Skip to content

Why AIEnergy started

AIEnergy started to correct the erroneous “Data centers are driving up electric rates” stories. AI was just an excuse for the companies involved to raise rates and use obsolete technologies. In very few places did AI raise overall demand more than 2 or 3% per year, less than renewables + batteries + demand response are able to serve. NYT & WSJ believed the hype.

Everyone in energy knows the tech is ready to handle the demand and actually lower prices, but the system is effed up. Good reporting in the energy press and from researchers makes that clear.

I learned from some very smart people and will write as long as I have something to contribute. I’m veteran tech reporter Dave Burstein (daveb@dslprime.com), and I love to hear from you. It’s a great service when you catch my mistakes.

Bloom’s 800V DC Fuel Cells: $Bs Savings, By Its Own Math

The numbers look good, but Bloom is far from proving them in the field. 800V will soon come on strong. Dave

Bloom Energy — press release, September 16, 2026. Source title: “Bloom Energy’s 800V DC-Native Power Can Cut Billions from AI Data Center Costs, Reduce Power Use, and Eliminate Need for Transformers.” Bloom Energy newsroom

Bloom Energy says fuel cells that generate 800V DC power directly, with no AC-to-DC conversion, could cut non-compute capital costs for a 1 GW AI data center by $3.6 billion (27%) and five-year total cost of ownership by $5.5 billion (9%).

Why the grid’s AC habit is breaking

The power grid was built around moving electricity long distances as AC, then stepping it down with transformers for AC-powered equipment. GPU racks don’t fit that mold: they’re dense enough to need large amounts of DC power delivered in one place, and chipmakers are already redesigning around it. NVIDIA has specified 800V DC starting with its Rubin Ultra generation and Kyber rack architecture, due in 2027, a move NVIDIA itself has detailed at GTC events and through OCP working groups — independent of anything Bloom says about it.

Skipping the conversion

Bloom’s pitch is that its solid oxide fuel cells generate 800V DC electrochemically on site, so there’s no AC stage to convert from. That removes transformers, switchgear, and the transport equipment that normally sit between generation and the rack — equipment facing multi-year lead times and tight copper supply. Less conversion hardware is also where Bloom’s cost model comes from: $3.6 billion less non-compute capital spending and $5.5 billion lower five-year ownership cost for a 1 GW site, versus a traditional AC build-out.

The numbers are Bloom’s own

Those figures come from Bloom’s internal “800V DC Total Cost of Ownership model,” using Bloom’s own next-generation AI rack assumptions — the release says actual project economics will vary by site. CEO KR Sridhar and CMO Natalie Sunderland are both quoted; Sunderland cites Bloom’s own “2026 Mid-Year Data Center Power Report” for the claim that DC-based architectures will make up 58% of new data center deployments by 2030. No independent cost study or third-party audit of the comparison is cited, and the release carries forward-looking-statement safe-harbor language aimed at investors, not just a technical readership — a signal this is as much a financial pitch as an engineering one.

This is Bloom Energy’s own press release and economic model for its own product; the 800V DC shift itself is independently confirmed by NVIDIA, but the $3.6B/$5.5B savings and the 58%-by-2030 figure are Bloom’s unaudited numbers, not a third-party finding.

authored by Dave Burstein with close collaboration by Claude