Energy · Deep dive

Mainspring Energy

Menlo Park maker of the linear generator — a flameless, fuel-flexible onsite power machine that reacts natural gas, hydrogen or ammonia at low temperature to shuttle magnets through copper coils — selling firm, fast-to-deploy power to data centers, utilities and industrial sites while gas turbines are sold out through 2030.

emerging

The question that decides it: Mainspring wins the spec sheet — fuel-flexible, near-zero NOx without aftertreatment, containerized units deliverable in months while GE Vernova's turbine slots are sold out through 2030 — but fifteen years in it has only tens of megawatts operating in the field, its Coraopolis factory does not reach volume until 2027, and Bloom Energy booked roughly $20B of backlog and single gigawatt-scale data-center orders in 2025-26 selling into the same time-to-power panic. Can Mainspring convert a 250kW-container product into repeatable 50-100MW data-center and utility blocks off the Pennsylvania line — with multi-year durability data its air-bearing architecture has never produced at fleet scale — before the 2026-2028 grid-shortage window closes and turbine lead times normalize, or does the linear generator remain the technically elegant niche machine while fuel cells and diesel-era incumbents take the AI power buildout?

My take ★★ Very interesting ★ Interesting − Maybe ✗ Not interesting

Snapshot

Mainspring Energy builds linear generators: containerized machines that convert natural gas, biogas, hydrogen or ammonia into electricity through a flameless, low-temperature reaction — no crankshaft, no turbine blades, two moving parts riding on air bearings. Founded in 2010 by three Stanford engineering PhDs as EtaGen, the Menlo Park company has raised more than $800M (company figure, April 2025), capped by a $258M Series F led by General Catalyst with Amazon’s Climate Pledge Fund and Temasek aboard, and is building a 292,000-square-foot factory outside Pittsburgh backed by an $87M Department of Energy grant. It matters now because AI data centers cannot get grid power — turbine slots are sold out through 2030 — and Mainspring sells firm onsite capacity in months. The uncomfortable counterpoint: after fifteen years it has deployed tens of megawatts, while Bloom Energy, selling into the same panic, booked a roughly $20B backlog.

Founding story

Shannon Miller went to Stanford for all three of her mechanical-engineering degrees and stayed for the thermodynamics. Her PhD research, funded by Stanford’s Global Climate and Energy Project, modeled what happens when fuel-air mixtures are compressed to extremes: react them at low temperature, without a flame, and you can extract more work per unit of fuel than any practical engine while producing almost no NOx. Matt Svrcek, a Stanford Graduate Fellow in the same lab, ran the companion experiments on reacting varied fuels under high compression. With fellow engineer Adam Simpson they incorporated EtaGen in 2010 — “eta” being the Greek letter engineers use for efficiency — to build the machine the dissertations implied. Khosla Ventures wrote the first check when, in its own later words, the linear generator “was just a theory”; Bill Gates joined early. What followed was a decade of hardware slog: the company did not launch a commercial product until January 2021, when it rebranded as Mainspring and unveiled the Linear Generator alongside a $150M financing agreement with NextEra Energy Resources. In June 2026 Miller executed the rarest founder move — initiating her own succession — handing the CEO seat to board chair Tom Linebarger, who ran Cummins for a decade, and becoming President over operations. The signal is explicit: the physics problem is solved; the manufacturing-scale problem is now the company.

How it works

A linear generator deletes the rotating machinery from power generation. Inside a Mainspring unit, two opposed pistons — the company calls them oscillators — sit in a shared cylinder. A cycle begins when energy stored in air springs at the outer ends drives both oscillators inward, compressing a fuel-air mixture between them. Compression continues until the mixture reacts uniformly at low temperature — no spark, no flame, no combustion in the conventional sense. The reaction shoves the oscillators apart; magnets mounted on each sweep through copper coils in the shell, inducing current directly. At the end of travel each oscillator compresses the air spring behind it, which rebounds it inward to start the next cycle, generating power in both directions. The oscillators never touch metal — air bearings flanking the coils float them on a cushion of air — so there is no oil, no lubrication circuit, and no scheduled core parts replacement. Two moving parts total. Because reaction temperature stays low, NOx formation is near zero without aftertreatment, and the company claims roughly 45% electrical efficiency (a California Energy Commission-supported figure), with software-controlled compression ratios letting the same hardware switch between natural gas, biogas, hydrogen and ammonia. Thunder Said Energy’s patent review flagged the honest caveats: low-temperature combustion faces a thermodynamic efficiency ceiling, and long-run durability of the piston-seal and air-bearing assemblies is the technology’s genuinely open engineering question — there is no fleet with decades of hours to point to.

Product and business overview

The commercial unit is a 250kW linear generator packaged in a standard 8-by-20-foot container, deliberately shaped like shipping infrastructure rather than a power plant. Mainspring sells it in factory-integrated blocks — a standardized 3MW skid combines twelve units with inverters and an optional step-up transformer — and scales by parallelization from a few hundred kilowatts to, the company says, 100MW-plus; observers peg the economic sweet spot at 1-10MW. Three named solution lines: commercial-and-industrial onsite power (grocery, cold storage, EV charging — Kroger and Lineage are the flagship logos), utility-scale distributed capacity (AEP pilot, a 48MW Utah Municipal Power Agency project announced February 2026), and, since roughly 2024, data centers, where the pitch is time-to-power: modular systems installed and commissioned in months while grid interconnection queues run years. Distribution runs through channel partners — Schneider Electric embeds the units in its microgrids (Mar 2024 partnership), plus regional resellers — and through financiers, principally NextEra.

Business model and pricing

Mainspring publishes no price list, and no $/kW figure has surfaced in press coverage — Revenue arrives two ways: direct equipment sales, and — more commonly for anchor customers — PPA-style structures where NextEra Energy Resources buys and finances the units (its January 2021 commitment was $150M) and the customer pays a recurring fee for power with minimal upfront commitment; both PG&E and Kroger chose that route. The value pitch: Mainspring’s Kroger case study claims up to 30% savings versus grid power at a Los Angeles-area store (2021); for data centers, any onsite premium beats revenue lost to a 24-72-month interconnection delay. The DOE grant ($87M, October 2024) and Pennsylvania’s $8.6M subsidize the capacity ramp rather than the product. What is not visible anywhere: margins, backlog, or whether units sold to date were profitable to build at Menlo Park volumes.

Traction over time

The shape of the curve is the story: fourteen years to first tens of megawatts, then an order book (48MW in one Utah project alone) that exceeds the historical installed base — all contingent on a factory that does not produce until 2027. Revenue has never been disclosed; third-party estimates range implausibly from ~$11M to ~$136M (2025-26).

Market analysis

The driver is the data-center power crunch. US data-center capacity is projected to roughly double from ~80GW in 2025 to ~150GW by 2028 (SemiAnalysis), grid interconnection queues hold 2,600GW of stalled projects, and gas-turbine lead times have blown out to three-to-five years, with GE Vernova’s ~100GW backlog (April 2026) effectively sold out through 2030 and turbines ordered today running around 2031 (Utility Dive, 2026). That converts onsite generation from niche to default: a Bloom Energy-commissioned operator survey found 27% of data-center facilities expect to be fully onsite-powered by 2030, up from 1% a year earlier. The addressable spend is variously sized: MarketsandMarkets puts data-center power at $35B in 2025 growing to $50.5B by 2030 (7.5% CAGR). Mainspring’s older markets — C&I resiliency, utility load pockets, biogas — add steadier demand. The structural risk runs the other way too: this is a shortage-priced market, and every competitor from turbine makers to fuel cells is adding capacity into the same window.

Competitive intel

Bloom Energy is the measuring stick and the humbling comparison: ~$28B market cap (October 2025) after a 1,000% run, ~$20B backlog at end-2025, a $2.65B twenty-year AEP offtake for up to 1GW (January 2026) and up to 2.8GW with Oracle — gigawatt-denominated deals against Mainspring’s megawatt-denominated ones. Bloom’s fuel cells prefer steady baseload; Mainspring counters with fast-ramping dispatchability and multi-fuel switching — a utility argument more than a hyperscaler one. Caterpillar and Cummins own the default alternative — gas recips with global dealer networks, lower capex, and durability records Mainspring cannot yet match; hiring Cummins’ ex-CEO is Mainspring’s answer to that credibility gap. INNIO’s Jenbacher business, which filed for a 2026 IPO on data-center demand, is the closest 1-10MW product rival with strong CHP economics. Generac crowds the commercial resiliency low end. GE Vernova and Siemens Energy define the ceiling: at 100MW+ scale turbines remain cheapest per kW, so Mainspring’s premium survives only while slots are scarce or air-permit limits (where its near-zero NOx genuinely differentiates) bind. FuelCell Energy’s July 2026 data-center order shows even also-rans are getting fed in this market.

History and evolution

• 2010 — Founded as EtaGen by Miller, Svrcek and Simpson out of Stanford’s thermodynamics lab; Khosla first money in.
• Jan 2018 — $83M Series C; strategics AEP, Centrica, Statoil join. Still no commercial product eight years in.
• Jan-Mar 2021 — Rebrands as Mainspring; launches the Linear Generator with the $150M NextEra agreement; Kroger disclosed as customer. Eleven years founding-to-launch.
• May 2021 — $95M Series D (Devonshire); Chevron joins.
• Aug 2021 — PG&E pilots a unit in a Napa County mobile microgrid — first utility client.
• May-Sep 2022 — $290M Series E (Lightrock); AEP Oklahoma pilot MOU (Jul 2022).
• Mar 2024 — Schneider Electric microgrid partnership at CERAWeek.
• Sep 2024 — Lineage signs for 33 units across five Texas facilities.
• Oct 2024 — $87M DOE grant and $8.6M PA package for the Coraopolis plant; ~600 jobs, 2027 production.
• Apr 2025 — $258M Series F led by General Catalyst; Linebarger (ex-Cummins) and Bethany Mayer join the board; total raised tops $800M.
• Feb-Jun 2026 — UMPA 48MW award; Travis Air Force Base multi-fuel pilot; Linebarger appointed CEO, Miller President (June 10).

What people say

The case for. Customers keep re-ordering, which is the strongest public evidence: Kroger went from one LA-area store (with claimed ~30% grid-cost savings, 2021) to a multi-site relationship; Lineage expanded from early deployments to a five-facility, 33-unit Texas commitment (September 2024). Utilities that piloted it — PG&E, AEP — kept engaging, and UMPA trusted it with 48MW of greenfield municipal capacity (February 2026). The investor list is unusually strategic — Shell, Chevron, Equinor, AEP, Hanwha, Amazon — energy operators who diligenced the physics. Glassdoor runs 4.1/5 across 66 reviews (2026), praising collaborative engineering culture and mission. The Air Force chose it to test hydrogen and ammonia operation (March 2026), third-party validation of the fuel-flex claim.

The complaints. Glassdoor’s rating fell about 10% over the past year; recurring gripes are below-market pay, 50-plus-hour weeks, 24/7 on-call expectations, and an older review’s cut of “cool tech, with a lack of vision”. Thunder Said Energy’s patent analysis questions long-run longevity and maintenance of the air-bearing/seal architecture and notes low-temperature combustion’s thermodynamic efficiency ceiling; The business-side critique writes itself: no disclosed revenue, no disclosed valuation, no disclosed backlog after $800M raised; deployment measured in tens of megawatts in a market rivals measure in gigawatts; and a factory-dependent 2027 ramp that must execute flawlessly inside a demand window competitors are already harvesting. The DOE grant, awarded October 2024, also ties a material chunk of the expansion to federal disbursement schedules the company does not control.

Outlook: the open question

Mainspring works if the Coraopolis factory turns a validated 250kW machine into repeatable 50-100MW blocks by 2027-28 — while the turbine shortage still holds the door open — and if early fleets accumulate durability hours that convert conservative buyers; it fails if the ramp slips or the window closes first. The bull case is genuinely strong: the technology is differentiated on physics rather than marketing (dispatchable like an engine, clean like a fuel cell, fuel-flexible like neither), permitting-constrained air basins structurally favor near-zero-NOx generation, the strategic cap table and NextEra financing solve the buy-side friction, and putting Cummins’ former CEO in charge is exactly the move a company makes when its constraint is manufacturing throughput, not invention. UMPA’s 48MW suggests the order book is finally denominated in real capacity.

What would sink it: time. Bloom demonstrated in 2025-26 that data-center buyers reward whoever can ship gigawatts now; every quarter Mainspring’s volume line slips, hyperscaler standardization hardens around fuel cells, Jenbacher recips and bridging turbines. If GE Vernova’s post-2030 slot availability arrives before Mainspring’s fleet-scale durability proof, the company is left competing on $/kW against machines with fifty-year track records. The tells to watch: whether Coraopolis ships production units on schedule in 2027; whether Mainspring announces a first named hyperscale or colocation deployment above 50MW (none exists publicly as of July 2026); whether a Series G arrives as growth capital or as a rescue; and whether the Air Force and UMPA projects publish performance data. Fifteen years of patient capital bought a working machine and a moment when the world desperately wants what it does. The next 24 months decide whether that coincidence becomes a company.

How a challenger would attack it

Attack the gap between the spec sheet and the fleet. Mainspring’s vulnerability isn’t the physics — it’s fifteen years for tens of megawatts, a factory that doesn’t produce until 2027, and zero multi-year durability data on the air-bearing architecture that Thunder Said Energy’s patent review flags as the open engineering question. A challenger doesn’t need a better machine; it needs to ship during the 2026-2028 window Mainspring can’t fully serve. Bloom is already running the play at gigawatt scale, but the sharper version targets Mainspring’s specific niches: an INNIO-style recip packager with a certified low-NOx aftertreatment stack neutralizes the permitting advantage in constrained air basins at lower $/kW with fifty years of durability data behind it, and Caterpillar or Cummins can bundle dealer-network service guarantees no 500-person Menlo Park company can match — the exact credibility gap Mainspring conceded by hiring Cummins’ ex-CEO. The commercial attack is disclosure: Mainspring publishes no price, no backlog, no revenue (estimates span $11M to $136M), so a rival that posts transparent $/kW, delivery dates and fleet-hours wins every conservative utility procurement. And the talent flank is open — Glassdoor down 10% in a year on below-market pay and 24/7 on-call — precisely when a challenger staffing a rival power-dense product would be recruiting from the same small pool of linear-machine and power-electronics engineers.

Same playbook, new buyer

Mainspring’s playbook — modular, containerized, fuel-flexible firm power sold through financiers like NextEra so the customer pays a PPA fee, not capex — ports to buyers the data-center gold rush is making it neglect. The most defensible shift is the one its own history proves: C&I resiliency for grocery, cold storage and EV charging (Kroger’s claimed 30% savings, Lineage’s 33-unit re-order) is a repeatable, credit-worthy segment that Generac serves with dirtier machines and Bloom ignores as too small — but Mainspring’s every incremental 2027 factory slot will be bid away by larger data-center and utility blocks, orphaning the beachhead that got it here. Second, geography: near-zero NOx without aftertreatment and ammonia capability are worth the most where air rules are strictest and gas infrastructure weakest — Japan and Korea (Hanwha and Marunouchi are already on the cap table), island grids, and ports facing ammonia-bunkering mandates — markets a Coraopolis-constrained company can’t chase before 2028. Third, defense and remote microgrids: the Travis AFB multi-fuel pilot shows the buyer exists, and a challenger productizing rapid-deploy, fuel-agnostic expeditionary power would face no Mainspring response while its factory is mortgaged to utility-scale blocks.

Sources and further reading

• Mainspring Secures $258 Million in Financing to Scale Linear Generator Business (PR Newswire, Apr 2025)
• Mainspring Closes Series E at $290 Million (PR Newswire, Sep 2022)
• Mainspring Announces $95 Million Series D Financing (PR Newswire, May 2021)
• EtaGen Closes $83 Million Series C Round (PR Newswire, Jan 2018)
• Mainspring Energy Launches New Power Generation Technology and $150 Million NextEra Agreement (PR Newswire, Mar 2021)
• PG&E is first utility client for Mainspring’s novel ‘linear generator’ (Trellis/GreenBiz, 2021)
• U.S. DOE Awards $87 Million Grant to Mainspring for Pennsylvania Manufacturing Facility (PR Newswire, Oct 2024)
• Mainspring Appoints Former Cummins Chairman and CEO Tom Linebarger as CEO (PR Newswire, Jun 2026)
• Utah Municipal Power Agency Selects Mainspring for 48 MW Local Power Generation Project (PR Newswire, Feb 2026)
• Mainspring Energy: linear generator breakthrough? (Thunder Said Energy patent review)
• High-efficiency and Ultra-low Emissions Linear Generator (California Energy Commission, 2024)
• 5-year waits and rising costs: How demand is redefining the gas turbine market (Utility Dive, 2026)
• Onsite Generation Expected to Fully Power 27% of Data Center Facilities by 2030 (Bloom Energy survey, 2026)