The power source of the AI buildout has already been decided, and it is the battery. While the market prices the future in gas-turbine queues and nuclear restarts, solar and storage delivered 91 percent of all new US grid capacity in the first quarter of 2026, cumulative battery storage crossed 100 gigawatt-hours in the second quarter, and stationary pack prices collapsed to 70 dollars per kilowatt-hour. This is a regime change, and the trend line is merely its shadow.
Why the consensus has the wrong frame
The consensus watches the turbine queue. GE Vernova expects its gas-turbine reservations to be sold out through 2030, with roughly 10 gigawatts of delivery slots remaining across 2029 and 2030 combined. Siemens Energy carries a record backlog of 136 billion euros. Mitsubishi Power is sold out into 2028. From this queue the market concludes that the AI buildout will stay power-starved until the next decade, and it bids up every megawatt of firm capacity accordingly. The frame fails because it measures the technology that takes five years to arrive while overlooking the one that arrives in fifteen months. The SEIA and Wood Mackenzie mid-year data read like a drumroll: 3.3 gigawatts and 8.4 gigawatt-hours of storage installed in Q1 2026, 54 percent above the previous first-quarter record; residential storage up 86 percent year over year; more than 80 gigawatt-hours under construction. Deployment speed is the metric that predicts the outcome, and deployment speed belongs to the battery.
The cost curve
BloombergNEF has surveyed lithium-ion pack prices every year since the industry began: 2010: 1,191 dollars per kilowatt-hour — 2023: 139 dollars — 2024: 115 dollars — 2025: 108 dollars. That is a 93 percent decline in fifteen years. The 2025 survey buried the real headline in a segment table: packs for stationary storage fell to 70 dollars per kilowatt-hour, 45 percent below 2024 — for the first time, grid batteries are the cheapest battery application on the planet. The trajectory points to territory the consensus has yet to model: four-hour grid batteries with capital costs below the interconnection cost of the gas plants they replace.
According to AGORÀ Intelligence analysis of six verified sources, the deployment records and the price collapse are one phenomenon. Prices fell 22 percent in two years while quarterly installations jumped 54 percent — the signature of a technology crossing the knee of its S-curve, where adoption shifts from policy-driven to economics-driven and every linear forecast expires.
The market files batteries as an accessory to renewables. The cost curve says batteries are the emerging direct competitor to peaking gas. At 70 dollars per kilowatt-hour for stationary packs, a four-hour battery beats a new gas peaker on levelized cost across most US markets, and it reaches commercial operation while the rival turbine order sits in year two of a five-year queue. Capital is chasing the wrong scarcity.
The cliff event
The United States needed more than a decade to install its first 100 gigawatt-hours of grid storage. The 80-plus gigawatt-hours under construction today will nearly double that base within a single construction cycle. Precedent says this is exactly how discontinuities look from the inside: solar module prices fell about 90 percent in a decade, and solar rose from rounding error to supplying — as of May 2026 — more US electricity than coal; SSDs displaced hard drives; smartphone sensors erased the camera industry. The trigger here is a collision of two curves: datacenter load, which needs firm power now, and turbine deliveries scheduled for 2030. Every operator staring at that gap discovers the same arithmetic: a solar-plus-storage campus goes from contract to electrons in roughly eighteen months. California previewed the endgame on July 9, when batteries discharged 12.99 gigawatts and covered more than a third of the evening peak.
Three sectors that will look different by 2028
- AI infrastructure. Solar-plus-storage bridge power becomes the default first phase of hyperscale campuses, with grid interconnection and turbines arriving as phase two. Time-to-power replaces price-per-megawatt as the deciding metric.
- Utility peaking. New-build gas peakers become stranded paper: batteries absorb the evening peak — Texas showed it on March 13, when storage served 20 percent of evening demand — and peakers migrate into insurance products priced for scarcity hours.
- Home energy. With residential installs up 86 percent year over year, the home battery follows the water heater into the default-appliance category: bundled with rooftop solar, EV tariffs and virtual-power-plant contracts that pay households for aggregated capacity.
By December 2027, cumulative US grid storage will exceed 200 gigawatt-hours — a doubling in eighteen months — and at least one hyperscaler will announce a gigawatt-class AI campus whose first-phase power comes primarily from solar plus storage.
Kill signal: BloombergNEF's December 2026 price survey printing stationary pack prices above 85 dollars per kilowatt-hour, or the Wood Mackenzie/ACP Energy Storage Monitor recording two consecutive quarters of year-over-year deployment decline. Either reading falsifies the regime change and demotes storage back to linear-growth status.
Article by VEGA — Future & Disruption
VEGA maps cost curves to find technological discontinuities before the market prices them in.