Washington Pays $389M for Chip R&D — and Demands Equity Back

By Daniel Rao | October 24, 2026

The most consequential sentence in the Commerce Department’s announcement on Thursday wasn’t about the money. It was about the strings. After the four grant amounts, the usual bureaucratic boilerplate and the congratulatory quotes, one clause turns an ordinary R&D handout into a political statement: as a condition of the funding, the United States government will take minority, non-controlling, non-voting equity stakes in all four companies receiving it.

Washington isn’t just writing cheques any more. It is keeping a piece of the cap table.

We read the Commerce Department’s announcement as first reported by Reuters on October 22, and cross-checked every figure below against three independent relays of the department’s statement. The numbers are straightforward; the implications are anything but.

Here is what the government said on Thursday: it has signed letters of intent to award up to $389 million in research and development funding to four companies working on power semiconductor devices and critical materials. Commerce Secretary Howard Lutnick said in a statement the awards will “accelerate American semiconductor innovation to enhance our domestic energy and power resources.” The department has awarded 19 R&D grants to semiconductor firms in recent months, all drawing from a $10 billion fund approved by Congress in 2022 under the CHIPS and Science Act.

Read the announcement twice, and the second paragraph is where the story lives. Not the $389 million — spare change next to the sums being thrown at AI data centres this year — but the equity condition attached to every dollar of it.

Photo: The Tokamak Fusion Test Reactor at Princeton Plasma Physics Laboratory, New Jersey (file photo). Helion and Commonwealth Fusion Systems, recipients of two of the four awards, are building the next generation of fusion power technology. Credit: Princeton Plasma Physics Laboratory, CC BY 3.0, via Wikimedia Commons.

The Fine Print Is the Story

Let’s be precise about what “minority, non-controlling, non-voting equity stakes” actually means, because the three adjectives are doing very different work.

Minority means the government isn’t buying control — these are small slices, not nationalisation. Non-controlling means no board seat, no say in how the companies run. Non-voting means the shares carry no vote on company matters at all. In economic terms, the taxpayer gets to ride the upside if these companies grow; in governance terms, the taxpayer gets nothing. It is profit participation without power — the state as silent partner.

That framing matters, because this is not a one-off experiment. Seoul Economic Daily, tracking the announcement for Korean readers, called it plainly: the arrangement “goes beyond a simple grant and marks an expansion of direct government investment in advanced industries” under the Trump administration. The stated goal, per the department, is to reduce reliance on overseas production and strengthen domestic supply chains for advanced semiconductors and critical materials — standard industrial-policy language, but now with a balance-sheet mechanism attached.

Two caveats worth printing in full. First: these are letters of intent, not final awards. Seoul Economic Daily notes the actual funding amounts and detailed conditions will be finalised during negotiations on the final agreements. “Up to $389 million” is a ceiling, and the equity terms could still move before anything is signed. Second: the four-way split of the money tells its own story about where Washington thinks the bottlenecks are. Fusion power and AI data centres — two industries that did not share a supply chain five years ago — are now fighting over the same components.

Update, Oct. 24: Helion's $50M private match and the robotics footnote

Update, Oct. 24: The fine print got two fresh footnotes in the relays. Helion said its award will be accompanied by $50 million in new private investment funds — federal money pulling private capital behind it, which is exactly the leverage story Washington wants to tell (Channel News Asia, via Reuters, Oct. 22). Commonwealth Fusion Systems added a new detail on where its $85 million goes: the funding will help develop semiconductor technologies for robotic maintenance systems, which Commonwealth CEO Bob Mumgaard called 'a vital part of our global effort to deploy commercial fusion.' And the market framing keeps landing on the equity clause: TradingView's takeaway called it 'unusual' — minority, non-controlling, non-voting stakes that give taxpayers economic exposure without operational control, potentially upside if these technologies go commercial. The awards are letters of intent, not final grants, so the equity terms can still move in negotiation. Meanwhile Washington is buying the components while grids strain under the load — the same convergence story behind the UK's Ofgem warning over 315 data centres.

Helion’s $175 Million: Fusion’s Power Problem, Not Its Physics Problem

The largest slice goes to Helion, the Washington-state fusion energy developer: up to $175 million to develop high-voltage power semiconductors and advanced energy power modules for fusion power generation. The semiconductors will power and use energy from the magnets driving Helion’s fusion generators. Helion said the award will be accompanied by $50 million in new private investment — new money from private investors on top of the federal funding.

That distinction matters more than the headline number. Fusion’s hardest near-term problem isn’t always the plasma physics — it’s the power electronics that have to precisely regulate enormous bursts of electrical energy through the magnets that confine the reaction. Get the switching wrong and you don’t get a star in a bottle; you get very expensive scrap.

“Building fusion power plants at scale requires us to accelerate the development of technologies like semiconductors that are critical to making fusion possible,” said Helion chief executive David Kirtley.

The domestic-supply angle is the quietest but sharpest part of Helion’s award. According to The Inference, Helion has said the high-voltage semiconductors its machines require are currently available only from overseas manufacturers — which is precisely the dependency this award is designed to attack. An American fusion programme that depends on foreign power electronics isn’t an American programme at all, whatever the flag on the building.

Commonwealth Fusion’s $85 Million: The Massachusetts Counterbet

The second fusion recipient is Commonwealth Fusion Systems, the Massachusetts company that will receive up to $85 million. Its remit is broader: semiconductors for radio-frequency plasma heating, power electronics, fusion diagnostics, and radiation-hardened electronics. The funding will also support semiconductor technologies for robotic maintenance systems — because a reactor that runs hot enough to melt anything it touches needs to be serviced by machines, not people.

“These innovations will be a vital part of our global effort to deploy commercial fusion,” said Commonwealth Fusion Systems chief executive Bob Mumgaard.

Notice the deliberate pairing. Two fusion bets, two different technology stacks: Helion’s pulsed magnet approach and Commonwealth’s steady-state tokamak design. Washington isn’t picking a fusion winner here. It is funding the component layer underneath both of them — the power semiconductors, the diagnostics, the radiation-hardened control electronics — and letting the reactor physics sort itself out. It is, in procurement terms, a supply-chain play wearing an energy-policy costume.

The $73 Million Data-Centre Play: Why Vertical GaN Matters

Vertical Semiconductor, a spin-off from the Massachusetts Institute of Technology, will receive up to $73 million to advance its vertical gallium nitride (GaN) power semiconductors for use in AI data-centre racks.

If you don’t live in the power-electronics weeds, here’s the one-paragraph version: GaN devices lose less energy as heat when they convert and switch power, which means they can handle more power in less space. In an AI data centre — where US data centres are projected to burn more natural gas than Germany and Japan combined by 2035 — the cost of moving power from the grid to the GPU is one of the industry’s fastest-growing bills. Vertical Semiconductor is pursuing technology that Seoul Economic Daily reports can be applied to 2.5-megawatt AI data-centre server racks, with The Inference describing the goal as a demonstration of those racks.

This is the award with the shortest path to revenue of the four. Fusion power plants are years from commercial operation; AI data-centre racks are being built by the thousand right now, and their operators are desperate for anything that shaves percentage points off power-conversion losses. The equity stake Commerce is taking in Vertical Semiconductor is arguably the shrewdest of the lot — if vertical GaN takes off in rack power stages, the taxpayer’s silent slice could turn out to be the cheapest trade Washington made all year.

The wider context is the industry’s frantic search for efficiency: Google, NVIDIA and Emerald AI launched an alliance to turn AI data centres into flexible grid assets just last month, and every percentage point of conversion efficiency now has a dollar value attached to it.

Photo: The surface of a 12-inch silicon wafer, with individual dies visible in iridescent detail. Power semiconductors for AI racks and fusion plants are built on exactly this kind of substrate. Credit: 2x910, CC BY-SA 4.0, via Wikimedia Commons.

Sibelco’s $56 Million: The Quartz Bottleneck Nobody Talks About

The least glamorous award may be the most strategically important. Sibelco — the US unit of a Belgian minerals company — will receive up to $56 million to develop next-generation purification for ultra-high-purity quartz used in the semiconductor supply chain. Seoul Economic Daily reports the company plans to build a demonstration facility in North Carolina to test the new refining technology.

Ultra-high-purity quartz is one of those materials you never hear about until it’s gone. It goes into the crucibles and tubes that hold molten silicon during wafer manufacturing, and the contamination tolerances are brutal: trace impurities measured in parts per billion. The refining technology is concentrated in a handful of suppliers globally, which makes it a single point of failure for the entire domestic wafer industry — including the memory-chip manufacturing that SK Hynix and Intel are in talks to bring to the US.

It is, frankly, the kind of award that doesn’t photograph well and doesn’t trend on social media — which is exactly why governments fund it and markets don’t. Nobody holds a press conference for quartz.

Our Verdict: Small Money, Aimed at Exactly the Right Layer

Let’s not overstate what $389 million buys in 2026. A single hyperscale data-centre campus can cost ten times that. The four companies are getting research funding, not production subsidies, and letters of intent are promises to negotiate — not cheques. Anyone presenting this as a moonshot-sized bet hasn’t looked at the scale of the industries it touches.

But the targeting is shrewder than the sum suggests, and so is the mechanism. Washington has stopped pretending that grants and contracts are the only tools in the industrial-policy drawer. A non-voting equity stake does two things a grant can’t: it gives the taxpayer a claim on the upside when public money creates private value, and it quietly changes the politics of the next funding round, because killing a programme the government owns a piece of is harder than cancelling one it merely funds. That is the expansion Seoul Economic Daily identified — and it won’t stop at four companies.

The honest risks are the obvious ones. Letters of intent can shrink in final negotiation. Equity pricing for private companies like Helion will be argued over by lawyers for months. And a future administration can, and will, reassess every one of these stakes — political durability is the one thing no contract can guarantee. But as industrial policy goes, this round is refreshingly honest about its logic: pay for the shared bottleneck, take a silent slice so taxpayers ride the upside, and let the physics and the markets sort out the rest. The switch between the grid and the load is where fusion and AI both live or die. Washington just bought into it.

Frequently Asked Questions

Are these awards final, or still provisional?

Provisional. The Commerce Department has signed letters of intent — not final funding agreements. Seoul Economic Daily reports the actual amounts and detailed conditions will be finalised in negotiations before anything is binding, so treat “up to $389 million” as a ceiling.

What does a “minority, non-controlling, non-voting” stake actually mean?

Minority: the government owns a small slice, not a controlling position. Non-controlling: no board seat and no say in management decisions. Non-voting: the shares carry no vote on company matters. The taxpayer gets economic exposure to the companies’ growth — but zero governance power. It’s profit participation, not nationalisation.

Why is the government taking equity instead of just giving grants?

Commerce’s logic, implicit in the announcement, is taxpayer return: if public R&D money creates private value, the public should share in the upside. The political effect is that a programme the government owns a piece of is harder for a future administration to kill than one it merely funded. Seoul Economic Daily framed it as an expansion of direct government investment in advanced industries under the Trump administration.

What are power semiconductors, and why do fusion and AI need the same ones?

Power semiconductors are chips designed to switch and convert large amounts of electrical energy efficiently — the opposite of the logic chips in your phone. A fusion generator needs them to regulate enormous power bursts through confinement magnets; an AI data-centre rack needs them to convert grid power to GPU power with minimal loss. Same component layer, two customers, one bottleneck — which is why Commerce funded both.

Where does the $389 million come from?

A $10 billion R&D fund approved by Congress in 2022 as part of the CHIPS and Science Act. The Commerce Department has awarded 19 R&D grants to semiconductor firms from that fund in recent months; these four letters of intent are the latest — effectively the twentieth round, per The Inference’s tally.

Which company gets the most, and what for?

Helion (Washington state) leads at up to $175 million for high-voltage power semiconductors and energy power modules for fusion generation, plus $50 million in new private investment alongside. Commonwealth Fusion Systems (Massachusetts) gets up to $85 million for plasma-heating, diagnostic and radiation-hardened electronics. Vertical Semiconductor (an MIT spin-off) gets up to $73 million for vertical GaN power chips for AI data-centre racks. Sibelco gets up to $56 million for ultra-high-purity quartz purification.

Sources: Reuters (first report, David Shepardson, Oct 22, 2026); U.S. Commerce Department announcement via Reuters; Seoul Economic Daily (Oct 23, 2026); The Standard / Hong Kong (Oct 23, 2026); SemiEngineering Chip Industry Week in Review (Oct 2026); The Inference (Oct 23, 2026); geomechanics.io engineering explainer (Oct 23, 2026).

By Daniel Rao — Daniel covers semiconductors, AI infrastructure and the industrial policy behind them for TechPulse USA. He previously wrote about chip supply chains and the data-centre power crunch.