Energy / Grid infrastructure · Deep dive

Heron Power

Solid-state transformers for the AI-era grid — a Tesla powertrain veteran betting that silicon carbide power electronics can replace the century-old iron-core transformer that data centers and renewables are now waiting years to get.

emerging

The question that decides it: Heron sells a semiconductor-based transformer into a market whose buyers — utilities and data-center operators — value nothing above multi-decade field reliability. The bet is that Heron Link hits cost parity with iron-core gear AND proves 20-year field life fast enough to win orders while the conventional-transformer backlog is still 2-4 years long. Does Heron ship enough real-world unit-hours to earn certification and field trust before the iron-core shortage eases and the urgency that is driving its 50GW of soft orders evaporates — or does it become the next grid-power-electronics startup that was technically right and commercially early?

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

Snapshot

Heron Power is a Scotts Valley, California startup building solid-state transformers — power-conversion hardware that swaps the copper-and-iron transformer for silicon-carbide semiconductors — under the brand Heron Link. Founded in 2024 by Drew Baglino, the former Tesla Senior VP of Powertrain and Energy Engineering, it has raised ~$178M across a $38M Series A (May 2025) and a $140M Series B (February 2026) co-led by a16z’s American Dynamism fund and Breakthrough Energy Ventures. The pitch lands amid an acute grid-transformer shortage: as of Q2 2025, Wood Mackenzie put the US power-transformer supply deficit at 30% with lead times of two-to-four years. Heron claims 50GW of soft orders from a dozen-plus prospective customers — including Intersect Power and Crusoe — and plans a 40GW-per-year US factory. As of mid-2026 it is pre-revenue.

Founding story

Baglino is the story. He joined Tesla in 2006, before Musk was CEO, designed the powertrain for the first Model S, and rose to run powertrain and energy engineering. He resigned in April 2024, in the same week Tesla laid off around 10% of its workforce, and by 2025 had surfaced with Heron Power.

His framing of why is the useful part. At Tesla he built hardware for the “edge” of the system — cars and batteries. His conclusion on leaving: the binding constraint has moved. Batteries are now cheap and fast to deploy; the wires, substations, and transformers that connect new load and generation are not. Transformers in particular have barely changed since the 1880s, have been commoditized and offshored, and are now backlogged for years. Public sourcing names him as founder and CEO; co-founders are undisclosed, though he has recruited heavily from Tesla’s power-electronics ranks. The Tesla halo is literal in the cap table — co-founder JB Straubel and ex-CFO Zach Kirkhorn both invested in the Series A.

How it works

A legacy transformer is passive electromagnetics — AC through copper windings around an iron core, the turns ratio setting whether voltage steps up or down; cheap, efficient, and dumb, with no monitoring or control, one-way flow, one job per device. A solid-state transformer throws that out. Heron Link uses silicon-carbide (and gallium-based) semiconductors to convert power digitally at high frequency in three stages: a rectifier (AC to DC), a converter (changes the DC voltage), and an inverter (back to AC, or left as DC where the load wants DC). Because the conversion is switched by semiconductors rather than fixed by winding ratios, the same box can regulate voltage, correct power factor, decouple faults, handle bidirectional flow, and stream telemetry.

The physical product is modular. Each Heron Link handles ~5 megawatts across tens of power-conversion modules; a failed module swaps out in ~10 minutes, versus the multi-year wait to replace a monolithic transformer. It converts medium-voltage grid power (~34.5 kV) down to the ~600-800V DC that Nvidia’s latest rack designs want, and each unit packs lithium-ion cells that discharge for ~30 seconds to bridge to backup power — letting a data center delete its UPS entirely. The claimed payoff: strip up to 70% of the gear from a data-center power train, occupy 70% less space, and in some cases cut cost by an order of magnitude — while riding the EV-scaled SiC/gallium supply chain rather than the constrained electrical steel that chokes iron-core production.

Product and business overview

Heron Link is a family of medium-voltage solid-state transformers sold as capital equipment. Heron is deliberately not starting with utilities: initial buyers are solar farms, grid-scale batteries, and data-center campuses — customers who need conversion anyway, move fast, and feel the backlog most acutely.

As of February 2026, data centers were ~one-third of Heron’s pipeline, the balance solar and grid-scale batteries. Baglino’s sequencing is explicit: prove the tech with fast-moving customers, then “go back to utilities” for the larger AC-to-AC distribution grid.

Business model and pricing

Heron is a capital-equipment manufacturer, not a subscription business: it books revenue by selling hardware, and the model hinges on high-volume automated manufacturing at competitive cost. The $140M Series B is explicitly a factory raise — money to build a US plant capable of 40GW of annual output, with pilot production targeted for early 2027 and full-scale ramp through 2027-2028.

On price, Baglino’s public claim is aggressive: Heron is “not asking for any premium” over the equipment customers buy today, and at the system level (counting eliminated gear) can be cheaper. That claim is the whole ballgame — and runs against a decade of SST economics in which solid-state units cost several times their iron-core equivalents. Heron’s counter is that it wins on total-system cost where an SST replaces four or five devices, not on a like-for-like swap. No unit prices are public; revenue is zero until production begins.

Traction over time

The shape is a well-funded, pre-commercial hardware company; the signal is demand-side. Interest reportedly jumped past 40GW, prompting Baglino to raise the Series B barely nine months after the Series A even though, he says, Heron “didn’t need the money.” But “soft orders” and “technical collaborations” are not backlog: none are binding, none have shipped, no revenue is recognized. The two named customers — Intersect Power (which Google agreed to acquire for $4.75B) and Crusoe (a 1.2GW campus in Abilene, Texas) — are flagships, both still prospective.

Market analysis

The tailwind is real and unusually sharp. Wood Mackenzie’s Q2 2025 data put the US power-transformer supply deficit at ~30% (10% for distribution), with standard power-transformer lead times of 128 weeks, generator step-up units 144 weeks, and high-capacity units quoted as far out as four years. Prices since 2019 are up ~77% for power transformers and 78-95% for distribution. Generator step-up demand grew 274% from 2019 to 2025, driven by AI data centers, electrification, and reshoring. The US grid hosts up to 80 million transformers, over half 35-plus years old, and the National Laboratory of the Rockies (formerly NREL) expects power through them to double by 2050.

The SST market itself is still tiny: one estimate pegs the data-center SST segment at ~$40.3M in 2025, growing to ~$572.4M by 2034 — a real curve, but a fraction of the tens of billions spent yearly on conventional transformers. Bull case: SSTs expand share as costs fall. Bear case: they stay a niche premium product while iron-core soaks up the shortage.

Competitive intel

The set spans incumbents and startups. Hitachi Energy, GE Vernova, Siemens Energy and Prolec GE are the legacy transformer OEMs — full order books, pricing power, deep utility trust, and every incentive to keep selling profitable iron-core units rather than cannibalize them. Hitachi already builds HVDC power electronics and could field its own SST if the category matures. Eaton signaled the incumbents’ playbook by acquiring SST startup Resilient Power Systems in 2025 for up to $150M — buy, don’t get disrupted. On the startup flank, DG Matrix (Raleigh; $60M Series A in Feb 2026, backed by ABB and Mitsubishi Heavy) ships a commercially available multi-port SST and is arguably further along on productization, while Amperesand (Singapore, also Tesla alumni, ~$80M raised) chases the same data-center buyers. And the quiet competitor is inertia: the cheap, proven iron-core transformer that most buyers will still choose wherever they can tolerate the wait.

Where Heron wins: Baglino’s proven power-electronics manufacturing, the largest fresh war chest among pure-play SST startups, and a system-cost story strongest exactly where demand is hottest. Where exposed: it is the least commercially proven serious player, with no shipped product against incumbents who own certifications and relationships.

History and evolution

• Apr 2024 — Baglino resigns from Tesla after ~18 years, the same week Tesla cuts ~10% of staff.
• 2024 — Founds Heron Power in Scotts Valley, CA; raises a small early round (~$5M implied).
• May 2025 — Closes $38M Series A led by Capricorn’s Technology Impact Fund (with Breakthrough Energy, Energy Impact Partners, Gigascale, Powerhouse, Valor Equity, JB Straubel and Zach Kirkhorn); total raised ~$43M. Uses funds to finish engineering Heron Link.
• Feb 2026 — Closes $140M Series B co-led by a16z American Dynamism and Breakthrough Energy Ventures; discloses 50GW of soft orders, names Intersect Power and Crusoe, and announces plans for a 40GW US factory. Peers DG Matrix ($60M) and Amperesand ($80M) raise in the same window.
• 2026-2027 (planned) — Full-system field deployments in late 2026; pilot production early 2027; full-scale production H2 2027; multi-year ramp thereafter.

No public crises yet — the company is too young. The relevant history is the sector’s: a decade of grid-power-electronics startups (GridCo Systems, Varentec/Sentient Energy) that were technically credible but commercially stranded.

What people say

The case for. The bull thesis is a rare alignment of team, timing, and money. Investors — a16z, Breakthrough Energy, Capricorn, and Tesla’s own alumni — are betting on Baglino’s manufacturing track record, the scarce thing in a field of good physics and bad production. TechCrunch and Canary Media frame a “supercycle”: aging transformers, a hard supply deficit, and hyperscaler demand that will pay for anything shipping power faster. The system-level logic is compelling in data centers, where one Heron Link deletes a UPS, an inverter skid, and multiple transformers while the SiC supply chain dodges the steel bottleneck.

The complaints. Here an investor should slow down; the criticism is engineering and adoption skepticism, not customer reviews (there are no customers yet). SSTs have been “almost ready” for over a decade, and trade press (POWER Magazine, Canary Media) is blunt that prior efforts “struggled to gain traction given high costs and technical challenges” and that grid-power-electronics firms have repeatedly “closed up shop.” The specific doubts: (1) Cost — SSTs have historically run several times iron-core prices at distribution scale, and Heron’s “no premium” claim is unproven. (2) Reliability and field life — buyers underwrite 20-40 year lifetimes, and power semiconductors and their thermal management have no such record at medium voltage, with more failure modes than a passive core. (3) Certification and inertia — utilities are famously slow, and type-testing, protection compatibility, and interoperability standards take years, which is why Baglino is starting with data centers.(4) Efficiency — high-frequency switching adds losses iron-core designs lack. (5) Execution — a 40GW factory from a pre-revenue company is a Tesla-sized bet, and today’s urgency is partly a shortage that new iron-core capacity is racing to close.

Outlook: the open question

Heron works if Heron Link reaches genuine cost parity with the gear it replaces AND accumulates enough real-world unit-hours to earn certification and multi-decade field trust before the iron-core backlog eases — and it stalls if it remains a technically elegant, commercially early device that buyers admire but keep deferring in favor of the cheap, proven, backlogged transformer. The team and timing are as good as this thesis gets: a manufacturing-proven Tesla operator, the sector’s biggest balance sheet, an acute shortage, and hyperscaler buyers with urgency. The demand signal — 50GW of soft interest, two flagship names — is real.

But everything that matters is still in the future tense: no shipped product, no recognized revenue, no field-reliability data, no independent certification; a “no premium” claim untested at volume; and a 40GW factory bet that has sunk better-funded hardware firms. What would settle it for Heron: late-2026 field deployments that actually run, a bankable order that converts to 2027 revenue, third-party validation of cost parity and efficiency, and — the real prize — a utility willing to certify and buy. What confirms the bear case: slipping timelines, customers quietly renewing iron-core orders as lead times shorten, or the familiar SST pattern of a great router-for-power the grid never plugs in. This is an early-breakout bet on a founder and a supercycle — and the next 18 months of shipping, not fundraising, decide it.

How a challenger would attack it

Ship before Heron does, and make the pipeline pay for it. Heron’s most exploitable weakness is sequencing: ~$178M raised, zero shipped product, prototypes in the lab, pilot production not until early 2027 — while 50GW of “soft orders” sit unbound, unbanked, and free for anyone to convert. DG Matrix is already running this attack with a commercially available multi-port SST and ABB/Mitsubishi backing; a challenger presses it by signing binding, penalty-backed delivery contracts with the same hyperscaler and data-center buyers Heron courts, turning Heron’s flagship prospects (Intersect Power, Crusoe) into references for someone else. The second vector is the field-life gap: buyers underwrite 20-40 year reliability, and Heron has no unit-hours; a rival that instruments early deployments, publishes failure-rate data, and offers a full-replacement warranty or transformation-as-a-service model (Heron sells capex hardware only) transfers exactly the risk Heron asks customers to swallow. Third, attack the factory bet — Heron’s Series B is committed to a 40GW plant before any commercial validation; a challenger that contract-manufactures on existing SiC lines stays capital-flexible if the iron-core backlog eases and the urgency premium evaporates, which is the scenario that strands Heron’s fixed capacity. Eaton’s Resilient acquisition shows the endgame: bundle SSTs into a full electrical package with a service network Heron won’t have for a decade.

Same playbook, new buyer

Take semiconductor power conversion to buyers Heron has explicitly deferred. Baglino’s sequencing — data centers and renewables first, utilities “later” — leaves the utility distribution grid, with up to 80 million installed transformers over half of which are 35+ years old, as a stated afterthought; a startup that starts with utility certification, type-testing and protection-compatibility work now owns the slow-but-enormous market by the time Heron circles back. The second shift is scale-down: Heron Link is a ~5MW medium-voltage campus device, but the same SiC conversion logic applies to EV fleet depots, commercial-building electrification and industrial microgrids — sub-megawatt buyers facing the same 128-week lead times with nobody building for them. Third, geography: the transformer shortage is global, but Heron’s factory, supply chain and pipeline are entirely US; Europe’s renewable interconnection queues and Asia’s data-center buildouts have no domestic SST champion, and Amperesand’s Singapore base shows the seat is contestable. Heron won’t follow any of these soon — its capital is sunk into one 40GW US plant aimed at one buyer profile, and its own roadmap says utilities wait until the data-center beachhead is proven.

Sources and further reading

• Former Tesla exec Drew Baglino’s new startup is rethinking the electrical transformer (TechCrunch, April 2025)
• Tesla alum’s Heron Power closes $38M Series A to transform key grid technology (TechCrunch, May 2025)
• The Solid-State Shift: Reinventing the Transformer for Modern Grids (POWER Magazine, 2025)
• Heron Power raises $140M to ramp production of grid-altering tech (TechCrunch, February 2026)
• Heron Power raises $140M to modernize electrical transformers (Canary Media, February 2026)
• Why investors are going gaga over solid-state transformers (TechCrunch, February 2026)
• Power and distribution transformers will face supply deficits of 30% and 10% in 2025 (Wood Mackenzie, August 2025)
• US transformer market faces severe supply constraints as lead times extend to four years (pv magazine USA, May 2026)