Zhiyang Protocol Raises 904M Yuan: A Traditional Infrastructure Playbook Meets Modular Blockchain

It starts with a number: 904 million yuan. That’s roughly $125 million at current exchange rates. For a company that was, until last week, a quiet player in China’s power grid digitization market, the announcement of a fundraise this size feels like a code injection into a legacy system. But the filing doesn’t mention Ethereum, validators, or staking. It talks about “multi-domain embodied intelligence” and “general-purpose AI perception terminals.” The tech stack is classical AI, not blockchain. Yet the structure of this capital raise mirrors exactly what I’ve seen in Layer 2 rollups: a well-defined allocation across R&D, infrastructure, and debt repayment, with a clear pivot from a mature core business into an adjacent high-growth frontier. The difference is that the frontier here is physical-world AI, not crypto. But the financial engineering is the same. And the market signals are something every blockchain analyst should watch. Zhiyang Innovation (Zhiyang) is a Chinese-listed company specializing in power grid intelligent operation and maintenance. Think transmission line monitoring, drone inspection platforms, and real-time fault detection. Its customer base is dominated by State Grid and China Southern Power Grid—two entities with deep pockets and low tolerance for downtime. The company’s historical revenue comes from selling hardware (sensors, cameras) and software (analytics platforms) to these utilities. The new fundraise, approved by the board on August 14, 2025, allocates the funds across four buckets: embodied intelligence R&D, general-purpose AI perception terminal industrialization, energy infrastructure supporting facilities, and repayment of interest-bearing debt. The catch is that the breakdown is intentionally vague, with a clause allowing the company to adjust the order and amount of each project “according to actual progress and capital needs.” This is a standard feature in Chinese capital markets, but it also means the company is buying itself flexibility to pivot if the AI or robotics market shifts. Let’s dig into the code-level logic of this allocation. The largest chunk likely goes to “multi-domain embodied intelligence and AI development.” In plain terms, this means building hardware-software systems that can perceive, reason, and act in physical environments—robots for power line inspection, substation patrol, and industrial safety. The technical challenge is not just training a large language model; it’s integrating computer vision, sensor fusion, and real-time control in a low-latency, high-reliability loop. Based on my experience auditing EigenLayer’s AVS specifications, I’d benchmark this against the security assumptions of off-chain compute. The power grid is a safety-critical system. A single misclassification in a vision model could cause a false alarm or, worse, a missed fault. The economic penalty for failure is not a slashed stake but a catastrophic outage. Zhiyang’s technical viability score for this project drops if the team lacks expertise in functional safety standards like IEC 61508 or ISO 13849. The filing does not disclose any such certifications. The second bucket—general-purpose AI perception terminals—is more concrete. These are edge devices that run inference locally, processing camera feeds at the substation level. The key metric is latency: from sensor input to actionable output, target sub-100 milliseconds. My simulations on Arbitrum Nitro’s WASM engine taught me that optimizing for deterministic execution in a hybrid environment is non-trivial. Zhiyang’s terminals will likely use specialized ASICs or FPGAs, not GPUs, to meet power and cost constraints. The energy infrastructure bucket is the most telling. Building dedicated power and cooling for AI compute clusters reveals that Zhiyang is planning to run its own training or inference pipelines, not rely solely on cloud APIs. This is a common pattern I’ve seen in enterprise blockchain rollups: the need for sovereignty over data and compute leads to on-premise infrastructure. The debt repayment portion is a red flag—it suggests the company’s balance sheet is under pressure, consistent with a capital-intensive pivot. Here is the contrarian angle: the market is framing this as a bullish AI play, but the real risk is not execution—it’s competition from blockchain-native solutions. Wait, that sounds counterintuitive. Let me explain. The power grid is a perfect candidate for decentralized physical infrastructure networks (DePIN). Projects like Helium, Hivemapper, and DIMO have proven that token incentives can bootstrap sensor networks faster than traditional capex. A grid operator could deploy a permissionless network of low-cost IoT nodes, each reporting voltage, temperature, and load data, secured by a decentralized validator set. The data would be immutable, auditable, and resistant to single-point failure. Zhiyang’s centralized approach, while faster to deploy in controlled environments, lacks the transparency and resilience of a blockchain-based system. Moreover, the embodied intelligence layer—robots patrolling substations—could be coordinated via smart contracts that automate payments for maintenance, penalize downtime, and reward accurate detection. The blind spot in Zhiyang’s plan is that it ignores the social consensus layer. Pure technical optimization cannot replace the trust guarantees of a decentralized protocol. If a State Grid executive asks, “Why should I pay for a robot that I cannot audit the code of?” the answer is not “read our whitepaper,” but “read the smart contract on a public chain.” Zhiyang’s entire value proposition relies on proprietary software and hardware. In a world where open-source DePIN protocols are maturing, that proprietary moat is vulnerable. From a competitive landscape perspective, Zhiyang is not competing with OpenAI or Baidu. It’s competing with decentralized sensor networks and open-source robot platforms. The power grid digitization market is a $10B+ opportunity, and the incumbents are all traditional industrial players. But the disruptors are not other AI companies—they are crypto-native protocols that offer lower cost, higher transparency, and global scalability. Zhiyang’s 904 million yuan raises its R&D budget, but it cannot outspend the collective development power of the Ethereum ecosystem. The typical DePIN project has a fractional budget compared to this, yet it achieves global coverage through token incentives. The financial efficiency of a token-incentivized network versus a centralized capital raise is an order of magnitude different. For example, a Helium hotspot costs $500 to deploy, and the network has over 1 million hotspots. That’s $500 million in distributed capex, but funded by individual miners, not a single balance sheet. Zhiyang’s approach is capital-intensive and slow. The risk is that by the time its robots are deployed, DePIN projects will have already captured the grid edge. Let’s talk about the investment implications. The market will reprice Zhiyang from a “power grid IT services” multiple (15x PE) to an “AI + robotics” multiple (30x PE) if the narrative holds. But the real test is whether the company can deliver revenue from the new products within 18 months. The debt repayment signals that cash flow is tight, and the R&D burn will accelerate. Given the lack of technical details in the filing, I assign a technical viability score of 4/10 for the embodied intelligence project. The missing pieces include: evidence of a functional prototype, partnerships with robot manufacturers, and a certified safety architecture. The greatest risk is that the company treats AI as a black box, paying lip service to “intelligence” without addressing the deterministic requirements of industrial control. Code is the only law that compiles without mercy. If the code is not open-source, the law is hidden. Takeaway: Zhiyang’s fundraise is a textbook case of a traditional company using capital markets to chase the AI narrative. The underlying logic is sound, but the execution risk is high. The vulnerability is that the same use case—power grid automation—is being attacked by crypto-native DePIN projects with lower cost structures and higher trust guarantees. Watch for the first demonstration: if Zhiyang shows a working robot patrolling a substation, the market will cheer. But the real signal is whether they open-source the perception algorithms. If they keep it closed, they are building a walled garden in a world that increasingly values open protocols. The smart money should track the DePIN counterparts, not the legacy pivot.