Understandably, the public discussion about artificial intelligence has centered on the “Tech Bros”, the chips, the software, and the geopolitical issues. However, investors, regulators, and policymakers need to recognize that AI extends far beyond digital concerns. The growing excitement around AI’s current and future capabilities is overshadowing fundamental problems in power infrastructure that must be addressed immediately. Some policymakers are beginning to recognize this, but many outside the energy sector still believe the issue will resolve itself.
While the euphoria is certainly understandable, the United States must quickly set it aside and recenter on its foundational strategy. In 2023, U.S. data centers accounted for about 4.4% of total electricity consumption. By 2026, we have seen that figure climb as we hurtle toward a projected 325 to 580 TWh by 2028, accounting for up to 12% of U.S. electricity use.
This is roughly 3x the total electricity used by U.S. data centers just a few years ago. Globally, the International Energy Agency estimates consumption will nearly double to around 945 TWh by 2030, enough to power 100 million homes. This highlights the substantial physical energy demands of the hyperscaler economy, yet hardly anyone outside the oil and gas industry is discussing how to efficiently deliver power from point A to point B.
At this point, it is already well known that data centers require large amounts of electricity. What is less recognized is the variation in the qualities of electricity. Many people assume all electricity is the same, but this is a risky and incorrect misconception. Data centers require high reliability, precise voltage control, and the ability to scale to enormous loads rapidly. A home can tolerate a brief flicker; a data center cannot. So, what does that mean? Electricity is not just about availability; it’s about affordable consistency.
Not many people outside of the energy industry fully acknowledge the importance of the midstream space. This is a dangerous fault. To most Americans, “midstream” typically sounds overly technical. In reality, it encompasses the energy logistics system that collects gas from the field, processes it, compresses it, stores it, and transports it via pipelines to power plants, industrial sites, and local distribution networks. Whether we count pipeline operators or storage facilities, we are discussing a network of providers that will play a crucial role in the country's future economic security.
Simply put, the midstream industry serves as a foundational backbone that facilitates molecules to megawatts. It underpins the current U.S. data center economy. Natural gas already supplies more than 40% of the electricity physically consumed by data centers in the United States, and it remains the largest source of additional electricity through 2030.
Oil and gas detractors often cite the IEA analysis, which states that renewables remain the fastest-growing source globally. Despite its growth, renewables remain less than one-fifth of the United States' energy mix and cannot be relied upon during adverse weather events, when the grid is most vulnerable. An impactful plan survives the most vulnerable of times, not just when the sun is shining. For example, during Winter Storm Heather in January 2024, wind output in Texas fluctuated wildly; the system was under the greatest stress during morning peaks when wind was at its lowest. The practical lesson is not that renewables have no role, but that adverse weather continues to expose the need for firm, dispatchable power when reliability matters most.
Reliability should never be taken for granted. For a hyperscaler, a brief outage isn't just an inconvenience. It’s a financial disaster. Recently, reports estimate that outages can cost technology companies about $9,000 per minute.
If just 5% of the hyperscaler data centers in the U.S. experienced a one-minute power disruption daily, the cost would approach $200 million. By the end of the decade, as infrastructure expands, that aggregated cost could jump to $2 billion annually. This validates why it is egregious to assume we can rely on anything other than fossil fuels, particularly natural gas, for the foreseeable future. This is not a marginal load issue. It is a bulk-power-system planning issue that requires our immediate attention.
Reliability is not just about having enough gas in the ground. It is about having the infrastructure to deliver it. The country is counting on gas-fired reliability while still underbuilding the infrastructure needed to make it truly dependable. The consequences are clear: the U.S. might develop more gas-fired power plants than its pipeline infrastructure can reliably support during peak demand.
Speed in the midstream world is critical. A data center can be operational in two to three years, while building new transmission lines in advanced economies can take four to eight years. Unless energy and grid risks are addressed, around 20% of planned data center projects will face delays, potentially deferring over $500 billion in investment.
There is a concerning disconnect in understanding behind-the-meter power. Many believe co-located generation allows data centers to bypass midstream needs. In reality, these on-site power solutions rely on natural gas, which requires pipelines, gas processing, compression, and fuel-delivery infrastructure for reliable operation.
The market is already moving toward these solutions. Reuters reports that developers are planning about 56 gigawatts of on-site generation for data centers, representing roughly 30% of all planned data-center capacity. Midstream’s role does not vanish when a data center connects to the grid; it simply shifts and adapts. Midstream companies are evolving from transporters to integrated energy partners. Leading firms are pursuing a “one-stop-shop” approach, spending more than $5 billion on digital infrastructure power projects and behind-the-meter generation to provide the certainty that hyperscalers value above almost everything else.
First, the “generalists” must understand that supporting midstream is not the same as rejecting renewables or storage. A serious energy policy should be technology-neutral and oriented around delivery rather than ideology. If we want global AI leadership without sacrificing affordability, we cannot afford to handicap one of the main infrastructure systems that currently carries the load.
Second, policy should make it easier to permit and expand the infrastructure that improves fuel assurance. Implementing bureaucratic fluff in the permitting process is egregiously counterproductive. We need faster approval for pipeline expansions and storage, especially in regions where power demand is rising faster than transmission can be built.
Third, policy should protect the public from a false choice between digital growth and household affordability. Better management of data-center flexibility on constrained grids could avoid $40 billion to $150 billion of capital investment over the next decade. The public will be more likely to support data-center expansion if they know the beneficiaries of that growth are bearing their share of the costs.
The future of AI is often described as a contest over whether the United States can build enough real-world infrastructure to support a radically more power-intensive digital economy. Midstream moves the molecules that keep the lights on when intermittent resources are unavailable and demand surges unexpectedly. The public does not need to love every pipeline project to understand that their standard of living depends on its success. You cannot expand economic output without a healthy, flourishing midstream industry. It’s time we treated it as a priority rather than an afterthought to the AI boom.
When you choose Opportune, you gain access to seasoned professionals who not only listen to your needs, but who will work hand in hand with you to achieve established goals. With a sense of urgency and a can-do mindset, we focus on taking the steps necessary to create a higher impact and achieve maximum results for your organization.