How Fabric intends to make advanced cryptography widespread.

August 21, 2024
Brian

Fabric Cryptography, a hardware startup founded by MIT and Stanford dropouts (and married couple) Michael Gao and Tina Ju, aims to make modern cryptographic techniques such as zero-knowledge proof (which allows you to prove things without revealing exactly what you know) and fully homomorphic encryption (which allows you to work with encrypted data without decrypting it) more widely used. This, the co-founders claim, will alleviate what they perceive as a fundamental tension between trust and privacy in an age where firms collect increasing amounts of data about consumers while also becoming increasingly unable to protect this data.

To demonstrate this, the researchers created a special RISC-V-based processor designed to run the algorithms required to construct zero-knowledge proofs and enable completely homomorphic encryption. Fabric describes this as a "verifiable processing unit" (VPU). According to the scientists, existing technology is simply too slow to make fully homomorphic encryption common, hence bespoke chips are required.

A number of venture capitalists are now backing the proposition, including Blockchain Capital and 1kx, who led the $33 million Series A funding round. Offchain Labs, Polygon, and Matter Labs were also involved. It follows Metaplanet's $6 million seed round, which included participation from Inflection and Liquid2 Ventures.

The presence of multiple crypto investors here investing in cryptography is not a coincidence. "We could not be more thrilled to partner with Fabric on their mission to accelerate all cryptographic operations with the world's first VPU to help bring about a world where privacy and verifiability are non-negotiable components of all digital systems," Yuan Han Li, a Blockchain Capital investor, said.

While Fabric is targeting corporations, the blockchain world has the most immediate need for its solution, which is why its first-generation chip will focus on zero-knowledge proofs.

“With zero-knowledge proofs and fully homomorphic encryption, it’s really starting to look like AI in the 2010s, with exponentially increasing compute demand,” Gao told me. “You can tell that although not everyone understands what it is, it’s increasing in ubiquity already within the crypto space. And we think that it goes beyond crypto. It goes to pervasive cryptography in the enterprise. And so we realized that this is really a mainstream moment, and hardware could be the final accelerant that pushes it over the edge and allows cryptographers to make their dreams come true. That’s why Tina and I teamed up and started this company.”

Fabric claims it has already received "tens of millions of dollars in preorders."

Enterprises also do not want to deal with "the toxic sludge of user data," according to Ju, because it is mostly a liability. So if they can retain that data encrypted even while it is in use, or if they don't even need to obtain it and can instead share only what is necessary via zero-knowledge proofs, that's a huge victory for them.

Gao is no stranger to ambitious hardware firms. He previously co-founded the photonic AI business Luminous Computing. He then co-founded the Bitcoin mining chip firm Katana in 2021, becoming CEO in 2022. Katana was rebranded as Fabric Systems. Fabric Systems is now defunct, yet some of its DNA appears to linger on in Fabric cryptography.

However, the couple's experience at Luminous may have served as a wake-up call. "We were both disillusioned, despite my personal excitement about large language models, and despite the vast potential of AI in general … the most practical go-to-market for photonic computing seemed to be giant AI machines for targeted ads on social media — a use case that hardly got me out of bed in the morning," according to Gao.

Fabric is preparing to roll out its first chips, and the business intends to utilize the fresh funding to develop the next generation of chips and expand its software and cryptographic teams. In the long run, Fabric intends to get its chips inside enterprise data centers and possibly sell to hyperscale clouds.