Winning the artificial intelligence (AI) race isn’t about compute capacity; it’s about who can get online first. For hyperscalers, colocation firms, and data center developers, speedy access to reliable energy has emerged as the defining competitive advantage in what is becoming an increasingly constrained power landscape. 

For example, a report by commercial real estate firm CBRE ranked adequate power availability as the top priority among data center developers. In a roundtable discussion held earlier this year, an executive from data center services provider JLL, Inc. stated the challenge clearly. “In 2025, site selection strategies are all about access to power [in] the shortest time frame.” 

Developers are increasingly facing uncomfortable questions that challenge their assumptions, surface vulnerabilities, and illuminate possible tradeoffs between speed, cost, sustainability, and long-term strategy.

Despite differences in scale and use case, data center developers have largely reached the same conclusions about viable (and not so viable) energy strategies. Here’s the consensus:

• The grid is not ready to handle data center power demands anytime soon. Just take a look at the map the National Renewable Energy Laboratory (NREL) developed showing that it takes between 3 and 7 years for large loads like data centers to begin receiving grid power. A survey of 149 senior data center employees conducted by Data Center Frontier found that 92% ranked grid constraints as the top obstacle slowing down data center projects.
• Natural gas-based power generation is the only viable near-term solution. We’ve all seen the announcements trumpeting the use of advanced modular nuclear reactors, hydrogen, and even fusion to power data centers. These may become meaningful power sources, but not in time to seize a competitive advantage in AI. Today, it’s all about natural gas. A story in the trade publication DatacenterDynamics bluntly summarized the prevailing belief. “Natural gas has emerged as the only viable and rapidly deployable power source capable of meeting the soaring energy demands of AI-driven data centers.”
• Onsite power is a must. There is an entirely rational response to the years-long wait for grid power facing data center projects today: build onsite power generation. Faced with long delays and unwelcome uncertainty, a recent survey of data center decision makers found that 27% expect to be fully powered by onsite generation by 2030, up from just 1% a year ago. 

There are real risks to accepting the consensus without rigorous self-assessment. The sprint to secure power as fast as possible can obscure problems that can surface before a data center ever goes online. Developers are increasingly facing uncomfortable questions that challenge their assumptions, surface vulnerabilities, and illuminate possible tradeoffs between speed, cost, sustainability, and long-term strategy.

Three uncomfortable questions to pressure-test your strategy

To scrutinize the consensus, here are three questions data centers should ask before locking in their energy strategy:

1. "What happens if your supply of firm gas isn’t so firm?" 

Natural gas powers multiple onsite generation technologies – from reciprocating engines and turbines to fuel cells and linear generators.

Reliably powering data centers with reciprocating engines, turbines, and fuel cells, however, depends on a non-interruptible, contracted supply of natural gas and the pipeline capacity to deliver it – which is known as firm gas. Gas supply interruptions can force data centers to use their backup diesel engines, negatively impacting local air quality and consuming their precious run-hour limits. 

While data centers require firm gas, the reality is that it is becoming increasingly difficult and expensive to secure. According to Bloomberg New Energy Finance (BNEF), data center natural gas demand is expected to increase by 470%, even as firm gas availability remains limited. This forces many data center developers to pursue more available and cheaper sites with non-firm gas. 

But there are risks to this approach, especially during peak demand or extreme weather events. For example, say a colocation provider runs a data center at a site with non-firm gas using 100 megawatts of reciprocating engines and has inked a service level agreement (SLA) with a 99.9% power availability guarantee – meaning that any interruption that lasts for even a short time can result in significant penalties. If a cold snap interrupts the supply of gas and the engines go offline, backup diesel generators will kick in. But diesel generators face strict air permit limitations on runtime, generate noise complaints, and produce visible smoke that triggers public opposition. 

True fuel flexibility requires the ability to seamlessly switch from one fuel to another. Would that flexibility make it worthwhile to revisit sites that had been dismissed for lack of firm gas?

A logical follow up question: What would it mean if a much cheaper data center location lacking firm gas could be powered by technology able to run on multiple readily available and economical fuels, including propane and hydrogen? Fuel flexibility offers firm power without the need for firm gas. True fuel flexibility requires the ability to seamlessly switch from one fuel to another. Would that flexibility make it worthwhile to revisit sites that had been dismissed for lack of firm gas? For developers willing to prioritize seamless fuel-flexible technologies, the answer could unlock significant cost advantages and improved energy resiliency.

2. "Is your speed-to-market strategy built on a permitting schedule outside of your control?"

Obtaining air permits for onsite generation is complex, with requirements varying by location and technology.

Dissecting all the permitting obligations associated with onsite generation takes time and expertise. But there are some necessary general steps to follow and realities to recognize. Keep in mind that the difficulty obtaining air permits will largely be dictated by where the data center is located, the type of generation, and how frequently it will run. Location matters because permitting in nonattainment areas – which are regions that fail to meet one of the U.S. Environmental Protection Agency’s (EPA) National Ambient Air Quality Standards – is more challenging. 

Onsite generation designed to run frequently is harder to permit than systems meant to operate only when circumstances demand more energy. With the caveat that location and operational details matter a lot, there is a clear hierarchy among onsite technologies – including those fueled by natural gas – when it comes to the speed and difficulty involved with air permitting. Local communities where data centers have been proposed have also been pushing for more transparency around necessary air permits, driven largely by worries about diesel generators. 

Here’s why that matters: Traditional natural gas turbines and engines designed for daily use can take over a year to permit – requiring public notices, hearings, emissions controls, and sometimes pollution offsets. Smaller turbines and engines move faster. Fuel cells and linear generators faster still. The clear lesson: If speed to market is your priority, permitting timelines should be a primary factor in technology selection – not an afterthought. Solutions that emit near-zero NOx without additional aftertreatment, for instance, can fast-track approvals even in challenging nonattainment zones, shaving months off project timelines. 

Technologies with natively low emissions also avoid the costs and compliance risks associated with aftertreatment that can malfunction and violate air quality permits. For example, Selective Catalytic Reduction (SCR) systems reduce NOx emissions from gas-powered engines and turbines but are susceptible to malfunctions that result in air quality noncompliance.   

3. "Will your energy strategy still make sense when the grid finally catches up?"

The pivot to onsite generation is meant to solve an immediate problem. Data centers need power fast, and onsite solutions tangibly address the speed to market dilemma. But is it wise to design your entire energy strategy to tackle a single, short-term problem?

The answer may very well be yes. The competitive necessity to access power quickly may overwhelm any consideration about powering data centers for decades of operation. Nevertheless, even if short-term needs prevail, it is worthwhile to view energy strategy as a long-term proposition. Central to those deliberations is what your response will be in three, four, five, or more years when grid power is available for your data center. In many cases, that grid power will be significantly cheaper than onsite power, which means that a long-term strategy for onsite assets is critical. 

Onsite power solutions are big investments. Some can evolve to complement grid power by providing backup power, peak shaving, and demand response.  If the generator can operate on zero-carbon fuels like hydrogen, it can transition into a hydrogen generator without new capital investment. Clearly understanding the operational characteristics, including limitations, of individual technologies will help clarify whether onsite power solutions can have a productive second life once grid power arrives.   

Balancing short-term needs with a long-term strategy 

None of this is to say that the availability of grid power is unwelcome. Rather, it’s to urge data centers to think holistically about their energy strategy to ensure a smart balance between short-term needs and long-term success.

Data center energy strategy shouldn't be dictated solely by today's grid constraints. The smartest approaches balance immediate power needs with long-term operational flexibility and overall cost effectiveness. That means asking hard questions now: Can you guarantee fuel supply? Will your technology clear permitting hurdles quickly? And what happens when the grid catches up? Speed matters – but so does building an energy strategy that works for decades, not just deadlines. 

Successful data center developers will be those willing to face the uncomfortable questions others avoid. These questions – and the issues beneath them – must eventually be answered. Tackling them now gives you control you won’t have if circumstances force the answers later.