The Grid Is Getting $550 Billion a Year. The IEA Says That Is Not Enough.
US data centre demand went from 23 GW to 42 GW in three years and utilities are planning $1.3 trillion of capex. The binding constraint is not capital — it is transformers, crews and permitting years, which is why storage keeps winning on speed.
Published: 11 September 2026
Grid storage · 6 min read
Where things stand
Written 11 September 2026, on IEA Electricity 2026, S&P Global utility capital expenditure research and J.P. Morgan grid analysis.
Grid investment is running at around $550 billion in 2026, roughly 20 per cent up, and global spending is projected at $5.8 trillion across 2026 to 2035 with China, the EU and the United States accounting for two thirds. US utilities alone are forecast to spend about $1.3 trillion between 2026 and 2030.
And it is not enough. The IEA puts the requirement at roughly 50 per cent above today’s $400 billion a year by 2030 to meet forecast demand.
What changed the forecast
US data centre electricity demand went from 23 GW in 2023 to 42 GW in 2026. S&P Global CERA has large loads adding 374 TWh and over 45 GW of peak through 2035.
Utilities spent two decades planning around flat demand. That assumption is gone, and the replacement arrived faster than a transmission line can be permitted.
The binding constraint is not money. Capital is available — the capex forecasts prove it. What is scarce is transformers, switchgear, skilled crews, and the years a transmission route takes to permit. This is the same shortage that shows up on a battery project as an eighteen-month wait for a connection.
Why storage is the fastest thing on the list
When a network is constrained, there are broadly three responses: build more lines, build generation nearer the load, or make better use of the capacity already there. The first is slow and contested. Storage is the third.
- •Deferring a network upgrade by shaving the peak that would have triggered it
- •Relieving a congested corridor by charging on one side and discharging on the other
- •Providing the fast frequency response an inverter-dominated grid needs and thermal plant cannot deliver
- •Being commissioned in months rather than the years a line takes — the Khavda build took about ten months on site
This is why grid investment programmes increasingly have a storage line in them rather than treating storage as an alternative to network spend. Our explainer on what grid-scale storage actually does works through each of those functions.
The Indian version
India’s constraint is differently shaped. The problem is less a sudden data-centre load than a very large solar fleet generating at midday against demand that peaks after sunset, plus distribution networks that were not built for either.
The response is visible in policy: viability gap funding widened for standalone storage, tenders moving from two-hour to four- and six-hour products, and the CEA now proposing mandatory co-located storage on new renewable plant. Different cause, same conclusion: the grid needs flexibility faster than it can be wired.
Sources
Reporting this piece draws on. Figures were correct as published; scheme terms and commodity prices move.
- Surging energy demand puts US utility capex forecast near $1.3T in 2026–30S&P Global
- Grids — Electricity 2026International Energy Agency
- Grid Resilience: Neglected No MoreJ.P. Morgan
Frequently asked questions
How much is being invested in electricity grids?+
Around $550 billion in 2026, roughly 20 per cent up, with $5.8 trillion projected globally across 2026–35 and about $1.3 trillion from US utilities alone between 2026 and 2030. The IEA nonetheless puts the requirement at roughly 50 per cent above today’s $400 billion a year by 2030.
What changed the demand forecast?+
Data centres. US data centre electricity demand rose from 23 GW in 2023 to 42 GW in 2026, and S&P Global CERA expects large loads to add 374 TWh and over 45 GW of peak through 2035. Utilities had planned around flat demand for two decades.
Why does storage help when the grid is constrained?+
Because it makes better use of capacity that already exists, and it can be built quickly. Storage defers upgrades by shaving the peak that would trigger them, relieves congested corridors, and supplies the fast frequency response an inverter-dominated grid needs. The Khavda plant took about ten months on site; a transmission line takes years to permit.
Want this applied to your own numbers?
Every solution page carries a calculator that shows its own arithmetic — storage savings, lithium against lead-acid, and home pack sizing.
Get in touch
