The Zama FHE Master Plan

March 7, 2024
Rand Hindi

Today marks an important milestone toward our mission of making the internet end-to-end encrypted using Fully Homomorphic Encryption (FHE). 

I’m proud to announce that Zama has raised $73 million in a Series A—one of largest venture rounds in France’s history—co-led by Multicoin Capital and Protocol Labs, two of the leading pioneers in the blockchain industry, with participation from Metaplanet, Blockchange, VSquared, Stake Capital and Portal Ventures, as well as several other strategic founders, including Juan Benet (Filecoin/IPFS), Gavin Wood (Ethereum/Polkadot), Anatoly Yakovenko (Solana), Julien Bouteloup (Stake Capital Group), and Tarun Chitra (Gauntlet). This round brings together the brightest minds in the industry to work toward a singular goal: making FHE ubiquitous in blockchain and AI. 

Four years ago, we founded Zama as an open source cryptography company focused on FHE, the “holy grail of cryptography.” At that time, FHE was more theoretical math than practical code—it was too slow, too expensive, required too much cryptography expertise, and could only support a handful of real-world applications. In the years that followed—largely thanks to the relentless efforts of our team—we methodically tackled each of these problems, making FHE generally accessible to non-cryptography developers for the first time. I’m incredibly proud of this work. It alone is a significant achievement. 

Today, Zama offers a robust suite of open source FHE libraries and solutions. Anyone, from solo developers to large enterprises, can build applications that deliver out-of-the-box, end-to-end encryption to their users. Developers don’t need to know anything about cryptography to get started. Our FHE scheme, which builds on top of TFHE, supports any type of application, no matter how complex, vastly expanding the applicability of FHE. 

Alongside security and developer experience, we maintained an unrelenting focus on performance. Since inception, we have improved the speed of our FHE scheme by a factor of 20x, and are on track to reach 100x soon. This key performance milestone has unlocked critical confidential blockchain and AI use cases. Dedicated FHE hardware accelerators, which are coming in the next two years, will bridge the final gap to enable web-scale applications such as confidential Large Language Models (LLMs) and encrypted Software-as-a-Service (SaaS).

In addition to advancing our research, this funding gives us several years of runway and the necessary resources to support our partners that are going into production with FHE today. While our partners span a wide cross section of industries, we found blockchain to be a key market for Zama to focus on this year. A major issue with blockchain is that all transactions and data are publicly visible, creating a challenge for developers of applications that require using sensitive information, such as personal or financial data. To address this, we created the fhEVM, a confidential smart contract solution that enables developers to create confidential onchain applications in Solidity. The fhEVM can be integrated into any EVM-compatible chain, ensuring that the state remains encrypted end-to-end, from the moment a transaction is submitted to when smart contracts are processing it.

Several incredible projects have already integrated the fhEVM: 

  • Fhenix is an FHE Layer 2 rollup which uses the fhEVM to enable confidential transactions in the Ethereum ecosystem, addressing the critical need for privacy in scaling solutions.
  • Shiba Inu, one of the largest tokens by market cap, is using the fhEVM to build an entire network state for their community, powering everything from their decentralized financial infrastructure to their metaverse.
  • Inco is a modular blockchain using the fhEVM to offer Confidentiality-as-a-Service to various Layer 1 networks, thereby enhancing the privacy of existing blockchains. 

Over the coming year, we expect FHE innovation to take off exponentially as more developers discover what they can now build. Some of the more popular blockchain use cases include:

  • Confidential Tokens — where balances and amounts are encrypted. This, for example, allows companies to manage payroll onchain without revealing the identities and salaries of their employees.
  • Decentralized Identity — where a user’s age, nationality, social security number, or credit score can be stored confidentially onchain. This solves the elusive identity problem in crypto and enables regulatory compliance in DeFi, Sybill resistance for airdrops, uncollateralized credit, and much more.
  • Gaming — where attributes can remain hidden during gameplay. Examples include poker, Wordle, or other strategy games.
  • Institutional Finance — where real world assets such as mutual funds or mortgages can be tokenized and traded without revealing the identity and amounts to other participants.

I am also excited about the applications of FHE in Artificial Intelligence. Using our Concrete framework, it is now possible to convert an AI model into an FHE equivalent, and to both train and run inference on encrypted data. Data scientists can already start building simple confidential AI applications today, and integrate more complex models such as LLMs as hardware acceleration becomes available in the near future. 

Blockchain and AI are just the start. Our long term goal is to make the entire internet encrypted end-to-end. Here is our masterplan:

  • First, make FHE extremely easy to use for developers, no matter their use case.
  • Second, make FHE extremely fast, scaling confidential blockchains to 1,000+ transactions per second and generating dozens of encrypted LLM tokens per second.
  • Third, bring FHE to cloud applications, from databases to SaaS.
  • Finally, make the entire internet encrypted end-to-end via a new, FHE-powered “HTTPZ” protocol.

If this plan resonates with you, whether you are considering incorporating FHE in your product or simply experimenting with it, get in touch! My DMs are always open.

Come build the future of the internet with us. 

Rand (@RandHindi)

Read more related posts

People shouldn’t care about privacy

A technique called Fully Homomorphic Encryption is about to radically change how we guarantee privacy online.

Read Article