A charged battery is stored WORK, not just stored backup. Honest math: a 13.5 kWh home battery (~12 kWh usable) running a local AI machine at 400-600W funds roughly 20-30 hours of continuous inference, a full overnight batch schedule with room to spare. In San Diego's post-NEM-3 rate world, where export credits pay a fraction of retail, spending surplus electrons on compute that replaces cloud subscriptions and outsourced hours usually beats selling them back. And an EV is this idea at 5x scale: a 60-100 kWh battery with wheels.
Backup was the first job. Rate arbitrage was the second. Funding the night shift of your own private AI is the third, and almost nobody is doing it yet.
Inference machine at 400-600W under load. Usable capacity discounted ~10-15% from nameplate. Bursty real-world work stretches every number here.
| Storage | Usable energy | Continuous AI work it funds |
|---|---|---|
| Small home battery (~10 kWh) | ~8.5-9 kWh | ~15-22 hours |
| Standard home battery (~13.5 kWh) | ~11.5-12 kWh | ~20-30 hours |
| Stacked pair (~27 kWh) | ~23-24 kWh | ~40-60 hours |
| EV with V2X (~60-100 kWh) | ~50-85 kWh | Multiple WEEKS of nightly batch work |
| Path for a surplus kWh | What it's worth |
|---|---|
| Export to grid (NEM-3-era credit) | A fraction of retail rate, often single-digit cents in low-value hours |
| Charge the battery, run compute at night | Replaces grid power you'd buy at premium evening rates AND cloud/AI subscription spend AND outsourced hours |
| The operator's arbitrage | Electrons in at near-zero marginal cost, work product out. The spread is the whole thesis |
Batteries earned their place as backup, then as rate arbitrage. The third job is quietly the most interesting: a battery lets an operator time-shift intelligence, charging on midday surplus and funding a night shift of AI work that would otherwise cost real money in subscriptions, outsourcing, or daylight hours. Start with the machine and a schedule; add the battery math when you see what the night shift produces. And watch V2X: the moment vehicles feed homes at scale, every driveway in San Diego holds a week of thinking.
We spec the inference machine, wire battery-aware scheduling, load the private AI stack, and hand you an operator-owned night shift: your storage, your electrons, your data doing work while you sleep.