Latest / Quickly Quantum / Fortaegis Bets $50M That Security Lives in the Chip, Not the Code — September 14, 2026
Transcript
- Fifty million dollars, oversubscribed, just went into a Dutch startup betting that the silicon itself — not the math running on top of it — is where quantum-era security should live. Is chip-fingerprint hardware the shortcut past software-based post-quantum cryptography, or just a well-funded story? That's our main story today. Before that, in the headlines: Quantum Machines and NVIDIA get a live quantum program running across qubits and GPUs in about a millionth of a second, Quantinuum's Helix architecture entangles two different error-correction codes together, Europe's research network GÉANT teams up with the Quantum Internet Alliance, and RIKEN, Cleveland Clinic and IBM model a twelve-thousand-atom protein. Welcome back to Quickly Quantum, your daily brief on the quantum frontier. It's Monday, September 14, 2026. Let's get into it. Now, Quantum Machines says it just became the first control-hardware company to run a complete NVIDIA CUDA-Q program — that's NVIDIA's platform for writing one piece of code that a classical computer and a quantum computer both understand — across live qubits, GPUs and CPUs, using NVIDIA's NVQLink connector, a microsecond-speed link between the quantum controller and the GPUs. The company demonstrated it live this week at IEEE Quantum Week in Toronto. The headline number: the full exchange between the quantum processor and the classical chips completed in about a millionth of a second, fast enough to support real-time error correction, where you need an answer back before the qubits lose their information. Quantum Machines CTO Yonatan Cohen called it a step toward developers moving faster on large-scale quantum computers. Now, this is a press release, and the 'first' claim is self-reported — nobody's independently benchmarked it yet. But the practical thread here is real: this is the same push we've seen all year to make quantum hardware programmable without needing a specialist to hand-code control pulses. If it holds up outside NVIDIA's own demo room, that's real infrastructure, not just a slide. Quantinuum's back too. Five days ago, this show covered Helix, the company's error-correction code, hitting a logical memory error rate of 4.6 times ten-to-the-negative-five on its ninety-eight-qubit Helios processor — logical qubits being the error-corrected qubits built from many noisy physical ones. Today's delta comes from the full published paper: Quantinuum entangled two different types of error-correction code together, linking a Helix logical qubit to a separate surface-code logical qubit, into a three-logical-qubit GHZ state — a specific kind of shared entangled state — with a fidelity of 99.925 percent. Independent write-ups at Quantum Computing Report and Quantum News back up that number. Why bother mixing codes at all? Because Helix is efficient at logic gates but weak at preparing the special 'magic states' that universal computing needs, so Quantinuum built a second, specialized code just for that job and taught the two to talk to each other. The catch, per independent technical analysis: they proved the interface works, but they haven't yet pushed an actual magic state through it. Quantinuum says the payoff shows up on its next-generation Apollo hardware. Across the Atlantic — or rather across the fiber running under it — GÉANT, the operator connecting forty European research and education networks, signed a memorandum of understanding on September eighth with the Quantum Internet Alliance, a fifty-plus-institution consortium led by TU Delft. The deal, signed at GÉANT's Amsterdam headquarters by CEO Lise Fuhr and QIA director Stephanie Wehner, pairs GÉANT's fiber backbone with QIA's quantum-networking expertise: the physical nodes, memory buffers, and entanglement-distribution algorithms you need to actually move quantum information between cities. The goal is moving lab-scale quantum communication prototypes into real cross-border field trials, working toward what QIA calls a sovereign, European-built quantum internet. GÉANT's own newsroom and Quantum Computing Report both put out detailed accounts of this agreement within days of the signing, agreeing on the specifics. But a memorandum of understanding is still intent, not a funded build-out with a start date. Signing a partnership is easy. Actually running entanglement across a border on live fiber is the test that matters, and nobody's put a date on that yet. Now, over in Japan, RIKEN, Cleveland Clinic and IBM modeled a protein with twelve thousand six hundred thirty-five atoms — the largest molecule anyone's simulated using a quantum computer, according to the team. They call the approach quantum-centric supercomputing: tightly coupling IBM's Heron quantum processors with classical supercomputers, here Japan's Fugaku and Miyabi-G systems, so each does the part it's actually good at. The team reports a 210-fold improvement in accuracy within less than a year of scaling the method roughly forty times, and the work is a finalist for the 2026 ACM Gordon Bell Prize, high-performance computing's top award. The target is drug discovery — modeling how atoms bind more accurately to speed up finding new therapies. Here's the caveat worth sitting with: modeling a big molecule isn't the same as proving you beat classical computers doing it alone, and that comparison isn't established in what's been published so far. Scale is a real, countable number. Advantage is still a different question — and that same question, what hardware actually proves versus what it merely claims, is exactly what's driving our main story today. Our main story today: the fifty-million-dollar bet that security belongs in the silicon, not the software running on top of it. Fortaegis, an Amsterdam startup founded in October 2023, just closed an oversubscribed Series A — fifty million dollars, meaning more investors wanted in than the round had room for — led by Singapore's Serendipity Capital, with Tokyo Electron's venture arm, TEL Venture Capital, also putting money in, alongside investors tied to manufacturing, defense, quantum computing and critical infrastructure. So what does Fortaegis actually make? Every silicon chip comes off the manufacturing line with tiny physical imperfections that are effectively unique to that chip, like a fingerprint baked in at the factory. Fortaegis reads that fingerprint to generate encryption keys that change constantly and, crucially, are never permanently stored on the device — so there's nothing sitting on the chip for an attacker to steal, even if they get physical access to it. The company calls the result the Fortaegis Silicon Platform, meant to handle identity and security across networks of potentially millions of connected devices or autonomous AI agents, deployable through server hardware, ruggedized field equipment, or compact edge devices — chips that process information locally instead of shipping everything to a data center. The money is earmarked for commercial manufacturing starting in 2027, expanding production of field-programmable gate arrays, or FPGAs — chips you can reconfigure after they're built — and developing application-specific chips, or ASICs, built for one job and one job only, which tend to run faster and cheaper at volume. Fortaegis says it's already working with more than 25 companies and governments, and founder and CEO Boudewijn Wijnands told the Financial Times the company is working with governments in the US, UK, Japan, Singapore, Germany, France and the Netherlands. The company has filed 17 patents, with another 14 applications in the pipeline. Now here's the question this whole story turns on: is a hardware fingerprint that never gets stored actually a stronger hedge against quantum attacks than the standardized post-quantum cryptography that NIST has already published — or is it a different tool solving a different problem, dressed up as a replacement? Nobody outside the company has run that head-to-head comparison yet. Fortaegis isn't shy about the stakes it thinks it's playing for. 'The security and prosperity of the West and its allies increasingly depend on who controls the foundations of compute,' Wijnands said in the company's announcement. 'AI and quantum are making that question more urgent. Fortaegis was founded on the conviction that security cannot remain something we add on top of digital systems — it must become part of the foundations of compute itself.' Ilyas Khan, Fortaegis's chairman and founding shareholder — who's also the founder of quantum-computing company Quantinuum — went further, telling The Quantum Insider: 'The concept of using nature's intrinsically intractable security system that is carried within silicon has finally been turned into a reality. Fortaegis' technology will become foundational across multiple industries and multiple platforms... Make no mistake this is not simply another tech development. This is a breakthrough that will affect a great many industries.' And Chris Miller, a Fortaegis US board member and the author of the book Chip War, tied it to the bigger geopolitical picture: 'Computing infrastructure is the central arena of geopolitical and economic competition today. As AI advances and quantum computing takes off, security and control of computing foundations are essential. Fortaegis is addressing that challenge at the architectural level.' Three people with a direct financial and reputational stake in this company, all reaching for the same word — foundational. That's worth noticing, and worth being a little skeptical of. Here's what none of those quotes settle: whether hardware-fingerprint keys actually beat standardized post-quantum cryptography in a real deployment, at scale, under attack — that comparison hasn't been run independently. And the company's own speed claims, including a widely repeated figure that its communication is up to two hundred times faster than existing approaches, come from unspecified internal testing, not an outside audit. My honest read is that Fortaegis is solving a real, narrow problem — getting rid of permanently stored keys that attackers can physically extract from a device — and it's dressing that up in bigger language about being the quantum-era answer to encryption itself. Those are two different claims. The narrow one is plausible and useful for exactly the markets it's chasing: defense hardware, satellites, autonomous systems, places where physical tampering is the real threat model. The bigger one, that this becomes the security layer for AI and quantum broadly, is the part riding on 'oversubscribed' and boardroom quotes rather than a shipped product — commercial manufacturing doesn't even start until 2027. Time for the Hype Check: I'm putting this at a 4 out of 10 on substance. The underlying hardware idea is real and the funding is real, corroborated by three separate outlets — but the quantum-resistance framing and the two-hundred-times speed claim are company assertions nobody's independently tested, and there's no product shipping for another year at least. If Fortaegis actually gets independently benchmarked against NIST-standardized post-quantum cryptography before that 2027 manufacturing date, that comparison will tell us whether hardware-rooted keys are a genuine alternative or just a good pitch deck. If you want the daily version of this kind of story — the funding round, the quote, and the part that still needs proving — follow Quickly Quantum wherever you're listening right now. This has been Quickly Quantum, an AI-voiced podcast, created and built by a real human using today's cutting-edge technology. Nothing you heard on this show is financial advice. I'm Brian Lampert, and I'll catch you all tomorrow — take care! I also host Space Stakes: the business of the new space race, every day. What actually flew, what the contract is really worth, and who has customers. Find it wherever you get your podcasts.