Deep Dive into the Investment Code of Web3 Infrastructure: Guiding You to Find Alpha (Part 1)
TechFlow Selected TechFlow Selected
Deep Dive into the Investment Code of Web3 Infrastructure: Guiding You to Find Alpha (Part 1)
Using Nym, Evmos, and Fission as examples, jointly explore popular Layer2 and privacy-focused projects; the three most promising infrastructure sectors in 2023.
Navigating Web3 tides with focused insightsSoftware is eating the world!
CRCapital continues to focus on the Web3 sector, gradually open-sourcing internal research insights and behind-the-scenes stories into a series of interviews and reports shared with our community.
We are honored to have invited Mr. Jason Xie, Senior Investment Manager at Huobi Ventures, for this conversation. He has led investments in prominent Web3 infrastructure projects such as Evmos, Arbitrum, zkSync, and Optimism PBC.
This article (Part 1) highlights key excerpts from Jason Xie’s insights on the investment rationale behind Layer2 project Arbitrum; explores popular Layer2 and privacy-focused projects using Nym, Evmos, and Fission as case studies; and identifies three most promising Infra sectors in 2023. (Part 2) will cover how an expert in Infra research and investment is shaped, along with Jason’s reflections on fundraising by Chinese-founded projects and investment strategies by Chinese funds.
This article is based on the dialogue and does not constitute investment advice. For original AMA records and future investor and project interviews, follow us on Twitter @CRCapitalcrc and @RLabsWeb3—we look forward to growing together with you.
Interview Summary:
01 About Jason Xie
02 Investment Logic Behind Layer2 Project Arbitrum
03 Exploring Hot Layer2 and Privacy Projects – Examples: Nym, Evmos, Fission
04 Three Promising Infra Sectors in 2023 According to Jason Xie
"Building infrastructure that yields massive returns usually requires founders to hold a strong conviction. Technological innovation is like crossing a river—you must wade through it, not just walk around." — Jason Xie, Senior Investment Manager, Huobi Ventures
"1. Introduce yourself?"
I'm Jason Xie, Senior Investment Manager at Huobi Ventures. I studied telecommunications engineering at Tianjin University, where my coursework covered both hardware and software topics. After graduation, I worked in software development, which laid the foundation for investing in technology-driven infrastructure projects. I’ve also had entrepreneurial experience before joining a top-tier exchange to focus on Infra investments. Recently, I’ve been focusing on the zero-knowledge (zk) space, particularly exploring use cases where zk integrates with other technologies and infrastructure.
Our portfolio spans several tracks. In Layer2, we cover major directions including Arbitrum, Optimism, and zkSync. In cross-chain, we invested early in Connext, which uses liquidity incentives to power its bridge model, and Debridge, another player in the cross-chain space. On privacy, we backed Nym—a project also supported by a16z. We also have exposure to derivatives and Web3 middleware among others.
"2. Why did you invest in Arbitrum? What was your investment logic?"
Arbitrum stands out with exceptional TVL and overall performance, leading the Layer2 race. The latest valuation reportedly exceeds $500 million, giving us conservatively a 50x–60x return.
Within Huobi Ventures, Ethereum scaling was a strategic priority—we just needed to decide among different technical approaches. At the time, Arbitrum was mainly compared against Optimism (OP) and zkSync.
zkSync 2.0 hadn’t launched yet, so there were significant unknowns despite its undeniable technical lead. We therefore set zkRollup aside temporarily and focused on comparing Arbitrum with OP. Broadly speaking, one key difference stood out: OP only achieved about 99% EVM compatibility because some instruction sets in their custom EVM execution layer weren’t fully supported, while Arbitrum was the first to achieve full 100% EVM compatibility. Later, they introduced high-performance WASM-based execution layers.
Though it seems like just a 1% gap, this made a huge difference in real-world adoption speed. By late 2021, users were already experimenting with new “meme” projects on Arbitrum. This eventually gave rise to quality protocols like GMX and various derivatives—financial instruments naturally suited for Layer2 environments. This case clearly demonstrates the importance of technological innovation and aggressive market strategy. OP lagged slightly here, adopting a more conservative approach compared to Arbitrum’s offensive go-to-market tactics. I find this contrast fascinating—Arbitrum succeeded by excelling in both tech innovation and market execution.
"3. ZK technology has gained significant traction recently, sparking intense debates between OP and zk camps. How do you frame your understanding?"
While the market often frames this as a binary competition, I believe it should be viewed dynamically. zkSync’s testnet Phase 2 still faces challenges. Zooming in, zk proofs—whether generation or verification—incur complex computations involving fast Fourier transforms, matrix operations like MSM, and hashing steps. Academic research remains active, with rapid algorithm iterations—recent examples include Halo2.
We expect zkRollup performance to improve over time, but this is a long-term view. In the short term, since no true mainnet launch has occurred yet, it remains unproven. Meanwhile, OP and Arbitrum have demonstrated solid TVL growth, validating them as optimal solutions at the current technological stage.
Looking further ahead, improvements in zk algorithms and dedicated zk acceleration chips could allow zkRollups to catch up or even surpass existing solutions. These two technologies are engaged in an ongoing competitive evolution. For instance, Arbitrum already enjoys capital advantages built during early adoption, fueling network effects. However, capital flows freely and rapidly on-chain without friction. If zkSync launches a breakout dApp, capital can quickly shift. Vitalik Buterin previously suggested combining zk and OP approaches, which reinforces my belief that we should maintain a dynamic perspective.
"4. Can you introduce some privacy-focused projects you've invested in?"
Let me talk about Nym, a project also backed by a16z. Its underlying technological breakthrough is truly innovative. Since our investment in mid-2021, it has remained highly cutting-edge within its niche. The project originates from Cmix, an algorithm developed in the 1980s by David Chaum, who later launched another project called XX Networks. When people think of encryption obfuscation protocols, they often recall Tor (the Onion Network), widely used in darknets to prevent tracking by entities like the FBI or regulators. However, Tor isn't inherently secure. Conceptually, Tor operates as a traffic pool with one entry and one exit point. The FBI can simply deploy numerous monitoring nodes at the exit points.
Here's an example attack scenario: How might the FBI identify the IP address and server location of a hidden service like "Silk Road"? Since Tor appears as a single-entry, single-exit traffic pool, the FBI can flood all websites with access requests and monitor normal traffic patterns. Given sufficient resources, they can blanket-cover many sites. Then, at a specific moment—say, late at night when traffic is low—if someone accesses the Silk Road site, the FBI detects the request. Using pre-deployed exit-node monitors, they trace which node transmitted the traffic. By analyzing outgoing traffic combined with AI-powered big data analytics, they can pinpoint the actual server IP beneath that node.
Nym’s obfuscation network is far more sophisticated—you cannot determine the sender’s IP at all. Suppose Node A in the Nym network sends a message. Normally, messages would be delivered in order of arrival. But Nym introduces a queuing mechanism where messages wait before being sent—not FIFO (first-in-first-out), but randomized. The first message in might be the last out, or dispatched mid-sequence. This shuffles transmission timing, preventing adversaries from correlating send times with identities. Additionally, Nym uses mixnodes—similar to onion routing—that encrypt messages in layers, fully obfuscating them. Data packets pass through these encrypted mixnodes without being decrypted until the final recipient, who peels off each layer sequentially. As a result, no intermediate node knows the origin or content, achieving complete anonymity: recipients receive data without knowing who sent it or from which IP.
The system leverages token economics effectively—mixnodes are incentivized with tokens. More mixnodes mean higher obfuscation complexity. To enhance bandwidth and network efficiency, Nym implements algorithmic optimizations. This decades-old algorithm wasn’t practical earlier due to limitations in hardware and algorithmic efficiency—it existed only in theory until now. Edward Snowden serves as one of their advisors.
Most users don’t perceive privacy needs directly. Zcash aimed to enable private transactions without revealing balances, yet saw limited adoption. However, anonymous communication networks are crucial in countries with strict policies, and exchanges sometimes block regional IPs. Such protocols exhibit strong network effects—users must install a client and often encourage peers to join, enhancing mutual security and driving viral growth. Nym offers absolute privacy—impossible to de-anonymize. Even Lightning Network packets are transmitted in plaintext. My investment decision stems from tangible demand, genuine technological innovation, and a scalable, expandable architecture.
"5. What was your investment rationale for Evmos and Fission?"
Evmos’ core value lies in integrating Cosmos’ cross-chain communication protocol (IBC) with the EVM. Unlike Solana’s VC-driven ecosystem—fueled by large capital inflows to build dApps—Cosmos resembles a tree with deep roots and sprawling branches, organically grown by its community.
Our investment in Evmos was driven by several factors: First, its core team originated from the Cosmos Foundation. Second, considering network effects, Ethereum already dominates with strong developer adoption. Without clear differentiation, competing directly with EVM chains would be difficult.
Simply being another EVM-compatible chain wouldn’t justify investment. What made Evmos compelling was combining EVM with IBC (Inter-Blockchain Communication), enabling extremely high composability. Within Cosmos, project valuations are relatively evenly distributed, allowing seamless integration across ABC (Application Blockchain Interface) and various EVM chains. This enables organic ecosystem-wide composability, while Ethereum developers can easily migrate. Thus, Evmos unlocks massive potential for network effect expansion.
At its essence, Evmos functions as a critical communication layer with broad application potential. It occupies a necessary role in the network stack—one that cannot be bypassed. Whether it ultimately succeeds depends on the team’s ability to incubate new projects. Jack Dorsey’s Damus, for instance, is similarly built atop a foundational protocol.
Fission represents a class of middleware projects. Their CTO is a core EVM developer. My thinking here is that Layer1 has become extremely competitive, with abundant base-layer infrastructure. Yet developers face numerous hurdles: connecting to RPC services, running nodes per blockchain, syncing blocks, deciding whether to use MySQL to cache on-chain data, or writing backend listeners in Go, etc. For example, syncing a full Ethereum node may require several terabytes of SSD storage and days of processing. Then come diverse programming languages—Rust, Solidity—or even custom languages for zk projects—making the learning curve steep and complex.
Middleware is mature in Web2 and cloud computing—their core function is bridging the gap between underlying platforms and developers. Chainlink is a classic example, helping developers fetch price feeds. The Graph helps index and query on-chain data. Chainlink even launched a Keeper service for scheduled tasks, addressing a key limitation of EVMs, which cannot natively support loops or timed executions—requiring external middleware for offloaded computation.
Middleware connects dApps, users, and platforms. Abstractly, imagine it as nodes linking a web-like structure—connecting public chains, dApps, developers, and end-users. Over time, this creates powerful network effects. I focus heavily on middleware due to its scalability and alignment with real market needs. Equally important is marketing and business development—how well the team drives adoption by top protocols and broad developer communities.
"6. Are there any specific Infra sub-sectors you’re particularly bullish on in 2023?"
First, non-interactive cross-chain bridges. Our 2021 cross-chain investments were motivated by the multi-chain ecosystem trend—cross-chain interoperability was inevitable. We mapped the entire landscape: early models like hash time-lock contracts (HTLC), primitive atomic swaps, relayer-based bidirectional monitoring (common among EVM chains), and MPC-based relay services.
When a cross-chain event is registered, systems periodically verify it. Different models manage relayers via MPC, consensus mechanisms, or hybrid designs like LayerZero combining IBC and PoS. However, repeated security breaches revealed a root cause: custodial fund models create honeypots for hackers. Funds are inevitably pooled somewhere—essentially announcing “here’s a vault.” Any misstep—contract upgrades without testing, inadequate audits, flawed processes, poor internal risk controls, or leaked relayer private keys—becomes an attack vector.
Recently, I came across an academic approach: non-interactive bridges using zk techniques—not to be confused with existing zk-based bridges like Dow Song’s model, which aggregates EVM chain transactions on a zkRollup Layer2 and settles back. The model I refer to generates zkProofs for every state change. For example, if a target chain undergoes balance and state changes, it generates a zkProof proving this change and sends it to the source chain. Once the source chain verifies the proof, it confirms the update and proceeds accordingly. The entire process relies on zkProofs, which contain no sensitive data or private keys—eliminating hacker incentives. This fully leverages zk’s potential—not just for privacy, but for secure, efficient interactions. Non-interactivity enhances both safety and efficiency, marking a paradigm shift in cross-chain bridge design.
Second, precompilation and streaming generation. Traditional development compiles all code into binaries upfront. But certain scenarios don’t require full compilation—doing so incurs unnecessary time and computational overhead, especially given large zkProof sizes and high gas costs. Streaming generation compiles circuits on-demand—only when needed—offering flexible, elastic proof generation and verification. Nodes can split compilation workloads across multiple machines to reduce individual resource usage. Algorithm-level innovations—like novel hash functions—can optimize various stages of zk proof generation, each offering room for improvement.
Third, network-effect-driven middleware. This category is broad, and applications haven’t exploded yet—but I believe it warrants continuous attention.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
![Axe Compute [NASDAQ: AGPU] completes corporate restructuring (formerly POAI), enterprise-grade decentralized GPU computing power Aethir officially enters the mainstream market](https://upload.techflowpost.com/upload/images/20251212/2025121221124297058230.png?x-oss-process=image/resize,p_50/quality,q_80)
