David Archer, CTO, Niobium Microsystems – Interview Series

Dr. Archer is the CTO of Niobium Microsystems and a Principal Scientist leading Cryptography and Multiparty Computation at Galois, Inc., whose customers include DARPA, the NSA, IARPA, and the Department of Homeland Security.

He has more than 30 years of R&D experience in complex ASICs, system hardware, software architectures, secure computation, and cryptography.

Dr. Archer holds a Ph.D. in Computer Science from Portland State University, an M.S. in Electrical Engineering, and a B.S. in Computer Engineering from the University of Illinois at Urbana-Champaign.

Niobium Microsystems pioneers Zero Trust Computing solutions, featuring the industry’s fastest homomorphic encryption (FHE) hardware accelerators. The company’s technologies enable organizations to harness the value of data while keeping it fully encrypted, ensuring sensitive information remains private and supporting regulatory compliance. By pushing FHE performance to unprecedented levels, Niobium Microsystems is making possible a new generation of machine learning and statistical analysis applications with mathematically guaranteed privacy.

You’ve had an incredible academic and professional journey. Was there a specific moment or influence that led you to dedicate your career to cryptography?

I’d say it’s two disagreeing, specific influences, both worth reading: The Right to Privacy by Warren and Brandeis and The Transparent Society by Brin. The former emphasizes the right to be let alone, arguing that law should protect people from invasive technologies. The latter argues the opposite: those invasive technologies can’t be avoided, and we fail when we try to blind the powerful few who use them, because law prevents us from seeing them. Increasingly, both views seem impractical in the face of cyber criminal industry, data aggregators, and nation-state abuses. Modern cryptography that for example can prove facts without revealing secrets, or share sensitive data without revealing it, have important parts to play in a workable solution.

For those unfamiliar, how would you explain Fully Homomorphic Encryption and what makes it such a game-changer for data privacy?

Fully Homomorphic Encryption, or FHE, is a cryptographic technology that allows you to perform computations on encrypted data without ever decrypting it. Imagine handing a locked box to someone—they can manipulate what's inside the box without ever opening it, and then only you can unlock the result. With FHE, data remains encrypted during storage, transit, and computation. That’s revolutionary: it mathematically guarantees that sensitive data remains opaque. No more leaks, no more exposure.

What are the biggest challenges in making FHE practical for enterprise AI applications, and how is Niobium addressing those challenges?

Historically, the biggest challenge of FHE has been computational efficiency—FHE operations have been thousands to millions of times slower than regular computation. But at Niobium, we've tackled this head-on by building purpose-built FHE hardware accelerators. Unlike traditional CPUs or hardware adapted from graphics or AI architectures, our design is ground-up optimized for FHE's computational needs.  Through a full-stack, hardware-software co-design approach, we’ve brought FHE performance into the realm of practical enterprise use, finally aligning security with the speed of business.

What types of real-world use cases—particularly in defense or finance—do you think are most ready to benefit from FHE today?

We're already seeing clear demand in defense and finance, where protecting data isn't just desirable—it's mission-critical. An important warfighting capability is  multi-spectral imagery analysis from autonomous aircraft (UAV), which may be used to detect specific chemical signatures in a landscape. However, the particular signatures of interest are often sensitive: an adversary capturing a UAV could extract those signatures and learn how to adapt to them, avoiding detection. Onboard FHE can assure that those signatures remain secure even in the event of aircraft capture.

In finance, FHE is a transformative solution. It enables private queries in dark pool trading, safeguarding trading strategies and ensuring market fairness. Moreover, institutions can securely collaborate on fraud detection across borders without revealing sensitive transaction data or customer identities. This collaborative approach significantly enhances detection capabilities, uncovering complex institutional fraud schemes, cross-border transaction anomalies, and patterns of illicit activities like money laundering, all while maintaining strict compliance with international data protection regulations. The ability to share insights without sharing raw data isn't merely beneficial—it's essential for a secure and compliant global financial ecosystem.

There’s been a lot of talk about “zero-math” or privacy-preserving computing. How do you define that, and what role does it play in Niobium’s technology stack?

“Zero-math” privacy-preserving computing refers to making complex cryptographic operations transparent to developers and users—they don't need a Ph.D. in cryptography to leverage the technology. That’s a laudable goal, but the truth is that FHE implementation does require specialized knowledge. That’s why Niobium doesn’t stop at hardware: we’re also building software stacks to seamlessly integrate FHE into existing workflows. We’re also founding members of FHETCH, an open-standards consortium working to build cross-platform APIs so that FHE hardware and software can interoperate without customization. These are critical steps on the path to a world where users can benefit from this next-generation technology without needing to become cryptography experts themselves.

Homomorphic encryption has a reputation for being computationally expensive. Where do you see the most promising breakthroughs in performance coming from?

The most promising breakthroughs are happening in hardware acceleration tailored specifically for FHE. Traditional CPUs and GPUs just aren't built for the intricate, high-precision arithmetic FHE demands. At Niobium, we’ve specifically engineered our designs to handle these computations with unprecedented efficiency. Coupled with our specialized compiler optimizations, we're seeing orders-of-magnitude performance improvements that were unthinkable just a few years ago.

Some security approaches use trusted execution environments or confidential computing. How does FHE compare to those, and when is it the clearly better option?

Trusted Execution Environments (TEEs), sometimes called confidential computing, rely on hardware enclaves to protect data, but they still require trust in the hardware vendor, the enclave, and the system software, and they still process the data unencrypted. The complexity of that “root of trust” reduces TEEs to “pretty good security”, and a large number of studies confirm that concern. FHE removes trust entirely from the equation—no data ever needs to be decrypted, so there’s literally no sensitive data to compromise. Whenever absolute, mathematically guaranteed security is required, especially in heavily regulated industries, FHE is the superior choice.

Side-channel attacks are a concern when working with sensitive data at the hardware level. What steps have you taken to minimize those risks in Niobium’s designs?

Side-channel attacks exploit unintended information leaks from hardware. Research demonstrates over and over that TEEs are vulnerable to this kind of attack. However, if the only information you operate on is encrypted, then leaking it doesn’t benefit the adversary. It’s not that our hardware minimizes risk of leakage. Instead, it’s that our hardware maximizes performance and efficiency, particularly because it doesn’t need to worry about leakage.

How do you see standardization efforts in the FHE space evolving, and what is Niobium doing to ensure compatibility with emerging norms?

Standardization is crucial for adoption, and we're actively participating in these efforts through collaborations with industry groups and standards bodies. Niobium recently co-founded the FHE Technical Consortium for Hardware (FHETCH), alongside Chain Reaction and Optalysys, precisely to drive interoperability and establish practical performance benchmarks for commercially viable FHE solutions. As the Technical Co-Chair of FHETCH, I'm personally involved in shaping initiatives such as developing an API abstraction layer, which enables seamless integration between diverse FHE hardware accelerators and software libraries. We’re also involved in FHE community standardization efforts like those by homomorphicencryption.org. These proactive efforts help ensure compatibility and interoperability from the outset, which is key for enterprises looking to future-proof their security investments. We encourage NIST to standardize FHE quickly, because as you know, NIST cryptographic standards set the stage for federal government technology acquisitions, effectively making FIPS validation a prerequisite for selling any cryptographic products to any U.S. government agency.

What excites you most about where the field is heading in the next few years, and what can we expect from Niobium on the roadmap ahead?

What excites me most is the shift from theoretical promise to practical implementation of secure computing at enterprise scale. In the coming years, I anticipate that FHE cloud services will become increasingly commonplace, transforming the way organizations approach data security. Niobium will continue to lead this charge—we're not just advancing our FHE accelerator's performance, but also expanding into broader Zero Trust Computing solutions like zero knowledge proofs and cryptographically verifiable computation. This combination of assuring confidentiality on one hand, while assuring integrity and non-repudiation on the other, speaks directly to why I’m in this effort: resolving the conundrum laid out by Brandeis and Brin. Expect to see even more powerful, versatile, and accessible privacy-preserving technologies from us, redefining what enterprises consider possible in secure data handling.

Thank you for the great interview, readers who wish to learn more should visit Niobium Microsystems.