Fortaegis Bets Security Belongs in the Silicon

Every security product sold in the last two decades assumed a human was somewhere in the loop. Fortaegis is betting that assumption just expired.

The pitch is unfashionably deep. Not a detection model, not an agent firewall, not another layer in the SOC stack — a claim that the root of trust itself has to move down into the physical properties of the chip, because the things authenticating to each other are no longer people or even servers, but millions of autonomous agents and edge devices transacting at machine speed.

On Saturday, Amsterdam-based Fortaegis Technologies said it had closed an oversubscribed $50 million Series A led by returning investor Serendipity Capital. The money moves the company from technology validation into commercial production of what it calls Secure Compute, across AI infrastructure, defense and autonomous systems, telecoms, advanced manufacturing, critical infrastructure and automotive.

What Fortaegis actually built

The product is the Fortaegis Silicon Platform, a full-stack architecture spanning hardware, firmware, cryptography and software. The load-bearing idea: instead of storing cryptographic keys on a device — where they can be extracted, cloned or coerced — the platform derives them from the unique physical characteristics of the silicon itself. Keys are generated rather than kept. There is no secret sitting on disk to steal.

From that root, the platform extends identity, security, policy and orchestration across every enabled node. Fortaegis says the design is quantum-safe by construction rather than by retrofit, which is a meaningfully different claim from "we added a post-quantum algorithm." If the key never exists in stored form, the migration problem that is currently consuming enterprise cryptography roadmaps looks different.

The Quantum Insider reports that commercial manufacturing is planned to begin in 2027, with FPGA production expanding in the meantime. The company reports having filed 17 patents with a further 14 in the pipeline, and says internal and customer testing has demonstrated connection performance more than 200 times faster than conventional approaches. That number is the company's own, from its own and customers' testing, and has not been independently verified — read it as a vendor claim, not a benchmark. But the direction matters: if you are authenticating millions of agents continuously, handshake latency stops being a footnote and becomes the constraint.

As The AI Insider notes in its account of the round, the approach turns the chip's own physical variation into the key material. The same architecture ships in four form factors — server rack-mounted, ruggedized edge, compact edge, and embedded silicon — which is how a security company sells to both a hyperscaler and a drone fleet without maintaining two products.

Follow the cap table

The investor list is the most informative document in this round, and it is not a normal software syndicate.

Serendipity Capital led. Strategic and supply-chain participation came from TEL Venture Capital — the corporate venture arm of Tokyo Electron, one of the handful of companies whose equipment is required to make advanced chips at all — alongside TNO, the Dutch applied-research organisation, and Prodrive Technologies, a Dutch high-end electronics manufacturer. NP-Hard Ventures, NovaCapital, Access Ventures and Coalition Capital also participated.

Then the individuals. Ilyas Khan, who founded Quantinuum, is Fortaegis chairman. Ian Fujiyama, who runs global aerospace, defense and government at Carlyle, is in. So is Peter Frankopan, the Oxford global historian. And Chris Miller — the author of Chip War — sits on the company's US board, describing computing infrastructure as "the central arena of geopolitical and economic competition today."

That is not a growth-equity roster. It is a manufacturing, defense-procurement and geopolitics roster. Read it as a statement about where the founders think demand comes from: governments and primes, not product-led growth. Serendipity's Rob Jesudason framed the thesis as building a new layer of infrastructure rather than improving an existing category — which is the honest description of both the opportunity and the risk.

Why now, and what's actually changed

Two forces are named in the announcement, and they pull in the same direction.

The first is autonomy. AI is collapsing the time between decision and action, and removing the human who used to be the implicit checkpoint. When an agent can provision infrastructure, call an API and move data without a person approving each step, identity and authorisation have to hold at machine speed and at machine scale. We covered the same pressure from the software side when HiddenLayer raised $100 million on the argument that agents are the new attack surface. Fortaegis is making the hardware-side version of that bet.

The second is quantum. The cryptography underwriting today's systems has a known expiry risk, and the migration is slow, expensive and incomplete. A root of trust that is quantum-safe by design rather than by patching is a genuinely different posture — assuming it works as described at production scale, which is precisely what this round is meant to prove.

There is a third force the release gestures at without naming: sovereignty. Control over compute has become a policy question, not just a procurement one. A European company selling a hardware-rooted trust layer into European and allied defense programs is a product decision and a positioning decision at once. Fortaegis says it is already working with more than 25 enterprises and governments, with engagements moving from customer-funded engineering into production qualification — the stage where deep-tech companies either convert or stall.

Who this moves

For AI infrastructure operators, this is a preview of an argument they will hear repeatedly: that securing an agent fleet with software identity alone is a category error. Whether that is true at the price point Fortaegis will need to charge is an open question. Hardware-rooted security has historically been sold on compliance mandates rather than on architecture.

For incumbents, the threat is slow but real. The security vendors have spent two years bolting AI onto detection and response — a pattern visible in everything from Nvidia's move into cybersecurity to Palo Alto's help-desk acquisition. None of that touches the root of trust. If the root moves into silicon, the layer above it gets commoditised.

For defense buyers, the appeal is obvious and the timeline is the problem. Resilience Media frames the target deployments as spanning military networks, drone fleets and NATO systems. Production qualification in defense runs in years. The round funds FPGA production across the US, Europe, Singapore and Japan plus an ASIC roadmap, which is the right sequence — FPGAs to win programs, ASICs to make the margins work — but ASICs are expensive and slow, and $50 million does not go far against a custom-silicon roadmap.

For investors, this is the deep-tech trade in its purest form. A validated architecture, a marquee cap table, real government engagement, and essentially all of the value dependent on execution in manufacturing — the part of the stack where capital intensity punishes optimism.

The takeaway

Three lessons sit inside this round. First, when the entities authenticating to each other stop being people, the assumptions underneath your security model expire quietly — nothing breaks, the model just stops describing reality. Second, a cap table is a demand forecast; when a security company raises from a chip-equipment maker, a national research institute and a defense-focused private-equity executive, it is telling you which buyer it believes in. Third, "quantum-safe by design" and "quantum-safe by upgrade" are different products with different costs, and the market has not yet priced the difference.

Fortaegis's founder argues that security cannot remain something added on top, but "must become part of the foundations of compute itself." That is either the most important architectural argument in security right now or an expensive way to rediscover why the trusted platform module never became a business. The root of trust is the one layer you cannot patch your way into. The next two years of manufacturing will decide which sentence applies.

Frequently Asked Questions

What is Fortaegis Secure Compute?

Secure Compute is Fortaegis's full-stack security architecture, delivered as the Fortaegis Silicon Platform. It spans hardware, firmware, cryptography and software, and roots trust in the physical properties of silicon rather than in stored cryptographic keys. It then extends identity, policy and orchestration across connected nodes, and is designed to be quantum-safe.

How much did Fortaegis raise and who led the round?

Fortaegis announced an oversubscribed $50 million Series A on September 13, 2026, led by returning investor Serendipity Capital. Participants included TEL Venture Capital, the corporate venture arm of Tokyo Electron, plus TNO, Prodrive Technologies, NP-Hard Ventures, NovaCapital, Access Ventures and Coalition Capital.

Why does AI make hardware-rooted security more relevant?

Autonomous agents act without a human approving each step, so identity and authorisation must hold at machine speed across very large device and agent populations. Fortaegis argues that software-only identity and stored keys are insufficient at that scale, and that deriving keys from silicon properties removes the stored secret an attacker would target.

What will Fortaegis do with the funding?

The company says the capital will scale production and commercial deployment of its FPGA products across the United States, Europe, Singapore and Japan, advance its ASIC roadmap, and grow engineering and commercial capacity. Founded in October 2023, it reports working with more than 25 enterprises and governments.

Editor's note — sources: Fortaegis Technologies Series A press release (primary), The Quantum Insider, Techzine, Resilience Media, The AI Insider. The 200x connection-performance figure and the 17/14 patent counts are company-reported and not independently verified. Strategic analysis is Edgewisely's own.