Why we're building a battery-backed AI cloud
AI workloads need 10x the power of traditional compute. The grid can't keep up — interconnection queues are 4 to 7 years long. We bet that owning the power layer is the only way to ship AI compute at the pace the market needs.
If you've tried to bring a new AI compute hall online in the last three years, you already know the problem. It's not chips. It's not networking. It's not even money, though capital is tight. It's power.
Interconnection queues at major US ISOs now run 4 to 7 years for new large-load customers. Some markets — Northern Virginia, Phoenix, parts of Texas — are functionally closed to new AI tenants until the late 2020s. ERCOT, PJM, MISO, CAISO: every single one has multi-year backlogs for projects over 100 MW.
The math has changed
A 5 MW AI compute hall used to be a mid-sized project. Today it's a hyperscaler-class facility. Power densities per rack have tripled. Power densities per square foot have increased 5x. The grid was not designed for this.
When we started SmartTec in 2023, the conventional wisdom was: build where the grid is, and wait. We chose the opposite bet. Build where the grid is weak, and bring your own power.
Why batteries, not gas
Gas turbines are the obvious answer — they're proven, they're fast to deploy, and they don't need grid interconnection. But they burn fuel continuously, they fail to start under load, and federal customers won't touch them under NDAA §889 / FEOC.
Megawatt-class battery storage (BESS) does the same job, cleaner and quieter. Lithium iron phosphate cells are now commodity-priced. Inverters are mature. The control problem — coordinating thousands of cells to behave like one stable generator — is a software problem, which is what AURA was built to solve.
What it buys you
- →Your compute hall comes online in 90 days, not 4 years.
- →You don't pay demand charges from the local utility.
- →You can site in places with cheap land and low disaster risk, not the few congested grid nodes.
- →When the grid drops (and it will), you keep running. Sub-10ms failover.
What it doesn't solve
Batteries are not free power. You still need to charge them, usually from the grid. The economics work when grid prices are low (overnight, weekends, shoulder seasons) and you discharge during peak. AURA does the arbitrage automatically. But you do still need a grid connection — just a smaller one than you would otherwise.
And batteries don't help with cooling. A 5 MW compute hall rejects 5 MW of heat. That's a separate, hard engineering problem. More on that soon.
Three design-partner slots open for Q4 2026 power-on. Locked launch pricing for 12 months. Direct engineering access.