Young startup Fabric Cryptography, based in Santa Clara, Calif., is building a new processor devoted to general purpose, programmable cryptography acceleration.
The team of 35 veteran supercomputing engineers is creating what they have dubbed a VPU, for Verifiable Processing Unit, that is set to go to tape out and production next year using an advanced process node with a Taiwan-based manufacturer, said Michael Gao, founder and CEO, in an interview with Fierce Electronics.
Gao said the company has designed a programmable processor to support different cryptography schemes including zero-knowledge proof, homomorphic, functional encryption and more. ZK proof is a cryptographic protocol that enables one person (the prover) to convince another (the verifier) that a particular claim is true without disclosing any details about the claim itself.
Already, Fabric Cryptography has secured signed purchase orders worth tens of millions of dollars, Gao said. In mid-summer, Fortune reported Fabric had one purchase order for $50 million.
Early end customers are in the blockchain space, but future customers could be banks, medical establishments or hyperscalers. The company is currently operating on $7 million in equity funding. One principle investor is Estonian billionaire Jaan Tallinn. Others reportedly include Nil Foundation, Inflection, Novawulf and Jasper Lau or Era.
“There’s a lot of investment in what we call cryptography compute,” Gao said. “Fabric Cryptography is basically building a GPU for cryptography.” The problem with today’s approaches is they run too slow. “Ours is orders of magnitude faster,” running in the megahertz range, above today’s kilohertz range.
One sign of that growth is hiring in the field. Fabric currently lists 22 openings, from general counsel to vp of technical operations and software engineers in cryptography, compilers, digital design and more.
Gao offered several examples of how the VPU would be used. In the case of facial recognition software, it could be used to verify known suspects in a crowd without the need to store faces, for example. To defend against a claim of racial bias in facial recognition, a facial recognition software company could create a ZK proof to show the bias for black faces was the same as for white faces without revealing the weights used to create the facial recognition model. “You don’t have to reveal the model to prove it’s not racist,” he said.
In the case of medical research, private patient encrypted data could be sent for assessments and diagnoses to a third- party researcher who could use the VPU to run a model so only the researcher could see the results.
With finance, an investor might want proof that a hedge fund is really making 20% on its investments. The hedge fund could cryptographically produce proof without revealing sensitive information.
“A ZK proof can be generated for any program, and you run the program to tally trading returns,” Gao said. “You produce a ZK-proof on top of our hardware and, from a hyperscaler or us, you can give proof to anybody and they can verify it.”
Gao said no other company is building anything similar to Fabric’s VPU, although RISC Zero is building something in the same arena: a platform for letting developers easily generate ZK proofs.
Gao, 27, first bought Bitcoin as a teen. He dropped out from MIT after a year and started Luminous Computing and raised $115 million in venture capital, according to his LinkedIn profile and other reports.