, often with a nod to attracting data center investments. Globally, we’re seeing similar interest. Nations such as the UK and Japan are evaluating dedicated nuclear solutions for high-tech industrial parks and data center hubs, recognizing that a reliable energy backbone will be crucial for AI competitiveness. New reactor designs are reportedly being tailored with data centers in mind, emphasizing modularity and enhanced safety to integrate more easily into an industrial campus environment.

In short, the resurgence of nuclear power comes at a pivotal time. It offers a path to scale up nuclear energy in step with AI’s growth and to do so with the reliability that critical digital systems demand. However, this approach to nuclear power generation for data centers brings its own set of challenges for stakeholders to carefully manage, including a complex landscape of regulations, engineering requirements, and public concerns.

• Regulatory and permitting hurdles

• Siting and community acceptance

• Safety and public perception

• Integration and infrastructure

• Development lead times

• Capital costs

• Timeline uncertainty

It’s clear that while the marriage of nuclear energy and AI data centers holds great promise, it must be pursued with careful planning, risk mitigation, and expert guidance at every step.

Navigating Your Strategy with the Right Expertise

For developers and investors evaluating their roles in AI data centers and supporting electricity and energy supply options, having experienced guidance is invaluable. This is where project technical advisory and independent engineering come into play.

A knowledgeable technical advisor provides an objective, in-depth evaluation of proposed energy solutions, from technical feasibility to economic viability to risk factors and risk mitigation, before key investments are made. In the rapidly evolving intersection of AI and energy, MSLLC offers exactly this kind of support to help clients evaluate and de-risk their energy strategies and decisions.

At MSLLC, we bring decades of experience in mission-critical infrastructure projects, including extensive work in the energy sector. Our team has provided technical oversight and advisory services for complex power generation initiatives (from renewables to advanced nuclear demonstrations) and large-scale data facilities alike. Our background gives us a 360° perspective on both sides of the equation, the power and energy engineering and data center operational needs. We leverage our extensive expertise and deep bench of SMEs to guide our clients in making informed decisions that balance innovation and growth with risk.

What sets us apart?

• End-to-End Technical Advisory

• Lifecycle Cost and Constructability Expertise

• Power and Energy Systems Know-How

• Broad Technology Perspective

  
  

Ultimately, MSLLC’s greatest differentiator is our track record of helping clients minimize risk in support of project finance and investment. We don’t just check boxes; we actively seek out potential failure points and identify solutions. Our team has supported mission-critical projects for the U.S. DOE, Department of Defense, and Fortune 500 clients, where failure was not an option. We bring that same diligence to commercial data center initiatives. In an industry where time is money and technical missteps can cost millions or derail funding, having MSLLC as your independent technical advisor is like an insurance policy—one that pays dividends by enabling smarter decisions, safer investments, and smoother project delivery.

Our work with DOE’s Office of Clean Energy Demonstrations (OCED), New York State Energy Research and Development Authority (NYSERDA), and major utilities shows that large-scale infrastructure and digital systems require more than good design; they demand systems thinking, collaborative planning, and credible, finance-ready reviews.

We’ve reviewed billions in infrastructure project value across sectors, including mission-critical facilities for the DOE and U.S. Army Corps of Engineers. We’ve also advised on verification systems for renewable energy infrastructure and designed data environments for energy market reporting. This breadth is what makes us effective in the data center space.

How can MSLLC add value to your data center project? It begins with rigorous independent engineering reviews. We dive into proposed projects and project portfolios, from review of proposed designs and FEED studies to PPAs to commissioning and startup reports, verifying that projects under consideration or development will meet expected performance and compliance. Our technical advisory support covers everything from energy and electricity supply to resiliency to integration with a data center’s power and cooling systems. By identifying potential technical pitfalls early, we help clients mitigate risk.

In addition, our project lifecycle advisory approach ensures that we’re looking at the full span of the project, from concept and permitting through construction, operation, and even decommissioning. We advise on permitting strategy, helping clients map out the regulatory pathway to prepare necessary documentation and studies to satisfy regulatory agencies and government authorities. MSLLC supports stakeholder engagement at every level. Introducing novel approaches to energy and electricity sources requires clear communication of risks and benefits.

Summary

In short, MSLLC is a trusted partner for organizations developing and investing in cutting-edge data center and energy supply solutions.

We are independent, meaning our priority is protecting our client’s interests and ensuring project viability.

We are comprehensive, bringing technical, regulatory, and financial insights under one roof.

We are experienced, having guided numerous energy infrastructure projects from planning to operation.

We can support you when the stakes are high.

As the AI revolution drives up demand for reliable, clean power, the opportunities for innovative solutions are immense. Yet, realizing these opportunities requires more than technology; it requires strategy, diligence, and expertise.

If your organization is developing or investing in the growth of AI data centers and supporting resilient power and energy needs, now is the time to evaluate all options. Engaging with experts early can help chart the right path. MSLLC can help you assess, plan, and de-risk your development and investment strategy, ensuring your decisions in supporting the AI infrastructure market are not only effective but also future-proof and responsible.

Together, let’s ensure that the rise of AI is matched by a strategy built for the long haul.