Opinion: Keeping Up With Big Tech: What Natural Gas Providers Must Do To Stay Competitive

According to Goldman Sachs, AI is poised to drive a 165% increase in data center energy demand by 2030. Big Tech needs so much power to fuel their AI investments, in fact, they are moving closer and closer toward integrating their own power supplies into their operational models. For example, Meta will become Entergy’s single largest customer once plans to build 3 power plants dedicated exclusively to supplying energy for Meta’s AI data center are completed. Chevron and GE recently announced a partnership to build the first multi gigawatt-scale co-located power plant and data center. Investors are taking note and incorporating a new type of asset into their technology investment strategy – Blackstone just acquired a power plant in Virginia’s Data Center Alley.

Given their demand for large, reliable energy supply, it is not surprising Big Tech, or as the industry has redubbed them “Hyperscale tech giants,” are turning to colocation and off-grid strategies to secure power – and control cost. According to Power Engineering hooking up to the grid can entail a 5-7 year wait time – which is out of the question for tech behemoths spending billions of dollars to ramp up AI datacenters this year.  

The result? A new type of relationship between energy and technology companies in which energy becomes a competitive edge in Big Tech’s supply chain.

In this new world order, there will be winners. But what will dictate who gets the spoils (e.g. lucrative off-grid contracts to supply power to data centers)? Nimble, fast solutions to ramp up energy supply is certainly part of the answer – but it’s not the whole equation. In addition to speed, energy providers who can meet the dual challenges of efficiency and sustainability stand to gain in this new paradigm.  

To win data center contracts, it will be increasingly important for energy companies to drive – and show – measurable operational efficiency gains. Even setting aside for a moment the sensitive cost dynamics of the AI market, it’s safe to assume that the closer “supply” gets to integrating into Big Tech’s supply chain, the greater the pressure for operational efficiency will be.  

AI is exponentially more resource and capital intensive than previous technology paradigms – which means Big Tech companies need to find every efficiency angle possible. A recent illustration of the sensitive dynamics of the AI market: when China’s DeepSeek announced that its models allegedly use 10-40x less electricity than US-developed AI models, the stock market reacted with chip and energy stocks dipping. Microsoft and Google dipped too – why? Because less energy means fewer dollars spent on energy, giving DeepSeek a competitive advantage in the AI race. Energy costs account for 46% of total expenditures for enterprise datacenters, and 60% for service provider datacenters, according to a recent IDC report. The operational cost of data centers account for 15-20% of total AI investment by Microsoft, Meta, Google, and Amazon according to expenditure numbers reported in Quartz. Thus, reducing energy costs is a huge lever in the AI race – one technology companies will do everything they can to pull.  

Technology companies have already invested significantly in improving operational efficiency within datacenters. To reduce the resource-intensity of cooling processes, one source reports Microsoft is exploring everything from liquid cooling technology to an out-of-the-box concept of building data centers under water. Existing technology, magnetic bearing chillers, have already been proven to yield significant efficiency gains — reducing cooling costs by up to 40%, according to one study.

Now, Big Tech will look to drive favorable deals with energy providers – if they don’t outright buy power plants themselves. Thus far, Big Tech companies have primarily entered into power purchase agreements (PPAs) with energy providers. But, with Amazon’s acquisition of a nuclear-powered datacenter and associated energy infrastructure from Talen Energy last March and their direct investment in X-Energy to support the development of small nuclear reactors (SMRs), it’s easy to see how the mother of all vertical integrators could move into the energy business one day. In fact, Tesla is already in the energy generation business through their virtual power plant program, which aggregates and sells clean energy produced by at-home chargers back to the grid. All this suggests that IPPs need to act now to lock down contracts with AI data centers – or risk less favorable dynamics when Big Tech companies fully realize their appetite for DIYing power generation.  

Historically, nuclear power has been talked about as a big winner in the AI data center game. However, natural gas power plants can generally be built in half the time as nuclear plants – or less. Technology like SMRs may be promising in the future – but for now, SMRs are not widely available and scalable. The real advantage nuclear has over natural gas is that it is both clean and reliable energy. This matters since Big Tech companies have yet to completely turn-face on their lofty environmental pledges. Knowing this, traditional energy companies like Exxon and Chevron are building carbon capture into their courtship of data center contracts.  

However, carbon capture can significantly increase the operational costs of gas-powered plants. For natural gas providers to win long-term they will have to find solutions that prioritize operational efficiency while also supporting decarbonization efforts – if for no other reason than to hedge future shifts in political and regulatory winds. Luckily, existing technology can help natural gas companies check off both operational efficiency and sustainability boxes in their proposals to data center operators. For example, carbon capture is great – what if it was done at no additional cost and the captured CO2 was used to power turbines? It’s possible! Already have a power plant? What if infrastructure could be retrofit with turboexpander generators operating on magnetic bearing technology that could capture waste energy created in natural gas expansion processes and convert it to clean electricity that could be sold back into the grid or used to reduce operational costs (and emissions)? Also possible!  

Which is to say, natural gas is a natural fit for data centers due to its reliability, scalability, speed-to-market, and cost-efficiency. If natural gas companies can figure out the last piece of the puzzle – decarbonization – without compromising any of these core value drivers, they stand to win big in the race to power AI.