Why two SpaceX alumni are betting on solar and batteries to power the AI craze

Ambrosia Energy says it can deliver power at about $100 per megawatt-hour by pairing solar panels with lithium-ion batteries, below the roughly $107/MWh cost of a new combined-cycle gas turbine. The startup claims it can build power plants in 12 months, has a project underway in West Texas, and is targeting 20-30 MW initially with ambitions for gigawatt-scale deployment by the end of the decade. It recently raised funding from DFJ Growth, signaling investor interest in its low-cost, fast-deployment power model.

Analysis

This is a credible threat to the marginal economics of firm power for hyperscalers, but the real disruption is not solar-plus-storage itself — it is the compression of delivery time and balance-sheet intensity for large-load customers. If a developer can genuinely contract, finance, and energize behind-the-meter capacity in roughly a year, it shortens the planning cycle for AI/data-center expansion and weakens the negotiating leverage of incumbent utilities, gas turbine OEMs, and merchant power developers. The second-order winner is likely the equipment-and-interconnection ecosystem around renewables: EPCs, inverter suppliers, switchgear, transformers, and battery integrators that can scale with modular deployment.

The market is probably underestimating how this changes hyperscaler capex timing. Instead of waiting years for gas turbine backlog relief, large buyers may increasingly prepay for dedicated capacity as a de-risked infrastructure wedge; that shifts the bottleneck from generation technology to land, transmission, and permitting. In that world, the scarce assets are grid interconnects and sites with cheap curtailment-heavy solar, which should benefit developers with early land-bank control while pressuring gas turbine pricing power and long-duration project IRRs.

The biggest risk is execution risk disguised as manufacturing scale: a small pilot can look superb, but multi-hundred-MW deployments will expose supply-chain constraints, warranty issues, battery degradation assumptions, and interconnection delays. The concept is also highly rate-sensitive because the economics rely on project finance and low-cost capital; a higher-for-longer cost of capital can erase much of the apparent cost advantage versus regulated or utility-scale alternatives. On the other hand, if the model works, the shock will likely arrive over 12-24 months rather than immediately, because hyperscalers validate infrastructure partners slowly and in staged commitments.

Contrarian angle: the market may be too focused on the batteries and not enough on the data-center load-growth that makes any fast-to-build capacity valuable. Even if this startup itself never scales to gigawatts, the signaling effect should raise the multiple on developers and suppliers that can deliver behind-the-meter power fast, while also increasing perceived obsolescence risk for turbine backlog economics. The more important question is whether this becomes a procurement standard for AI campuses, not whether one startup wins the market.

Key Decisions for Investors

• Long NEE / short a basket of gas-turbine and conventional power-equipment exposure for 3-6 months: the trade captures a shift toward fast-build renewable capacity, but size modestly because execution risk remains high.
• Buy near-dated calls on BEP or FSLR into any data-center power headlines over the next 1-2 quarters: if hyperscaler procurement shifts toward modular clean power, the re-rating can happen before revenue fully shows up.
• Short GEV or other turbine-adjacent beneficiaries on strength if backlog-driven optimism extends another 5-10%: the setup is attractive as a relative-value hedge against a faster-than-expected alternative supply path.
• Long ITRI or grid-infrastructure names on 6-12 month horizon: regardless of the winner, interconnection, metering, and distribution upgrades remain the bottleneck and should monetize best.
• Avoid chasing the startup theme itself; instead, express the thesis through public-market enablers and land-bank developers until there is proof of scalable deployments and signed multi-hundred-MW contracts.