Skip to content

The CCS Interview

Fhenix CEO: Trusted hardware can't secure DeFi, so privacy has to come from math

Guy Itzhaki, co-founder and CEO of Fhenix and a former leader of Intel's trusted execution environment team, explains how fully homomorphic encryption (FHE) can bring compliant privacy to smart contracts and DeFi. He covers why privacy is back on the agenda, how developers can add it, and where the project stands. The conversation was recorded in March 2026, when Fhenix was on testnet and planning mainnet for the end of the year.

Interview by Ashton Addison Recorded March 2026 7 min read

Key takeaways

  • Fully homomorphic encryption enables computation on encrypted data without decryption, maintaining mathematical structure while preserving confidentiality.
  • FHE differs from zero-knowledge proofs which provide anonymity, and from trusted execution environments which require trusting hardware security.
  • Privacy in blockchain includes both confidentiality of transaction data and compliance capabilities needed for institutional adoption.
  • Current blockchain transparency exposes sensitive information like whale balances and trading strategies that institutions want to keep confidential.
  • Guy Itzhaki previously led Intel's trusted execution environment projects before co-founding Fhenix to build privacy infrastructure using FHE.

From Intel enclaves to FHE

Ashton Addison: What have you and the team built at Fhenix, and how does it relate to encryption?▶ 00:49

Guy Itzhaki: In one sentence, we're building the fastest privacy infrastructure for blockchain, powered by fully homomorphic encryption. I spent a big part of my career at Intel, leading the team that built trusted execution environments, so a lot of the SGX projects you see in the blockchain space are kind of my kids. I love TEEs. Later I managed the team building hardware acceleration for FHE, and I saw the trajectory it was gaining. FHE has been around for many years. It was always considered magical cryptography that lets you compute over encrypted data, but also very complex and very slow. A couple of years ago I left corporate America to join Guy Zyskind, who built Secret Network, one of the earlier pioneers of privacy. The vision for Fhenix was to solve the privacy challenges we see on blockchain today. If you want to bring real money and large players in, you need maturity, and that includes privacy but also compliance, and the ability to do both. Many people ask whether privacy is just for tax evasion. It's really not. The goal is to bring privacy in a way that doesn't impact the business and stays compliant.

Ashton: What is the difference between ordinary encryption and FHE, and why is FHE better?▶ 04:31

Guy: Privacy comes in different flavors. Anonymity, hiding who made a transaction, is more a proving mechanism than encryption, and that's where zero-knowledge proofs play a big role, in projects like Railgun. Where we come in is operating over encrypted data. Everything on blockchain is transparent. Open a block explorer and you can see everything from day one. Institutions doing payments, DeFi projects that want to keep their strategy confidential and whales who want to keep their balances private all tell us that's a problem. Trusted execution environments try to tackle this, but you have to trust the hardware. A couple of months ago we saw that, with not a lot of effort, you can build hardware that listens to the memory and sees what's being computed inside the TEE. They're great for specific applications, but in a decentralized ecosystem where people can get their hands on the hardware, they're probably not the best fit. So we solve it through math. FHE is a specialized type of encryption. We encrypt the data in a special way that maintains its mathematical structure, so we can compute over it while it's encrypted. That sounds like magic, because in school we were taught that encrypted data is meaningless until you decrypt it. FHE changes that dynamic.

I don't think we will ever operate in a fully encrypted ecosystem.

Why privacy matters now

Ashton: Privacy has been a growing narrative since late 2025, with Zcash, Monero and Vitalik talking about it. Why does it matter more than ever, and is AI part of it?▶ 08:41

Guy: I'm very happy that the rise of privacy is here. When we started two and a half to three years ago, we had to convince people it was important. They'd say, "I use Gmail and everybody knows everything about me." I see a few reasons for the change. The first is the stablecoin narrative. I think stablecoin brought a product-market fit to crypto. Once you have that, you start doing payments, and then privacy matters. With meme coins and NFTs it was hard to explain why it mattered. The second is institutional demand. Institutions such as JPMorgan Chase and Robinhood come in with a privacy mindset, but it has to let them maintain compliance. They still want KYC and AML, so the solution has to be flexible enough to give them the privacy levels they need without opening the door for the regulator to push back. The third is the rise of AI, though I think it's still a bit early. We see a lot of agent-to-agent payments, but once there's a very clear product-market fit there, I think privacy will probably be the next order.

Ashton: Crypto payments are increasingly framed as being built for AI agents. Are you seeing that, and how does privacy talk turn into production?▶ 11:51

Guy: We definitely see the same trend in the number of projects looking into agent payments. Conversations on privacy have been happening since around Q3 2025, but getting from those conversations to production implementations is a process. Part of it is making sure the infrastructure is ready for what customers need. When we started Fhenix, we started it as an L2, on the assumption that the market needed another L2. Very quickly we realized it doesn't, and that we want to meet developers where they already are. So we switched to a coprocessor architecture, which is basically privacy as a service. It taps into EVM chains like Ethereum, Arbitrum and Base, and developers just import the libraries and start encrypting and decrypting assets in their smart contracts. We also kept Solidity support, because developers shouldn't have to become cryptographers. If I have to count on developers to know FHE, we can close the shop.

Compliant private DeFi in practice

Ashton: If a protocol like Aave or Uniswap wanted to add compliant private DeFi, how simple would the upgrade be?▶ 15:00

Guy: The mechanics of the integration are very simple. It's Solidity and the same tools developers already know. The challenge is the mindset and the design. If the pool is public and the amount you extract is private, it's easy to work out how much you took out. So you have to design it so that data isn't exposed through other mechanics. What it gives them is the ability to operate over encrypted data while still doing KYC and AML, and in certain conditions what we call KYT, know your transactions. Sometimes a regulator or auditor will want to see the content of an encrypted transaction. We don't have access to the key, but we can help smart contract developers define the conditions under which a third party could see the data. That helps close the compliance gap.

Ashton: How do you deal with the last mile, where an encrypted ecosystem still has to connect to everyday payments?▶ 21:13

Guy: The best approach is to be very practical. I don't think we will ever operate in a fully encrypted ecosystem. You define what you want to protect, operate around that, and make it very easy to deploy. One showcase we released a few months ago lets stablecoin issuers create a stablecoin with a dual mode, private and public. A purely private stablecoin is hard to plug into yield-bearing DeFi, because the likes of Aave can't use it yet. With dual mode, I can hold a balance or make a transfer privately. If I want yield from an entity that doesn't support private tokens, I switch to public mode, and they don't need to change anything.

Ashton: Could we see private versions of USDC and USDT?▶ 23:16

Guy: We definitely hope so, and we're reaching out to stablecoin issuers for exactly that. The integration is really simple. People used to think moving to a private stablecoin was very complex, but it's not. With four lines of code you can issue a new stablecoin that is either fully private or has this dual mode. Now it's about building the momentum and the demand.

If I have to count on developers to know FHE, we can close the shop.

Testnet today, mainnet by year-end

Ashton: What stage is Fhenix at, and is it deployable and usable right now?▶ 17:07

Guy: We're on testnet on Ethereum, Arbitrum and Base, and we're looking at additional chains. The first step is making sure developers can access the coprocessor on those chains. Then we need protocols to build on it. We have around eight projects actively building on us, including a private lending protocol, confidential transactions and confidential payments, and we're starting a buildathon to bring in more. We're also in discussions with DeFi protocols working on payments, dark pools and OTC. And there are active conversations with some of the largest institutional players out there, which is heartwarming to see after so many years of talk.

Ashton: What comes next for adoption on the building and implementation side?▶ 24:09

Guy: We're planning to move to mainnet toward the end of the year. That's a lot of work on our part, because we want to make sure the product is secure, safe and fast. On the builder side, we keep releasing new SDK versions that make things easier for developers and reduce the impact on user experience. A major SDK update was due at the end of that week. Beyond that, we'll keep onboarding builders and working on a few in-house initiatives that we can hopefully share next time.

Ashton: What's the best way for developers to get started, especially now that AI makes everyone a developer?▶ 25:49

Guy: All the information is on the Fhenix website. People are vibe coding very quickly, and I wouldn't encourage going into production with a product that was vibe coded, maybe in the future. But for ramping up, definitely. About a week before this interview we held an open house with Arbitrum in New York, and projects building on Arbitrum were able to integrate it into their products very quickly. Use your favorite LLM tool or just read the documentation. We have a great dev rel team, and anyone who needs support can reach out.

This interview has been edited for length and clarity. Recorded March 2026.

Guy Itzhaki
Guy ItzhakiGuy Itzhaki is Co-Founder and CEO of Fhenix, a blockchain research and development company that builds fully homomorphic encryption (FHE) infrastructure for Ethereum and other EVM chains. Before Fhenix he worked at Intel, where he was Director of Homomorphic Encryption and Blockchain Business Development. At… Full profile →
Share this interviewXLinkedIn
▶ Watching the interview