Demand Flexibility Has Replaced Generation as the Grid's Scarcest Asset
University of Utah modelling values real-time data centre flexibility at up to $590 million annually in Western US grid savings.
A University of Utah study published on July 6 (2026) modelled the Western US power system and found that scheduling data centre workloads off-peak could reduce grid operating costs by an estimated $62 million a year. Add real-time demand response, where data centres adjust consumption dynamically to match grid conditions, and the annual saving rises to $590 million. Those figures describe what is achievable with existing infrastructure, not a future technology.4
The scale of the underlying demand surge gives those numbers weight. The IEA estimated in May 2026 that global electricity consumption will grow at an average of 3.6% annually between 2026 and 2030, the fastest sustained pace in 15 years, driven by AI infrastructure, electric vehicles, industrial heat and air conditioning. Closing the resulting capacity gap would require annual grid investment of roughly $600 billion, about 50% above the $400 billion currently being deployed.2
Grid operators are running out of margin before that investment arrives. ENTSO-E, the body representing European transmission system operators, warned in May 2026 that unmanaged data centre growth could force grid operators to curtail renewable energy penetration to maintain frequency stability — a reversal of the trajectory European power markets have pursued for a decade. More dispatchable gas plant, not less, would be the practical consequence.3
Japan's intraday price curve shows how acute the mismatch has become. Day-ahead markets in the Tokyo area have recorded prices around ¥60 per kWh in the late afternoon as solar output fades. Hokkaido and Tohoku meanwhile registered prices near ¥3 per kWh at the solar midday peak. Japan's Ministry of Economy, Trade and Industry opened a demand response framework on April 27 (2026) directed at using large, shiftable loads to smooth that spread, rather than building fast-start generation to cover it.5
Investors have started placing capital accordingly. Fluence Energy shares closed at $24.16 on May 8 (2026), up 98.2% in a single week, after the battery storage and grid software company disclosed master supply agreements with two hyperscalers and a record $5.6 billion order backlog. The move reflected a broader rotation into companies positioned at the intersection of storage, software and AI power demand.1
But Fluence's underlying financials keep the enthusiasm calibrated. The company posted adjusted EBITDA of just $2.0 million in the first quarter of 2026, its fourth consecutive quarter in marginal positive territory, with non-GAAP gross margins of 52%. Before the May 2026 surge, shares had declined roughly 39% year to date. Its PowerTrack solar management platform administers 37.5 gigawatts of assets under management, with annual recurring revenue guided to $65 million to $70 million by year-end, a modest base relative to the capital deployment the IEA's investment numbers imply.1
Jeanine Johnson, a former PJM Interconnection board member, argued in Utility Dive on July 15 (2026) that grid operators need resources capable of responding in seconds as large loads come online across the country, not the multi-year timelines of conventional generation additions. Utilities are already revising load forecasts upward faster than interconnection queues can process new projects.6
The gap between what flexible data centres could theoretically provide and what they currently deliver comes down to market design. The University of Utah's $590 million figure assumes grid operators have the settlement rules, software and dispatch authority to call on data centre flexibility in real time, infrastructure that is absent or underdeveloped across most US and European markets. How quickly regulators build those frameworks, relative to the pace at which hyperscalers commission new capacity, will determine whether demand flexibility stays a modelling assumption or becomes a resource with a clearing price.4,6