Sixty-One Percent of Energy Investment Is Now Going Into Electricity
On sunny spring afternoons in California, something counterintuitive happens: the grid operator tells solar farms to make less electricity, not more. In 2024, the California Independent System Operator curtailed 3.4 million megawatt-hours of utility-scale wind and solar output, a 29% jump from the year before, mostly because transmission lines didn’t have room to carry the power away. The panels were fine. The sun was fine. The wires were the problem.
That small, specific breakdown is a preview of the argument the head of the International Energy Agency made in Seoul on September 17. Fatih Birol told reporters that electricity demand is now growing three times faster than overall energy demand, and that 61% of global energy investment this year has gone into electricity, versus 39% into fossil fuels. It sounds like a clean verdict: the money has already decided that wires and switchgear matter more than wells and pipelines. But Birol’s own description of the situation was less triumphant than the number suggests. Electrification, he said, is “not a rose garden.” The two things actually holding it back, he added, are grid capacity and cost, not a shortage of capital looking for somewhere to go.
That distinction is the real story here, and it matters well beyond energy-policy circles, because it’s a direct answer to a question fund managers and utility executives are also asking: whether the better decade-long bet is on the companies that generate electricity or the ones that move it. Investors have made up their minds about electricity in general. What they haven’t solved is which part of the electricity system the money goes to, and the part that keeps getting underfunded is exactly the part that decides whether all that new generation ever reaches anyone.
Where the 61% figure comes from
The number wasn’t from a report. Birol gave it at a joint news conference with South Korea’s energy minister, Kim Sung-whan, where the two launched a regional partnership called RISE ASIA, short for Resilient & Integrated Strategy for Energy Security in Asia. Its immediate purpose is coordinating Asian responses to the energy disruptions caused by the war in Iran; its longer-term purpose is nudging Asian economies toward electrification. South Korea brings a practical asset to that second goal: it’s a major manufacturer of the batteries, wires, cables and transformers the buildout requires, which Birol framed as a genuine head start for electrifying the wider region. Seoul and the IEA are now targeting 35% electrification across participating economies by 2035, a goal to be formally announced around the COP31 climate summit later this year.
None of the underlying demand is coming from climate policy alone anymore, which is part of why the number keeps climbing regardless of which government is in power. Solar and wind account for 98% of the growth in global electricity-generation investment over the past decade, but the newer growth showing up on the demand side is dominated by things people simply buy and plug in: electric vehicles, heat pumps, air conditioning, industrial motors. Investment tied to EVs and combustion-engine efficiency alone is estimated to have roughly quintupled since 2015, reaching about $220 billion in a recent year. An electric motor or a heat pump is just a more efficient converter of energy than the engine or furnace it replaces, and that efficiency shows up in the data as demand for a wire rather than a fuel truck.
The figure also fits a trend the IEA has been tracking for a while in its own numbers. The agency’s flagship investment report projects total global energy spending will hit $3.4 trillion in 2026, a 5% rise from 2025. Of that, roughly $2.2 trillion is expected to go collectively to renewables, nuclear, grids, storage, low-emission fuels, efficiency and electrification, versus about $1.2 trillion for oil, gas and coal combined. That’s a broader bucket than Birol’s 61%, since it folds in efficiency spending and fuels alongside pure electricity infrastructure, but the direction matches, and it isn’t new. The point where combined clean and grid spending pulled decisively ahead of fossil fuels arrived only a few years ago. Before that, the two had been running close to even.
Birol was also asked, pointedly, whether all this amounted to overinvestment in power infrastructure, given growing doubts in some quarters about whether the AI spending driving part of that demand can actually be sustained. He didn’t back away from the trend. He just didn’t pretend it was simple, either.
What the 61% number leaves out
Here’s what gets lost inside a single headline percentage: “electricity investment” is not one category. It covers power plants, transmission lines, distribution networks, battery storage, and the substations and transformers that stitch it all together. For years, IEA officials have pointed out that spending has piled up disproportionately on the generation side of that list while the delivery side lags behind it.
The lag shows up as an actual, physical queue. At the launch of the IEA’s most recent renewables report, Birol said a record 750 gigawatts of new renewable capacity would be added globally in 2025, while 1,700 gigawatts, more than double that amount, sat stuck waiting for a grid connection. Two years earlier, that queue stood at 1,500 gigawatts; it has kept growing even as capacity gets built, not shrinking. Birol called the backlog “economically criminal,” and by the IEA’s own accounting, only about fifty cents now get spent on grids for every dollar spent on new power plants, a ratio that was closer to sixty cents a decade ago and has been sliding since. He’s used a blunter image for the same imbalance elsewhere: engineering an impressively fast, capable car, and forgetting to build any roads for it to drive on.
California’s curtailed solar farms are what that imbalance looks like once it stops being an abstraction. So is South Africa, where a lack of grid capacity has contributed to ongoing power shortages, and where, in one recent renewable-energy auction, no wind projects were awarded at all because there was nowhere to connect them. Generation gets financed, permitted and built in three to five years. A major new transmission line, once permitting is included, routinely takes far longer, in parts of Europe over a decade. Multiply that mismatch across enough countries and you get exactly what the 61% figure obscures: a lot of money chasing megawatts that have nowhere to go once they’re made.
Why AI keeps entering this conversation
Part of the reason this argument is happening in 2026, and not five years ago, is sitting inside data centers. IEA figures published earlier this year found that electricity demand from data centers rose 17% in 2025, while demand from AI-focused facilities specifically jumped 50%, both far outpacing the roughly 3% growth in electricity demand overall. The agency’s broader modeling has data-center electricity use more than doubling by 2030, to something in the range of 945 terawatt-hours, more than Japan’s entire electricity consumption today.
That demand doesn’t spread out evenly the way a growing population’s electricity use does. It shows up concentrated: a handful of gigawatt-scale campuses need a specific substation, in a specific place, upgraded on a specific timeline, right when a chipmaker’s construction schedule says so. A grid that looks merely adequate on a national average can still fail an individual data-center developer waiting on one substation upgrade, which is precisely the kind of localized bottleneck a figure like “61% of investment” can’t see.
That mismatch is already changing corporate behavior in ways the 61% figure doesn’t capture either. Rather than wait years for a grid interconnection, some of the largest data-center developers have started contracting directly with power plants or building generation on-site, effectively routing around the queue instead of standing in it. That’s a rational response to a slow-moving grid. It’s also a sign that the grid itself, not electricity in the abstract, has become the thing companies with enough capital are trying to engineer their way past.
What it would actually cost to close the gap
The IEA has attached a number to what closing this would take: global grid investment needs to reach roughly $600 billion a year by around 2030. Recent annual spending has been running closer to $390 billion to $400 billion, which puts the shortfall somewhere in the neighborhood of $200 billion a year, every year, for most of the rest of this decade.
Even where that money exists, there isn’t enough industrial capacity to spend it quickly. Transformer lead times have nearly doubled over the past two to three years, and manufacturers reported record order backlogs as recently as 2024. The IEA has separately flagged a shortage of roughly 1.5 million skilled workers globally, linemen and grid engineers among them, needed to build and maintain the system at the pace demand now requires. By the agency’s own estimate, the world needs to add or refurbish something like 80 million kilometers of power lines by 2040, roughly the length of rebuilding the entire grid that exists today.
None of that is cheap in any market. IEA modeling puts the cost of adding one terawatt-hour of combined generation and grid capacity at $30 million to $110 million in emerging economies and $75 million to $150 million in advanced ones. That’s the affordability half of Birol’s two-part problem: even fully financed, the power has to reach a household or a factory at a price low enough that they choose it over burning fuel directly, rather than a grid connection that costs more than the alternative it’s supposed to replace.
The market has already started pricing this in
None of this is a secret to public markets. GE Vernova, which sells the gas turbines, transformers, switchgear and high-voltage systems this buildout requires, ended the second quarter of 2026 with a total backlog of $176 billion, up $13 billion in three months. Its Electrification segment, the part of the business selling grid equipment rather than generators, grew even faster: equipment orders rose 86% in the first quarter of 2026 and 66% in the second, pushing that segment’s book-to-bill ratio, orders taken divided by revenue shipped in the same period, as high as 2.5 at one point in the year. That means the company was booking well over twice as much new grid-equipment business as it could physically ship.
Siemens Energy and Baker Hughes reported similar patterns the same year: record quarterly order intake, backlogs stretching years into the future, and executives at all three companies pointing to the same three demand drivers, data-center power, LNG buildout, and grid reliability. It’s a notable reversal. For most of the past decade, turbine and grid-equipment manufacturers were managing overcapacity left over from a slower-growth era. Now they’re rationing manufacturing slots years in advance. That’s the same imbalance Birol described in Seoul, translated into order books: the world wants grid capacity considerably faster than the industry that builds it can currently supply, and the companies that make transformers and switchgear are the ones actually cashing in on that gap, not necessarily the ones building new power plants.
So, are grids and electrical equipment becoming more important investments than oil and gas production? At the level Birol was describing in Seoul, yes, clearly, and that’s been true for at least a couple of years now, not just this one. But the more useful answer sits one layer down. The world has decided, with real money, to build the generation half of an electrified economy considerably faster than it’s decided to build the half that actually delivers that electricity anywhere. Capital isn’t the constraint anymore. The wires are.
Whatever happens next depends less on a press conference in Seoul than on transformer factories, permitting offices, and the people willing to become linemen. Until the delivery side catches up, the extra dollar going into electricity this year is only as useful as the wire it eventually finds to travel down.