The Korean Startup Trying to Shrink the Grid's Heaviest Box

The transformer sitting on your street corner is one of the oldest ideas still running the modern world. Its basic design - iron core, copper coils, a bath of oil - has barely changed since the 1880s. It is heavy, it wastes a little energy every second, and almost nobody thinks about it. Crocus does. The company, founded in 2016 with a lab in Seoul and an office in Redwood City, has spent nearly a decade arguing that this unglamorous box is one of the most valuable things on the grid to reinvent.

Crocus is a power-electronics and energy-software company. That description sounds narrow, but the product line is wide. It sells EV fast chargers, an AI platform that reads and optimizes a facility's electricity use, and - the piece that now defines its ambition - a solid-state transformer that swaps iron and oil for semiconductors and code. The three products look unrelated on a shelf. Under the hood, they are the same problem: converting and controlling power efficiently, in real time, with software in the loop.

That is the throughline worth stealing if you build hardware. Crocus did not launch by announcing it would rebuild the grid. It started with EV chargers, a market with clear buyers and visible demand, and used the same conversion technology to climb toward harder, higher-value problems. Chargers first. Then the software that watches the meter. Then the transformer itself.

The ladderOne core, three markets

The company organizes everything under the name Acelo. Acelo EV is the charging business - fast and standard chargers with an operations system and a user app. Its distinguishing trick is "smart charging," which actively distributes available power across sessions instead of dumbly maxing out each plug. The chargers speak the industry's standards, including OCPP 1.6 and ISO 15118, and support vehicle-to-grid, meaning a parked car can push power back when the grid needs it. That last capability is often marketed louder than it ships; Crocus built it into the product.

Acelo Grid is the software layer. It analyzes a site's consumption patterns, predicts voltage, and adjusts equipment to trim waste. The headline claim is modest and specific: an average five percent reduction in power consumption through voltage optimization. Five percent sounds like a rounding error until you apply it to a factory, a logistics hub, or a data center - and then to every energy-intensive site on a grid.

Acelo SST is where the company gets interesting. A solid-state transformer replaces the passive iron block with an active, semiconductor-based unit that can be controlled by software. Crocus builds its versions on silicon-carbide (SiC) MOSFETs and high-voltage IGBTs, using power-conversion topologies with names only an electrical engineer loves - MMC, DAB, CHB, AFE. The point of all that machinery is control: the company says its units respond in under a millisecond and can hold voltage to within a single volt.

The transformer on your street corner hasn't fundamentally changed since the 1880s. Crocus thinks that's the point - and the opportunity.The Crocus thesis, in one line

Why nowThe AI data center problem

Timing helps. AI data centers are straining local grids, and the bottleneck is not only compute - it is power delivery. The Acelo SST DC800 is built for exactly this. It converts medium-voltage AC directly to 800 volts of DC, the level modern high-density racks increasingly want, skipping the wasteful conversion steps that traditional infrastructure stacks up. Crocus puts numbers on the payoff: 40 to 50 percent less floor space for power infrastructure, roughly 30 percent lower total cost of ownership, and up to 70 percent lower maintenance compared with conventional gear.

Those figures are the company's own, and belong in the "promising, not yet independently audited" column. But the direction is credible enough that the market has moved with it. Solid-state transformers have become one of the hotter corners of climate hardware, with venture money flowing to peers building the same category. Crocus's counterpoint is time: it has been working on this since 2016, while much of the field arrived recently.

ValidationA government put its name on it

In 2025, Korea's Ministry of Trade, Industry and Energy selected Crocus to lead a national project developing an AI autonomous-control-based semiconductor transformer system to optimize power quality in data centers. Being handed the lead role on a government program is not the same as commercial traction, but it is a specific, checkable signal that the technology cleared a serious technical bar. The same year, the company earned a TI-2 investment-grade technology rating - a Korean evaluation that sits near the top of the scale.

Both the EV charging and power control markets are growing amid worldwide energy imbalance. We highly evaluated Crocus' technical expertise, which is at the highest level in Korea.Yeom Seung-won, E&Investment

The money followed the same arc. Early rounds drew institutional backers including Samsung Venture Investment, Shinhan Asset Management and E&Investment. A Series B in March 2023 pushed cumulative funding past roughly $12 million. By late 2025, Crocus reported a Series C of about $22.5 million, bringing the total to near $31 million. CEO and co-founder Daniel Lim has said the company intends to pursue a KOSDAQ listing in 2027.

The businessSell the box, keep the software

Crocus is a B2B company, and its model is the classic hardware-plus-software pairing done in energy. It sells equipment - chargers and transformers - and pairs that equipment with control and monitoring software, operations platforms and apps. The hardware wins the site; the software keeps the relationship. A transformer or charger is a one-time sale, but a customer optimizing power quality with Acelo Grid keeps paying to make that power smarter. Government and utility R&D programs add a third revenue stream and a steady stream of validation.

The customer list points at anywhere electricity is expensive and abundant: EV charging operators and fleet depots, industrial complexes and factories, distributed-generation sites tying together solar and energy storage, and now AI data centers. The primary market is South Korea, with the Redwood City office giving the company a foothold in the U.S. and a line into Silicon Valley's power-hungry customers.

The competitionIron, incumbents and a crowded new field

Crocus competes on two fronts. In solid-state transformers, it lines up against a new wave of well-funded startups - names like DG Matrix, Amperesand and Heron Power - and, more fundamentally, against the century of installed iron transformers made by incumbents such as ABB and Hitachi Energy. Replacing a proven, cheap, dumb-but-reliable box with an expensive, clever, unproven one is a hard sell, and the industry moves slowly for good reasons. In EV charging and grid integration, the field is broader still, from ChargePoint and Wallbox to grid-software specialists.

So where does an eight-person company fit against that? Its edge is not scale; it is the combination. Few competitors do EV charging, grid AI and solid-state transformers from a single power-electronics core. That breadth is a genuine advantage if the pieces reinforce each other - and a risk if a tiny team spreads itself across three demanding markets at once. Which of those it turns out to be is the open question hanging over the next two years.

The honest partWhat could break this

Copyable lessons first: pick a beachhead market with real buyers, build on a technology core that ports to bigger problems, and let unglamorous efficiency gains compound into a product line. The five-percent insight - that small, boring savings become a business when multiplied across enough sites - is the kind of thing worth borrowing.

Now the conditions under which it does not work. Solid-state transformers must beat iron not in a lab but on cost, lifespan and reliability at grid scale, and that case is not yet closed. A team of eight chasing three markets can be a strength or a fracture. Much of the performance data is self-reported. And a 2027 listing target is a plan, not a fact. Crocus has built a coherent argument and earned real signals - a government mandate, a top technology rating, patient capital. Whether that argument becomes an industry standard is the wager still being placed.