On September 8, The Wall Street Journal reported that the US government is becoming a part-owner of three publicly traded quantum computing companies.

The Commerce Department struck deals worth up to $100 million each for D-Wave Quantum (QBTS), Rigetti Computing (RGTI), and Quantinuum (QNT). The money comes from the CHIPS Act, the 2022 law best known for paying to build chip fabs in the US. In return, Washington gets a small, non-controlling equity stake in each company.

The stocks jumped on the news. That’s noise.

What I care about is the structure of the deals, because that tells you what Washington is really buying. (I’m leaving the “should the government own stock” debate for another day.)

Take Rigetti’s deal. The money doesn’t arrive in one lump sum. About $44 million is upfront. Two more chunks of about $30 million and $26 million only come if Commerce decides Rigetti hit specific technical milestones by certain dates.

In exchange, Rigetti agreed to issue the government about 7.7 million new shares (with some limits on voting rights), valued at $100 million.

Commerce gets all 7.7 million shares upfront, but it can only sell the slice that matches cash it has already sent Rigetti. And if it walks away early, Rigetti can buy back all the shares the government hasn’t paid for yet for one dollar. Not one dollar per share. One dollar total.

A research grant basically says, “Here’s money. Go do science.” This says, “Build something specific, in America, on a schedule—and the government keeps a sliver of the upside.” It’s essentially the same semiconductor playbook Washington ran last year when it turned Intel’s CHIPS grants into an ownership stake.

Importantly, the deals reported on September 8 sit inside a bigger push. In May, Commerce signed letters of intent with nine companies for about $2 billion, each with a minority, non-controlling government stake attached.

The biggest check went to Anderon, a new IBM subsidiary that’s a pure-play quantum wafer foundry headquartered in Albany, New York. On September 16, Anderon finalized a $1 billion CHIPS award, and IBM is adding $1 billion of its own.

Anderon makes quantum chips for customers across the industry on 300-millimeter wafers—the same dinner-plate-sized discs modern chip plants use. And the first quantum wafers are already moving through the plant.

A foundry is like a commercial bakery that will make anyone’s recipe. Most quantum companies still bake their own bread in a back kitchen.

Also on September 8, GlobalFoundries (GFS) finalized a deal with the US government worth up to $375 million over five years—tied to milestones—to make quantum chips and the super-cold electronics and packaging around them. Privately held PsiQuantum finalized an award of up to $100 million, too.

I should note that Google turned the government down in June because it thought the attached strings might slow its progress toward a working quantum computer. IonQ (IONQ) and Microsoft (MSFT) were also not among the nine companies I mentioned earlier.

The struck deals tell us Washington is *not *buying qubits. It’s buying factories, other infrastructure, and process R&D to build up the domestic quantum supply chain.

In other words, the government has stopped asking, “Who has the most qubits?” Now it’s asking, “Who can build the necessary hardware, repeatedly, in the US?”

If you need some background on quantum before we discuss more of the important things that have happened recently, you can find a dozen or so pieces I’ve written about the space under the “Frontier Tech” tab of our Grow or Die Substack site, including a four-part intro investor series.

Now, on with the show…

We don’t have a general-purpose, fault-tolerant quantum computer yet of course. But in the past year, the technologies needed to build these things at scale truly started coming together.

Error Correction

A raw qubit is sort of like a pencil balanced on its point. The slightest vibration knocks it over, and a useful quantum computer needs to be able to perform millions of steps without a fall. Error correction is the fix. You spread one piece of information across many shaky qubits and have them constantly check on each other. When one starts to fall, the group catches it and sets it upright before it does.

The catch has always been that every extra pencil is one more thing that can fall. For decades, adding more made things worse. Google changed the game with its Willow chip in December 2024. It showed you could add more qubits to make the system more reliable, not less, and that this repair method gets better as it scales. In other words, Google showed for the first time that scalable error correction is possible.

And error correction became a lot more concrete over the past year.

Google basically taught its Willow chip to tune and re-tune itself as it works. In July, it published a Nature paper on an AI system that adjusts more than 1,000 control settings on its chip while it runs. When researchers deliberately knocked the chip out of tune, it held performance 3.5X more steadily than conventional calibration.

IBM went after stamina. After unveiling the 120-qubit Nighthawk chip last November, an updated version can now run a calculation with over 7,500 two-qubit operations. And the company says Nighthawk-based systems could reach 15,000 operations by 2028.

That’s important because while qubit count tells you how big your workspace is, operation count tells you how many steps you can take before errors pile up and ruin the answer—and the problems we need quantum computers for, like designing new drugs and materials, require long chains of steps.

IBM also showed a “decoder,” the ordinary computer that reads the error clues, working in under half a millionth of a second. That matters because even the most brilliant error-fixing code is worthless if the decoder can’t keep up with the errors.

Continuous Control

Continuous control is necessary for practical error correction.

When you run a quantum calculation, you can’t press Start and walk away. While the quantum chip runs, a classical computer must monitor, control, and correct it in real time. Measure, decide, correct, measure again, about a million times a second—faster than the errors pile up.

Think of it like keeping a spinning plate (the quantum chip) balanced on a stick. The catch: you can’t look at the plate. One glance and it falls. All you can do is feel the “wobble” through the stick and adjust your hand. That’s continuous control. And without that low-latency classical computer feedback loop, error correction is not practical.

In late October 2025, continuous control became much more real when Nvidia introduced NVQLink, an open architecture for connecting quantum processors to GPU-accelerated supercomputers through CUDA-Q. It enables superfast back-and-forth communication so GPUs can perform calibration, control, and error correction while the quantum calculation is running.

The broader point here is that a useful quantum computer needs classical computing to handle certain things while it focuses on the “strange” part of the problem. That means useful quantum computers will almost certainly be hybrid systems with quantum processors for specialized operations and GPUs/CPUs (and other classical chips) for everything else.

“Stitching” quantum chips to classical supercomputers so they can talk to each other in millionths of a second takes you from a science demo to something you can put in a data center to do real-world work. It also means quantum computers will need much more ordinary computing than almost anyone realizes.

Manufacturing

Until this year, most companies building a quantum computer also had to build their own chips, a few at a time, in their own research fab. Like if Nvidia had to own a factory to make every chip it designed. This has been a major bottleneck in quantum since the beginning.

Anderon—the company that spun out of IBM and received a $1 billion check from the US government I mentioned earlier—is the first 300mm wafer pure-play foundry

built exclusively for quantum.

It will make quantum chips for anyone who wants them and its first wafers are going through the line now.

GlobalFoundries is building a similar business that will work with several kinds of quantum chips.

You can think of Anderon and GlobalFoundries as the TSMCs of quantum. Startups can now focus on design and hand off the manufacturing, which is how the regular chip industry got cheap, fast, and reliable.

Quantum Security

Another important development in the past year: quantum-safe cybersecurity moved from standards documents into live networks and federal deadlines.

See, pretty much everything you do online relies on math to scramble your data so outsiders can’t read it and prove that the website you’re visiting or software update you’re getting is the real thing. A powerful enough quantum computer running Shor’s algorithm could break the math (the locks) behind both. That machine doesn’t exist yet, but the risk already does because spies can simply steal and store scrambled data today and unlock it later.

Much of the replacement math (new locks) has been ready since 2024, when NIST finalized new quantum-resistant standards. The scrambling half is already widely used. Most human web traffic through Cloudflare (NET) uses it, and nobody notices. The proving half is just getting started.

This June, the standards became homework with a due date. The federal government’s most sensitive systems must upgrade their scrambling by the end of 2030 and their proving by the end of 2031. Federal contractors are on the same clock.

You should care as an investor because a deadline turns “someday” into a budget line. Every agency, contractor, bank, and hospital that follows Washington’s lead now has to find and replace old locks across millions of systems.

So we’re talking about an enormous upgrade cycle that will take years and doesn’t even need a working quantum computer to happen. It’s why quantum security gets its own list in my upcoming power ranking.

Next time, we’ll talk about two important quantum computing companies that went public since my last quantum stock power ranking in late May. Then I’ll give you a brand-new power ranking.

Related: The Growth-Stock Trap That Destroyed Zillow and Groupon