Fhenix activated its CoFHE coprocessor on Ethereum and Arbitrum mainnets on October 21, 2026, marking the first production deployment of fully homomorphic encryption (FHE) as a coprocessor service on general-purpose EVM chains. The system enables smart contracts to perform arithmetic operations —...
Fhenix activated its CoFHE coprocessor on Ethereum and Arbitrum mainnets on October 21, 2026, marking the first production deployment of fully homomorphic encryption (FHE) as a coprocessor service on general-purpose EVM chains. The system enables smart contracts to perform arithmetic operations — addition, multiplication, comparison — on encrypted data without decryption. Two design-partner applications, Fluton and AlphaEngine, went live at launch. Three additional dApps are scheduled for Q4 2026.
The launch arrives at a specific inflection point for on-chain privacy infrastructure. Tokenized real-world assets have reached approximately $14.5 billion (as of October 20, 2026), and McKinsey projects the tokenized market to reach $1–4 trillion by 2030. Institutional participants consistently cite on-chain data exposure — visible balances, trade sizes, portfolio compositions — as a structural barrier to adoption. Fhenix's dual-mode token standard, which allows a single asset to carry both a public ERC-20 balance and an encrypted confidential balance within one contract, represents the first production attempt to address this on live EVM infrastructure.
Fhenix has raised $22 million across seed and Series A rounds. The system launches in permissioned mode: contracts must be allowlisted before accessing CoFHE in production. No sustained throughput, latency distribution, or cost benchmarks have been published. Consensys Diligence completed a security audit in September 2026, though the public report has not been released.
Fully homomorphic encryption allows computations to be performed directly on ciphertext. The result, when decrypted, matches what would have been produced by running the same operations on plaintext. In practical terms: a smart contract can add two encrypted balances together and produce a valid encrypted sum without ever seeing either input.
Fhenix's CoFHE is a coprocessor, not a standalone chain. Smart contracts on Ethereum or Arbitrum emit an FHE computation request. Off-chain aggregators detect the request and route it to the FHE engine. CoFHE performs the computation on ciphertext off-chain. Results return to the host chain as a hash (pointer) referencing the encrypted output. An optional threshold network decryption step makes data available to authorized parties.
The integration surface is deliberately narrow. Developers add encrypted data types to standard Solidity contracts: euint8, euint16, euint32, euint64, euint128, euint256, ebool, and eaddress. Converting a standard variable to an encrypted one requires changing uint256 to euint256 — a single type substitution.
There is a meaningful constraint: encrypted booleans cannot drive standard Solidity if statements. Developers must use encrypted selection operations instead, which alters contract logic design.
CoFHE separates the FHE computational workload from the EVM execution environment. The host chain (Ethereum or Arbitrum) handles transaction ordering, state commitments, and gas accounting. The coprocessor handles encrypted arithmetic.
This design reflects a pragmatic trade-off. FHE operations remain computationally expensive — orders of magnitude slower than plaintext equivalents. Running them inside the EVM would be prohibitively costly in gas terms. By offloading to dedicated off-chain infrastructure, Fhenix avoids the gas ceiling problem while preserving the security guarantees of the host chain for state finality.
Fhenix claims its threshold decryption method delivers approximately 20,000x higher throughput and up to 37x lower latency compared to "traditional approaches," though the baseline for this comparison has not been specified. No independent benchmarks have been published.
The system records commitments associated with resulting ciphertexts on-chain, providing a verification trail. Shielding (converting a public balance to confidential) requires one transaction. Unshielding (confidential to public) requires a request followed by a claim after decryption completes — a two-step process.
Two design partners launched alongside the mainnet:
Fluton operates as a confidentiality engine built on encrypted intents. Users can express transaction intentions in encrypted form, with execution occurring without exposing the underlying parameters.
AlphaEngine is a strategy platform where trading methods remain encrypted. Capital deployment endpoints were not live at launch, according to reporting from HackerNoon, meaning the platform functions as a strategy evaluation tool with execution capabilities pending.
Fhenix selected these two from approximately 60 candidates. Three additional applications are planned for Q4 2026, though names have not been disclosed. The company is also exploring Solana integration through a Colosseum hackathon.
The most commercially relevant feature in the launch is the dual-mode token standard. A single smart contract holds two states for the same asset: a public balance that behaves identically to a standard ERC-20 token, and a confidential balance encrypted via FHE.
Holders can shift between modes in a single transaction. No bridge, no wrapper, no separate chain. The public balance participates in standard DeFi composability. The confidential balance shields amounts, positions, and activity from on-chain observers.
This addresses a specific structural problem in institutional adoption. Fund managers, corporate treasuries, and regulated entities cannot place assets on public ledgers where competitors can track their positions in real time. The dual-mode approach preserves the asset's existing distribution and market infrastructure while adding a private layer on demand.
The limitation is equally specific: sender and recipient addresses and transaction timing remain publicly visible. Confidentiality applies only to amounts and balances. Metadata leakage — the fact that address A transacted with address B at time T — is not mitigated.
Fhenix does not operate in isolation. The FHE-for-blockchain sector has seen rapid development in 2025–2026:
Zama is the dominant player. The Paris-based company reached unicorn status ($1 billion+ valuation) after raising over $150 million, including a $57 million Series B in June 2025. Zama launched its TFHE-based protocol on Ethereum mainnet on December 31, 2025, and claims approximately 70% market share in blockchain FHE with over 5,000 developers in its open-source ecosystem. Zama provides the underlying encryption libraries that multiple projects, including early Fhenix iterations, have used.
Inco Network (USA) builds a standalone confidentiality layer for Web3 with encrypted smart contract execution and programmable access control.
Mind Network focuses on encrypted AI and DeFi applications with an FHE-based restaking infrastructure.
The broader privacy infrastructure market spans three competing technologies: zero-knowledge proofs (ZK), FHE, and trusted execution environments (TEE). ZK remains the most mature, with over $11.7 billion in combined project market capitalization and an estimated 2,000+ applications by 2026. TEEs power over 50 blockchain projects. FHE is the newest entrant to production, with the highest privacy guarantees but the most significant performance overhead.
According to a February 2026 analysis by Bex.co, the most sophisticated production systems in 2026 now combine two or three of these approaches, using each technology where its specific trade-offs matter least.
Fhenix has raised $22 million in total across two rounds:
| Round | Amount | Lead Investor | Key Participants | |-------|--------|---------------|-----------------| | Seed | $7M | Multicoin Capital | Collider Ventures | | Series A | $15M | Hack VC | Amber Group, Primitive Ventures, GSR, Stake Capital, Dao5 |
Strategic investors include BIPROGY (Japan, $2.6 billion annual revenue) and Tandem, the venture arm of Offchain Labs (the team behind Arbitrum). The Tandem relationship extends beyond capital: Fhenix founder Guy Zyskind stated the two entities are "working on a confidential stablecoin together."
Offchain Labs CEO Steven Goldfeder described the launch as "the beginning of something big: a programmable economy." Two additional institutional partnerships are scheduled for announcement, per Fhenix.
The company was founded by Dr. Guy Zyskind, an MIT researcher and co-founder of Secret Network, and is led by CEO Guy Itzhaki, a former Intel Director who oversaw the Homomorphic Encryption and Blockchain Group.
Consensys Diligence completed a pre-launch audit in September 2026. One acknowledged finding: approved operators can read exact holder balances — a design choice that enables compliance use cases but limits privacy guarantees for counterparty-facing interactions.
The gaps in the current deployment are material:
No published benchmarks. Fhenix describes throughput as "dynamic" and "scaling with demand" but has released no sustained throughput figures, latency distributions, or per-operation gas cost data. For a technology where performance is the primary constraint, this is a significant omission. The broader FHE field faces this issue — vendor claims of 20,000x improvements lack standardized baselines for comparison.
Permissioned launch. All contracts must be allowlisted to access CoFHE. This gating mechanism limits composability — the core value proposition of public chain deployment. Fhenix has not disclosed criteria or timeline for permissionless access.
Metadata leakage. Addresses and timing remain public. For institutional use cases involving counterparty privacy, this limits confidentiality to balance and amount data only.
Audit transparency. The Consensys Diligence audit report has not been made public. Two wrapper hosts added after the audit are excluded from its scope (per HackerNoon reporting).
Economic model undefined. No token has launched. Fee structures for CoFHE computation are not published. The cost of FHE operations relative to standard EVM execution — a critical variable for developer adoption — remains unknown.
Application maturity. Of the two launch partners, AlphaEngine's capital deployment endpoints were not live at launch. The practical utility of the launch ecosystem is limited until full-featured applications are operational.
The timing of Fhenix's launch reflects a growing structural tension in on-chain finance. As tokenized assets expand — $14.5 billion in tokenized U.S. Treasuries alone as of October 2026 — the demand for confidentiality infrastructure grows proportionally.
According to a Chainlink privacy-preserving tokenization report, institutional participants face a concrete problem: investor allocations, fund subscriptions, lending agreements, and transaction details contain sensitive information that cannot be exposed on a public ledger. Competitors can reverse-engineer trading strategies from on-chain activity. Regulatory frameworks (GDPR, HIPAA, securities law) impose confidentiality requirements that conflict with blockchain transparency.
The current market offers three approaches to this problem:
Fhenix's bet is that the coprocessor architecture — offloading FHE computation while anchoring state on public chains — threads the needle between performance limitations and institutional requirements. Whether the current performance envelope is sufficient for production workloads at scale remains unproven.
Fhenix's mainnet deployment is a concrete engineering milestone: FHE computation on encrypted smart contract data is now available on Ethereum and Arbitrum as a production service. The dual-mode token standard offers a structurally elegant solution to the tokenization privacy problem, allowing assets to move between transparent and confidential states without changing venues.
The economic case is less clear. FHE remains the most computationally expensive privacy technology in production. No benchmarks, fee structures, or cost comparisons have been published. The launch is permissioned. The application ecosystem consists of two partially operational dApps. And Zama, with 7x the funding and an 8-month head start on Ethereum mainnet, holds the dominant position in the FHE infrastructure market.
What Fhenix has demonstrated is that the coprocessor model works — encrypted computation can be offloaded from the EVM while preserving host-chain security guarantees. Whether this architecture generates sufficient throughput at acceptable cost for institutional-scale tokenization workloads is the question the market will answer over the coming quarters. The $14.5 billion tokenized asset market (and the trillions projected behind it) needs confidentiality. The open question is whether FHE, in its current form, can deliver it at the speed and cost that money demands.