Slots Beat Specs: GE Vernova vs Mitsubishi vs Siemens
Who's winning and why?
ReadMeteor Labs follows the generators, turbines, transformers, switchgear, cable and memory behind every data center, from supplier order books to site permits, down to individual units.
Source: US Energy Information Administration
Spending on US data center construction has never been higher, and utilities and developers are ordering gas turbines and engines at a pace not seen in decades. New capacity comes online on the schedule of transformers, switchgear, cable and generators, and those orders show up in public records months before a campus is announced.
| 2014 | $0.2bn |
| 2014 | $0.2bn |
| 2014 | $0.2bn |
| 2014 | $0.2bn |
| 2014 | $0.2bn |
| 2014 | $0.2bn |
| 2014 | $0.2bn |
| 2014 | $0.2bn |
| 2014 | $0.2bn |
| 2014 | $0.2bn |
| 2014 | $0.2bn |
| 2014 | $0.2bn |
| 2015 | $0.2bn |
| 2015 | $0.2bn |
| 2015 | $0.2bn |
| 2015 | $0.2bn |
| 2015 | $0.4bn |
| 2015 | $0.4bn |
| 2015 | $0.3bn |
| 2015 | $0.3bn |
| 2015 | $0.3bn |
| 2015 | $0.3bn |
| 2015 | $0.3bn |
| 2015 | $0.3bn |
| 2016 | $0.3bn |
| 2016 | $0.4bn |
| 2016 | $0.4bn |
| 2016 | $0.4bn |
| 2016 | $0.4bn |
| 2016 | $0.5bn |
| 2016 | $0.5bn |
| 2016 | $0.5bn |
| 2016 | $0.5bn |
| 2016 | $0.4bn |
| 2016 | $0.4bn |
| 2016 | $0.4bn |
| 2017 | $0.4bn |
| 2017 | $0.4bn |
| 2017 | $0.4bn |
| 2017 | $0.4bn |
| 2017 | $0.5bn |
| 2017 | $0.5bn |
| 2017 | $0.5bn |
| 2017 | $0.5bn |
| 2017 | $0.5bn |
| 2017 | $0.5bn |
| 2017 | $0.5bn |
| 2017 | $0.5bn |
| 2018 | $0.5bn |
| 2018 | $0.6bn |
| 2018 | $0.6bn |
| 2018 | $0.6bn |
| 2018 | $0.7bn |
| 2018 | $0.7bn |
| 2018 | $0.8bn |
| 2018 | $0.7bn |
| 2018 | $0.6bn |
| 2018 | $0.7bn |
| 2018 | $0.6bn |
| 2018 | $0.7bn |
| 2019 | $0.7bn |
| 2019 | $0.7bn |
| 2019 | $0.7bn |
| 2019 | $0.7bn |
| 2019 | $0.8bn |
| 2019 | $0.9bn |
| 2019 | $0.8bn |
| 2019 | $0.8bn |
| 2019 | $0.8bn |
| 2019 | $0.8bn |
| 2019 | $0.8bn |
| 2019 | $0.7bn |
| 2020 | $0.8bn |
| 2020 | $0.8bn |
| 2020 | $0.8bn |
| 2020 | $0.8bn |
| 2020 | $0.8bn |
| 2020 | $0.8bn |
| 2020 | $0.8bn |
| 2020 | $0.8bn |
| 2020 | $0.9bn |
| 2020 | $0.8bn |
| 2020 | $0.8bn |
| 2020 | $0.8bn |
| 2021 | $0.8bn |
| 2021 | $0.7bn |
| 2021 | $0.9bn |
| 2021 | $0.8bn |
| 2021 | $0.8bn |
| 2021 | $0.8bn |
| 2021 | $0.9bn |
| 2021 | $0.9bn |
| 2021 | $1.0bn |
| 2021 | $0.7bn |
| 2021 | $0.8bn |
| 2021 | $0.8bn |
| 2022 | $0.8bn |
| 2022 | $0.7bn |
| 2022 | $0.8bn |
| 2022 | $0.8bn |
| 2022 | $0.8bn |
| 2022 | $0.9bn |
| 2022 | $0.9bn |
| 2022 | $0.9bn |
| 2022 | $1.0bn |
| 2022 | $0.8bn |
| 2022 | $0.8bn |
| 2022 | $0.9bn |
| 2023 | $0.9bn |
| 2023 | $1.0bn |
| 2023 | $1.1bn |
| 2023 | $1.1bn |
| 2023 | $1.2bn |
| 2023 | $1.2bn |
| 2023 | $1.3bn |
| 2023 | $1.4bn |
| 2023 | $1.4bn |
| 2023 | $1.5bn |
| 2023 | $1.5bn |
| 2023 | $1.5bn |
| 2024 | $1.6bn |
| 2024 | $1.6bn |
| 2024 | $1.8bn |
| 2024 | $1.8bn |
| 2024 | $2.1bn |
| 2024 | $2.3bn |
| 2024 | $2.4bn |
| 2024 | $2.4bn |
| 2024 | $2.5bn |
| 2024 | $2.5bn |
| 2024 | $2.5bn |
| 2024 | $2.7bn |
| 2025 | $2.7bn |
| 2025 | $2.7bn |
| 2025 | $2.8bn |
| 2025 | $3.3bn |
| 2025 | $3.1bn |
| 2025 | $3.0bn |
| 2025 | $3.1bn |
| 2025 | $3.2bn |
| 2025 | $3.3bn |
| 2025 | $3.4bn |
| 2025 | $3.2bn |
| 2025 | $3.3bn |
| 2026 | $3.3bn |
| 2026 | $3.3bn |
| 2026 | $3.5bn |
| 2026 | $3.8bn |
| 2026 | $4.1bn |
| 2026 | $4.6bn |
| 2026 | $4.9bn |
| 2026 | $5.4bn |
Source: US Census Bureau, via Our World in Data
| Period | Value |
|---|---|
| 2026 | 3.3 |
| 2027 | 12.3 |
| 2028 | 17.1 |
| 2029 | 14.5 |
| 2030 | 18.1 |
| 2031+ | 4.1 |
Source: US Energy Information Administration
Most data center datasets count announced megawatts. We measure the physical inputs instead: at the factory, where equipment is ordered and built, and at the site, where it is installed. Both sides share one equipment taxonomy, one set of company identifiers and one point-in-time clock.
Order books, factory utilization, capacity expansions and contract wins at the manufacturers that supply power equipment, cable and memory to US data centers.
backlog, new orders, utilization, capacity, capex, R&D, export mix, contracts
Site-level detail on 500+ US data center projects: how many generators, how large, whose engines, what fuel, how much power is made on site, and whether the project is on schedule.
site, operator, unit count, unit rating, engine make, fuel, on-site MW, status
Supply Chain Intelligence covers the equipment that decides when new data center capacity comes online.
Every backup generator, turbine, engine and fuel cell in US data center permit filings, with counts, ratings, makes, fuel and operating limits.
Order backlog, utilization, capacity and contracts at the makers of transformers, switchgear and cable, including the Korean suppliers that ship much of it to the US.
Facility investment, capacity and R&D at Korean memory makers, read from the same filings.
A site-level view of US data center projects and how their plans change: new filings, expansions, resizings, stalls and withdrawals.
We add sources continuously. If you are looking for something specific in the data center supply chain, ask us.
Press releases give a headline number. Permit filings and supplier disclosures give the parts list: how many units, of what size, from which manufacturer, running for how many hours a year. That is the level our data works at, for every site and supplier we cover.
| Headline | What the filing shows |
|---|---|
| "1 GW AI campus" | 54 diesel gensets at 2.5 MW each, Caterpillar, Tier 2, 100 hours a year |
| "Powered by natural gas" | 6 gas turbines at 57 MW each, permitted for prime duty |
| "Phase 2 underway" | Modification adds 24 units, filed eleven months before the announcement |
| "Record orders" | Transformer backlog up 6.1% q/q, North America 58% of new orders, plant at 96% utilization |
Example records.
Every number traces back to a document, every history can be rebuilt as of any date, and every company maps to something you can trade or benchmark.
Order flow and site demand across US and Korean equipment names, ahead of earnings.
Where demand is forming, and what equipment constraints mean for timelines.
Bottom-up, site-level data for market sizing and supply chain maps.
Share of permitted units by make and model, and competitor wins by region.
We send our research notes and analysis by email when they are published.
Tell us the names, regions or equipment you follow. We will walk through live records and send a sample built around them.