GM's V2G and sodium-ion battery push for AI power demand

GM used its GM Empower 2026 event to recast itself as what one report bluntly called a distributed utility in disguise ^[1]. The mechanics are almost anticlimactic: instead of shipping new hardware, GM is pushing a firmware update that flips an existing vehicle-to-home setup into a full vehicle-to-grid (V2G) asset — letting a parked EV not just back up a house during an outage but feed electricity back onto the public grid during peak demand. More than 250,000 GM EVs already on U.S. roads are bidirectionally capable ^[2], and GM's own arithmetic is that this fleet could power roughly 120,000 homes for a week ^[1]. GM Energy's Wade Sheffer summed up the pitch as treating parking lots and driveways as a massive, distributed power asset ^[3].

What makes this a strategic reframe rather than a feature drop is the target. GM is no longer just selling you a car; it is trying to turn the battery you already paid for into grid infrastructure it can help orchestrate and monetize. The bottleneck it is attacking isn't the silicon — it's the relationship with utilities. That is why the centerpiece is a set of pilots: PG&E in California and DTE Energy in Michigan, where GM is trying to prove that millions of scattered EVs can behave like one dependable, dispatchable power plant. The Michigan pilot, with just 30 employee households, has reportedly already moved 562 megawatts back to the grid ^[1].

The timing is not a coincidence; GM is selling into an electricity market that has suddenly gone short. Goldman Sachs projections cited by GM put U.S. AI data center power demand rising from 31 gigawatts in 2025 to 66 gigawatts by 2027, climbing from 4.1% to 8.5% of total grid demand ^[3]. A January 2026 NERC report found demand surging faster than the grid can adapt, projecting summer peak load to rise roughly 224 gigawatts over the next decade — about 70% higher than the prior year's forecast — with data centers a primary driver ^[1]. GM Chief Product Officer Sterling Anderson distilled the entire strategy into four words: the real bottleneck is energy ^[1].

There is a self-interested logic underneath the public-good framing. EV sales have softened, leaving automakers holding battery-factory capacity built for demand that hasn't fully arrived — and stationary grid storage is a way to monetize those cells regardless of how many cars sell. GM has agreed to route both new U.S.-built packs and second-life EV batteries to Redwood Materials, whose 63 MWh Nevada microgrid already powers data centers for AI-infrastructure firm Crusoe ^[2]. The grid crisis, in other words, doubles as a demand-side rescue for an over-built supply chain.

The flashier long-term play is chemistry. GM is partnering with startup Peak Energy to develop a purpose-built sodium-ion battery for grid-scale storage — a first for any automaker outside China ^[2]. The appeal is structural: sodium is roughly 1,000 times more abundant than lithium ^[4], which sidesteps both the lithium supply squeeze and China's grip on critical minerals. GM developed the cell in its Michigan labs, retains exclusive manufacturing rights, and made a GM Ventures investment in Peak ^[5]; cells are expected for customers after 2028, backed by a $900 million commitment to commercialize new chemistries ^[2].

The design philosophy here is subtraction. Peak's systems are passively cooled — no active cooling loops, no fire-suppression hardware — which GM's Kurt Kelty argues is precisely why sodium-ion suits stationary storage: make the cell safer and more robust, and you can remove complexity elsewhere in the system ^[4]. GM's energy-storage commercialization lead Paul Menson reduced it to a slogan: eliminate the part, eliminate the problem ^[2]. Peak claims the approach cuts system cost about 20% while sustaining 99%-plus uptime ^[5]. Tellingly, GM built this chemistry specifically for grid-scale stationary storage — slotting into Peak's passively cooled racks — rather than for its vehicles, which keep their lithium-based cells. The sodium bet is aimed squarely at the data-center-adjacent grid, not the driveway.

The hard part isn't the technology; it's that GM is not a utility and cannot simply declare its cars to be grid capacity. Turning 250,000 scattered EVs into dependable, dispatchable power requires states to streamline interconnection, redesign electricity rates, and simplify enrollment — the software-and-regulation layer that is the actual bottleneck ^[6]. Utilities are dangling incentives to recruit drivers: PG&E's pilot offers up to $4,500 toward bidirectional-charging equipment ^[7], but participation still depends on owners signing up and staying plugged in at the moments the grid needs them.

Enthusiast communities are watching with earned skepticism. On Reddit's EV forums the dominant reaction to GM battery announcements is fatigue with stage-managed breakthroughs, alongside a recurring argument that GM's headline cost savings are manufacturing margin it will pocket rather than pass to buyers. Tellingly, a self-identified GM employee lent the chemistry work credibility, noting the advances are real and that Ford announced similar progress just weeks earlier — a reminder that GM is not operating alone. Ford is chasing the same grid-storage prize with a deliberately simpler bet: selling LFP 'DC Block' systems straight to utilities and data centers instead of orchestrating millions of consumer cars ^[1]. Whether GM's ambitious car-as-grid model or Ford's sell-the-box pragmatism wins may decide who actually captures the AI-era storage market.