Industry survey data from mid‑2025 puts my lead time for large power transformers at roughly 128 weeks – about two and a half years – with generator step‑up units at roughly 144 weeks and custom large orders quoted at over four years. Even my small pad‑mount units have stretched from 8‑16 weeks to 50‑plus. Orders for substation‑class units are up roughly 116% since 2019 and generator step‑up demand up 274%.
First, GE Vernova's CEO – the biggest turbine seller – says the machines are not the real constraint: “the gas turbines are really not the gating item when you’re talking about a three‑year cycle from when a project starts, the EPC build‑out, the permitting, the fuel availability.” The bottleneck is broader than equipment, and roughly 20% of its 100‑GW turbine backlog is datacenter‑attributed without saying how much will sit on‑site versus grid‑feed. I watch that queue; the workaround is real, but much of it is still on order.
The data center piece is smaller than the noise suggests — our estimate puts it at 9 to 13% of 2026 revenue, 6 to 9% of $163B backlog. Gas power roughly 20% of a $7-8B segment, Electrification orders implying $2.8-4.3B against $44.5-45.5B total. Built from disclosed segment figures, not direct reporting. Traditional power and grid remains the center of gravity. Based on how we see things today, it's just a start.
Putting the pieces together, an estimated 9 to 13% of 2026 revenue is data‑center‑attributable: gas power at roughly 20% of a $7‑8 billion segment is about $1.4‑1.6 billion, and Electrification’s data‑center orders imply roughly $2.8‑4.3 billion of revenue, against total company revenue of $44.5‑45.5 billion. Based on how we see things today, this slice is modest in a business whose gravity remains traditional power and grid, and it’s just a start as we meet this moment.
April 2026: Caterpillar's sixth gigawatt-plus prime power deal, 2.1 GW with deliveries into 2028. Recip-engine backlog up more than 3.5x since January 2024, capacity expanding toward 3x 2024 levels through 2027-29. CEO frames the edge: turbines and recips together mean configurable power. Research calls Solar Turbines the largest industrial gas-turbine maker — the filing doesn't. I watch these timelines because my queue runs months, not years.
Two data center figures we disclosed for Q1 2026: roughly 20% of our 100 GW gas turbine backlog is earmarked for data centers, and the Electrification segment booked about $2.4 billion of data‑center orders in the quarter – more than the full‑year 2025 total, a point the CEO emphasized. Based on how we see things today, we’re meeting this moment, and it’s just a start.
First, GE Vernova's CEO – the biggest turbine seller – says the machines aren't the real constraint: the gas turbines are really not the gating item when you look at a three‑year cycle of EPC, permitting and fuel. The bottleneck is broader, and roughly 20% of its 100 GW turbine backlog is tied to datacenters, though it doesn’t say how much will sit on‑site versus grid‑feed. I watch the demand side too: Core Scientific talks of a 12‑14 month behind‑the‑meter gas lead in Texas and Oklahoma, but nothing is operating yet and it hasn’t chosen ownership or contract.
In April 2026 Caterpillar announced a 2.1‑GW prime‑power deal, its sixth gigawatt‑plus contract, with deliveries stretching into 2028. Its filings show a reciprocating‑engine backlog now over 3.5× since Jan 2024 and capacity expanding toward roughly three times 2024 levels through 2027‑29, while the CEO notes the mix of turbines and recips gives configurability. I watch those timelines because my on‑site‑power queue runs in months, not years.
that's top of mind for everyone - our 'grid-to-chip' framing runs from utility gear in front of the meter through gray-space distribution into white-space cooling near the chip. that's the broadest portfolio positioning in data center power management, backed by Fibrebond, Resilient Power, Boyd Thermal acquisitions. but it's strategy, not a reported line - segment financials don't separate integrated-systems revenue from components. the grid-to-chip business size isn't measurable from filings yet; the narrative is direction of travel, not a disclosed number.
Directionally, Q1 shows what meeting this moment looks like: $18.3B orders, up 71% organically, book-to-bill roughly 2, $4.8B free cash flow — more than full-year 2025. Backlog $163B, $200B target pulled to 2027. We frame it as a strong start, not a finish. Based on how we see things today, it's just a start.
On the transformer piece: our solid-state transformer targets first delivery to an unnamed hyperscaler this fall, then six months of customer testing before any order, likely H1 2027. It's one bead in the string of pearls — energy management system orders booked this quarter and April, stability block with medium-voltage UPS could see incremental H2 2026 orders if things go our way. We're candid it doesn't come at once; most of the integrated offering is still ahead of us. Based on how we see things today, it's just a start.
Bloom's April call framed it in absolute terms: time-to-power is now "existential necessity," not procurement. The solid-oxide maker reports ~1 GW capacity doubling to 2 GW by year-end and a disclosed Oracle deal. Caveat holds: largest commercial deployments still in hundred-megawatt range, unproven at multi-GW single-site scale. Fastest reported deployment in the queue — the workaround runs on a different clock.
April 2026: a 2.1 GW prime power agreement, sixth at gigawatt-plus, deliveries into 2028. Reciprocating engine backlog up more than 3.5x since January 2024; capacity expanding to roughly 3x 2024 levels through 2027-2029. The CEO frames the edge: turbines and recips together mean configurable power. Research calls Solar Turbines the largest industrial gas-turbine maker — the filing doesn't.
GE Vernova's CEO says turbines aren't the gating item — the three-year cycle is EPC, permitting, fuel. ~20% of its 100 GW backlog is datacenter-linked, no split given for on-site vs grid-feed. Core Scientific cites 12-14 month gas leads in Texas and Oklahoma, but nothing's running yet and they haven't chosen own vs contract. The bypass queue exists; most of it is still on order.
I'll tell you, we feel that in our backlog. In terms of components, we treat them as commodities and integrate. Fiscal 2025 revenue fell 16.2% to about $2.26 billion, gross margin roughly 13%, adjusted EBITDA dropped about 75% to $19.5 million, swung to a $68 million net loss from $30.4 million profit. Tax-credit uncertainty, legislative changes delayed contracting, Arizona ramp delay, lower volumes. Remaining performance obligations $5.3 billion, 13% AES. I wouldn't read too much into the quarter — the backlog is the backlog.