Arcium, MagicBlock, Umbra, Darklake... A comprehensive overview of the private DeFi and AI ecosystem being built on Solana.

Author: Castle Labs

Article translated by AididiaoJP, Foresight News

Solana's privacy ecosystem is still in its early stages.

We spoke with Helius CEO @mert to hear his thoughts on Solana’s privacy. In his own words, Solana is “a bit behind” in privacy.

What should a mature privacy ecosystem look like?

Required elements include:

• Formal verification
• No committee
• Immutability
• Open-source code

Solana’s unique architecture also enables different privacy priorities compared to EVM chains. Mert specifically highlighted ZK compression: “On Solana, we can achieve large-scale and composable privacy protocols without persistent rollups—at least not when using base rollups.”

In the specific context of Solana, the two most relevant verticals for privacy development are neobanks and Private DeFi. However, in terms of tools and user experience, Solana still has a significant gap to overcome before achieving a fully functional and composable privacy ecosystem.

We also asked Mert about his views on the privacy technology stack. As the report emphasizes, privacy should not be viewed as a single technology, but rather as a “final privacy stack” where all primitives ultimately work together. For Mert, the endgame will be the combination of FHE (Fully Homomorphic Encryption) and ZK (Zero-Knowledge Proofs). TEEs and MPC are practical for certain use cases, but they cannot provide sufficient guarantees in adversarial systems.

Finally, we asked him for his thoughts on Helius Privacy.

Helius Privacy will be developed as a ZK-based UTXO privacy layer on Solana, utilizing "Zones" to allow individual companies to choose their own trade-offs.

A public zone will also be provided for all general users, offering complete anonymity in an immutable and formally verified manner. More details will be announced soon.

Based on the above context, this article focuses on how the Solana privacy ecosystem addresses various privacy challenges.

Private Compute

Currently, there are two main providers in this space: @Arcium and @magicblock.

Both are addressing similar issues: privacy-preserving computation.

Arcium processes arbitrary data using MPC (Multi-Party Computation). It splits and distributes data across independent clusters of nodes that collectively compute the result without seeing any individual input. Meanwhile, Arcium operates as an independent computation network, with final settlement handled on Solana, which manages task ordering, network security, and fee payments.

All of these privacy-preserving computations occur within Multi-Party Execution Environments (MXEs)—customizable, parallelized execution environments.

Beyond solving computational problems, Arcium’s products also serve the broader privacy ecosystem of Solana. They are building the Confidential SPL (C-SPL) token standard, enabling confidential tokens, transfers, and transactions on Solana.

We asked the Arcium team about the sources of demand. As expected, the primary demand comes from payments and cryptocurrency data analytics, with growing institutional interest, particularly in healthcare—enabling the training of models on cryptocurrency datasets. C-SPL also enables seamless private transfers, further attracting institutional interest.

In terms of data, since the Alpha mainnet launched in early February 2026, Arcium has processed over 900,000 computations and 3.5 million transactions, with most of the growth occurring in early May.

Most current demand comes from early applications like Umbra. Over the coming weeks, applications such as ZINC and Crafts will also launch, and we expect demand to increase further.

ZINC is performing proof-of-work mining for cryptocurrencies, while Crafts is utilizing Arcium to conduct sealed-bid auction fundraising, enabling startups to tokenize a portion of their equity through fair price discovery.

Some exciting new developments being built on Arcium include capital formation through sealed-bid auctions, opportunity markets, crypto settlement in prediction markets, and other financial privacy applications.

Many of these use cases are creating new or improved markets where users are unaware of privacy.

Magic Block addresses privacy computation through TEE (Trusted Execution Environment), while Arcium relies on cryptographic guarantees provided by MPC. Its product operates by using Intel TDX to create a hardware-verified black box—Private Ephemeral Rollup (PER)—where transactions are aggregated and processed before being submitted back to Solana.

MagicBlock helps developers preserve these properties throughout the stack, including: confidentiality (protected state), scalability (high throughput), composability (still interoperable with other Solana programs), and compliance (access control layer).

Although they take different approaches to solving the problem, both can generate deployable private order books, dark pools, and private DeFi lanes with minimal code changes. This is already evident in the ecosystem built on Arcium, spanning areas such as DeFi, prediction markets, and neobanks.

Private Transfers and Balances

As privacy-preserving computing infrastructures like Arcium and MagicBlock continue to mature, use cases built on these infrastructures are also growing, including private transfers.

@UmbraPrivacy is the first to be built on Arcium’s MPC infrastructure. Umbra introduces Encrypted Token Accounts (ETAs), the direct equivalent of Solana’s standard Associated Token Accounts, but with balances stored in encrypted form, providing:

• Amount privacy: Transaction amounts are encrypted using Rescue cryptography.
• Balance privacy: Balances are stored in encrypted form.
• Privacy linkage: Completely sever on-chain connections between sender and receiver using a shielded pool combined with ZK proofs.

In addition, Umbra provides compliance features that allow users to grant selective viewing access to auditors and compliance systems without exposing their full transaction history. This is crucial for institutional workflows and for users who wish to prove fund ownership without revealing detailed transaction histories.

In the privacy wallet space, @theprivacycash and Hush are two others.

Privacy Cash uses a Tornado-style mixing pool for SOL: users deposit SOL to generate a commitment and add it to a Merkle tree, then use a ZK proof to withdraw to any recipient, completely severing the on-chain link between the deposit and withdrawal addresses.

Hush is inspired by Zcash but enhances DeFi utility. Users deposit SOL into Hush’s shielded pool, where it is automatically converted to jitoSOL, enabling passive earnings from staking rewards and MEV income while maintaining privacy. Within the pool, users can send and receive funds to and from other Hush participants and conduct multiple transactions without interacting with Solana’s public ledger. Upon withdrawal, funds are unlinked from the original deposit through mixing. Transfers within the pool incur a fee of 0.01 jitoSOL, and withdrawals are charged a 50 bps fee. Hush also integrates Jupiter for private swaps and implements geographic restrictions for sanctioned regions, fostering a strong compliance profile among institutional users.

No Onchain Trail

Private transfers address the issue of companies and institutions paying employees or conducting private transactions on-chain. But to go further, we need to embed privacy into everyday on-chain activities, especially transactions.

Every order placed on a public AMM is a signal that frontrunners, copy traders, and MEV bots can read and exploit. Several protocols on Solana are already addressing this issue.

@encifherio is a privacy-first DeFi interface that routes trades through Jupiter while keeping transaction details private. The team stated: “Execution quality remains unchanged, as we are not adding custom routes—we use the same routes and liquidity provided by Jupiter.”

It wraps the tokens the user wishes to swap and encrypts the swap details using ElGamal encryption. On-chain records only show state changes of the wrapped asset types, sufficient for Jupiter to verify that the correct tokens are being routed. The number of transactions, counterparties, participants, and even whether a transaction was executed are all processed within a TEE environment (AWS Nitro Enclaves) and never publicly broadcast. This enables large-scale private swaps.

VanishTrade takes a different approach by building a private transaction infrastructure that uses obscured transaction routing to protect trading strategies from leaving on-chain traces. Unlike encifherio’s wrapped tokens, Vanish obscures liquidity routing for transactions. Additionally, Vanish has launched the Vanish Integrity Framework (VIF), supported by Elliptic and Range, which embeds safeguards to prevent the routing of any illicit transactions.

Darklake is another competitor in this category, building ZK-native liquidity infrastructure and dark pools. Its zk-AMM, called the "Blind Spot Slippage Pool," adds a commit layer on top of AMMs to hide slippage data prior to execution. Searchers cannot read order intent before trades are executed, but can verify outcomes afterward. This delayed asymmetry prevents sandwich attacks while preserving verifiability. They have extended this model to private perpetual contracts (zk-Perps) using the Arcium compute layer, and to a reasoning framework called Zyga, which abstracts proof complexity to provide builders with a foundation for secure logic and coordination. Recently, they have further expanded into an infrastructure protocol that enables applications and users to privately verify, connect, and compute using the "Proof as Intelligence" model.

Private Prediction Markets

Private prediction markets are a more advanced privacy application because users' strategies can easily be copied, leading to a loss of advantage. To address this, the protocol is using the Arcium infrastructure to build dedicated dark pools for prediction markets.

@meleemarkets is building a prediction market that supports private order flow. They encrypt the order book using Arcium’s MPC infrastructure, allowing participants to place orders without revealing their direction on the public market until settlement.

Private AI

As AI agents increasingly operate on-chain, will every query they make and every piece of PII they consume be permanently public?

Loyal answers this question through its decentralized, censorship-resistant smart protocol, built using Magic Block’s temporary rollup execution and Arcium’s encrypted computation. They are building on-chain AI that protects user data—conversations, queries, preferences, and activities—all encrypted and stored on Solana with strict access controls. Users own and can export their encrypted conversation history, and can self-host the frontend without losing any data. Additionally, Loyal supports private transactions and fund management, allowing depositors to earn yields privately.