Trump’s crusade against Big Tech’s energy spending highlights a problem with no easy solutions
Donald Trump’s renewed broadsides against Big Tech have expanded beyond speech and antitrust to the escalating electricity appetite of the digital economy—especially AI data centers. The critique resonates: power demand that was flat for a decade is rising again, driven by cloud computing, AI training and inference, and, in some regions, crypto mining. Utilities are revising forecasts upward, local communities are balking at land and water use, and climate goals are colliding with the speed at which new load is arriving.
But while it is politically simple to lambaste Silicon Valley’s power bills, the underlying challenge defies quick fixes. The United States wants more computing and more reliable, affordable, cleaner power—all at once. The tools to square that circle are real but slow, fragmented, and full of trade‑offs.
What’s actually happening to demand
– Data centers are a fast‑growing slice of load. Global electricity use by data centers, AI, and crypto could exceed 1,000 TWh by the mid‑2020s, roughly the consumption of a large industrialized country. In the U.S., grid operators and the Energy Information Administration expect data centers’ share of demand to push into the mid‑single digits within a few years; in some hotspots—Northern Virginia, parts of Texas and Ohio—new computing load dominates utilities’ growth forecasts.
– AI changes the shape of demand. Training giant models arrives in bursts that can be scheduled; inference is persistent and scales with usage, turning computing into a round‑the‑clock load. The industry is also shifting to higher‑power chips and liquid cooling, increasing site‑level density even as efficiency per computation improves.
– The system is already strained. Interconnection queues are backlogged, existing transmission is congested, permitting for new lines is slow, and resource adequacy is tight in several regions. Communities are increasingly wary of large data campuses that occupy prime land, add heavy distribution and transmission costs, and consume water.
Trump’s instincts and the policy fault lines
Trump’s energy posture is clear: favor fossil fuels, ease permitting for pipelines and gas plants, pare back environmental rules, and question “green” subsidies. Applied to the data‑center problem, that translates into a straightforward if controversial path: meet new load quickly with more gas‑fired generation, postpone or loosen climate constraints, and allow markets to sort the rest out.
That approach would likely accelerate deployment where siting is least contentious and interconnection fastest. It also dulls the edge of a genuine reliability risk. But it collides with several realities:
– Corporate climate commitments are now hard constraints. The largest cloud providers have pledged 24/7 carbon‑free energy by 2030 and net‑zero supply chains. Their investors, customers, and employees expect progress. If federal policy pulls back on clean‑energy incentives, their cost of meeting those goals rises—but the goals remain.
– Gas is fast, but not frictionless. Local air permits face lawsuits, many regions are short on pipeline capacity, and peaking units without fuel assurance are increasingly scrutinized after winter outages. New gas can improve reliability, but it risks locking in emissions and stranded assets as states and companies advance decarbonization.
– The federal government cannot wish away grid physics. Transmission planning, interconnection reform, and distribution upgrades take time. Even with aggressive streamlining, multistate lines and substations move on multi‑year timelines.
On the other side, a purely “build only renewables and batteries” answer also runs into hard limits. Wind and solar are now the cheapest new megawatt‑hours in many places, but delivering firm, around‑the‑clock power at the speed and scale AI wants requires complementary solutions—storage, demand flexibility, clean firm generation—and, crucially, transmission to move energy from where it’s abundant to where data centers cluster. Those ingredients are not yet arriving fast enough.
Why Big Tech’s own playbook isn’t enough
Over the past decade, hyperscalers used long‑term power contracts to catalyze wind and solar, earning credit for decarbonizing the grid. That model is straining:
– Additionality and timing matter more. Buying generic annual renewable credits does not guarantee that a data center’s every hour is carbon‑free. The industry is shifting to “24/7” matching and to projects that add new clean capacity where and when it’s needed. Those deals are fewer, more complex, and more expensive.
– Siting is the new frontier. The cheap land, close to fiber and skilled labor, isn’t always where the cleanest, firmest power is. Utilities are starting to steer loads to zones with spare capacity, while local opposition in “data center alleys” grows.
– Betting on new tech carries risk. Hyperscalers are exploring nuclear (including advanced designs), long‑duration storage, geothermal, demand‑responsive computing, and even fusion PPAs. These bets could pay off, but they won’t meet next year’s load.
What a workable path looks like
There is no single lever. The practical answer is a portfolio that blends speed, cleanliness, and cost discipline—accepting that different regions will emphasize different tools.
– Targeted gas, with guardrails. Fast‑track a limited set of efficient gas plants or upgrades tied to clear reliability needs and paired with tight emissions controls, fuel security plans, and retirement schedules. Use contracts that prevent socializing all grid‑upgrade costs onto general ratepayers when a single large customer drives them.
– Accelerate transmission and interconnection. Build out high‑capacity lines between renewables‑rich regions and major load centers; standardize interconnection studies; move to “connect and manage” where safe; and adopt performance‑based incentives for utilities to deliver upgrades on time.
– Push “24/7” clean power from niche to norm. Encourage hourly matching through market products and disclosure, so corporate buyers pay for carbon‑free megawatt‑hours when they are scarce, not just cheap. That price signal pulls storage, geothermal, hydro uprates, and clean firm power into the mix.
– Site smarter. Incentivize data centers to locate near existing clean power and robust grids—adjacent to hydro and nuclear plants, in regions with surplus wind and solar, or in industrial parks built around transmission hubs. Where communities host new campuses, align benefits with impacts: tax agreements that fund schools and water infrastructure, noise and traffic standards, and public transparency on energy and water use.
– Make demand flexible by design. Treat computation as a grid resource: schedule training jobs around renewable peaks; curtail non‑urgent workloads during scarcity; deploy on‑site batteries for fast frequency response; and adopt dynamic tariffs that reward flexibility. Build flexibility into service‑level agreements so customers opt into carbon‑aware or price‑aware modes.
– Squeeze more from every watt. Advance chip and system efficiency (specialized accelerators, lower‑precision math, better compilers), liquid cooling, and waste‑heat reuse for district energy. Efficiency won’t outrun demand growth on its own, but it changes how much supply we need to find.
– Diversify clean firm supply. Extend life of safe existing nuclear, pilot advanced nuclear where community consent and site readiness exist, scale enhanced geothermal where resource and drilling capability align, and expand utility‑scale storage (both lithium‑ion and long‑duration) to cover evening and multi‑day gaps.
– Clarify who pays. Update tariffs so large new loads fund the marginal grid investments they trigger, without deterring economically valuable projects. Cost‑allocation clarity reduces backlash from residential ratepayers and makes local approvals easier.
The politics won’t get easier
Trump’s critique taps into real anxieties: rising bills, fears of blackouts, and suspicion that elite digital firms are commandeering public infrastructure. But any administration will face the same physics and timelines. Rolling back clean‑energy incentives raises the cost of decarbonizing big loads; green‑lighting only renewables without transmission invites bottlenecks; blocking data centers pushes investment—and jobs—elsewhere without shrinking global emissions.
The more pragmatic question is not whether Big Tech should use less power—demand will grow because the economy wants what that power enables—but how to channel that demand to upgrade America’s power system in ways that improve reliability, keep costs in check, and cut emissions. That means pairing near‑term, sometimes unglamorous steps (selective gas, grid upgrades, siting discipline) with structural reforms (hourly carbon accounting, transmission build‑out, flexible demand) and a steady pipeline of clean firm resources.
In short: the crusade spotlights a genuine problem. The answer is neither to slam the brakes on computing nor to wish away grid constraints. It is to use the leverage of large, creditworthy buyers and clear public policy to build a power system that can handle the AI era—reliably, affordably, and, over time, cleanly. There are no easy solutions, but there is a workable path if politics can tolerate nuance and sequence.
