Hook

On August 14, 2025, a little-known Chinese power sector software firm—Zhiyang Innovation—filed a plan to raise no more than 904 million yuan (≈$125M) for what it calls “multi-domain embodied intelligence and AI development.” On the surface, this is a routine capital markets event: a traditional industrial IT company trying to pivot into AI. But for anyone tracking the crypto-AI narrative, this is a canary in the coal mine. The capital is moving from zero-rate bonds into real-world AI infrastructure, and the chain of custody for that value will eventually touch decentralized compute, tokenized asset markets, and on-chain agent economies.

I don’t believe in narratives that lack infrastructure backing. This raise is infrastructure backing—for a stack that crypto will need to bridge.

Context

Zhiyang Innovation is a mid-cap Chinese firm (estimated market cap $300-$700M) that historically provided power line monitoring and smart grid software. Its pivot to “embodied intelligence” (robots + AI that interact with physical environments) and “general AI perception terminals” mirrors a broader trend: traditional industrial companies are using cheap debt/equity to buy a ticket to the AI revolution. The crypto market, meanwhile, is stuck in a sideways grind. AI agent tokens (like FET, AGIX, and newer L1s for agents) have been range-bound, waiting for a catalyst. The narrative of “AI moving from cloud to edge” is well-known, but the capital required to build that edge infrastructure is only now being deployed.

Zhiyang’s plan allocates funds across four buckets: embodied intelligence R&D, generic AI perception terminal upgrades, energy facility supporting infrastructure, and debt repayment. The first three are directly relevant to crypto’s thesis of decentralized physical infrastructure networks (DePIN) and AI compute marketplaces. The last signals that even traditional firms are optimizing their balance sheets for a higher-for-longer rate environment—a macro backdrop that crypto native projects must also navigate.

Core

Now, let’s dissect the implications through the lens of blockchain narrative mechanics.

Narrative Mechanism #1: Capital Allocation as Validation

When a non-crypto company raises $125M for embodied AI, it validates the thesis that “AI agents will need to interact with the physical world”—a core premise for projects like Render Network (compute for AI rendering), Akash Network (decentralized compute), and Hivemapper (physical world mapping). More importantly, it signals that the demand for perception hardware (cameras, LiDAR, sensors) and on-device AI inference is real. These hardware feeds will generate data that needs to be verified, stored, and monetized on-chain. I’ve been tracking this since my 2022 deep-dive on Celestia’s data availability sampling: modular blockchains are the settlement layer for data from millions of AI agents.

Zhiyang’s “general AI perception terminals” are exactly the kind of edge devices that will produce provenance-required data for DePIN networks. Every detection of a power line fault, every robot movement, every sensor reading—if this data is to be used for insurance, carbon credits, or automated settlements, it needs an immutable record. The company’s plan to build proprietary analysis platforms on top of these terminals suggests they intend to keep the data siloed. But that’s where the crypto twist comes: siloed data creates arbitrage opportunities for middleware that bridges to public blockchains.

Narrative Mechanism #2: The “Multi-Domain” Trap and Opportunity

The announcement uses “multi-domain” repeatedly. This is a classic signal of scope creep—a company that knows its single vertical (power) is capped and is grasping for adjacent markets. Crypto’s AI narrative has a similar problem: too many projects claim to be “multi-domain” without a clear integration path. The contrarian reality is that specialization wins. Zhiyang will likely succeed in power industry AI first, then struggle in transportation and manufacturing. For crypto, this means the most valuable AI infrastructure will be sector-specific DePINs (e.g., energy-specific compute, manufacturing-specific oracle networks), not general-purpose platforms.

Based on my experience auditing tokenomics for 20+ DePIN projects, I’ve seen the same pattern: generalists fail to capture liquidity; specialists build moats. The Zhiyang raise confirms that capital is flowing into vertical AI stacks—which is bullish for crypto projects that can provide the financial layer for those stacks (tokenized revenue streams, cross-chain settlement, AI agent wallets).

Narrative Mechanism #3: Debt as a Crypto Narrative Accelerator

Zhiyang’s plan includes repaying “interest-bearing debt.” This is a subtle but critical detail. The company is using the capital raise to deleverage before entering a high-risk R&D phase. In crypto, we’ve seen the opposite: protocols borrow to fund liquidity mining, then get wrecked when token prices fall. The traditional approach—clean up the balance sheet first—is what institutional investors want to see. This sets a precedent: future AI-related token raises should include a debt repayment component to signal financial discipline. I expect the next wave of crypto AI projects (especially those tokenizing real-world AI assets) to mimic this structure.

Contrarian

Here’s the counter-intuitive angle: the Zhiyang raise is actually bearish for the “AI on chain” narrative in the short term.

Why? Because $125M is going to a centralized, regulated entity that will build proprietary AI stacks. That capital could have gone to decentralized compute networks or tokenized AI agent platforms. Instead, it’s reinforcing the traditional model: one company owns the hardware, the software, and the data. This delays the “decentralized AI” thesis by proving that the easiest path to market is still centralized vertical integration.

The blind spot in most crypto AI analysis is assuming that traditional companies will need to use blockchains. They won’t—until they face a trust or coordination problem they can’t solve internally. Zhiyang’s “multi-domain” strategy is internally focused; they will build their own perception terminals, their own AI models, their own energy infrastructure. The incentives to open up to a permissionless network are weak. Crypto’s value proposition only becomes urgent when these closed systems hit a bottleneck: interoperability between different vendors’ terminals, data provenance for regulatory compliance, or cross-company machine-to-machine payments.

I’ve seen this playbook before. In 2022, modular blockchain advocates claimed everyone would need Celestia’s data availability. The reality was that most L2s built their own. The same will happen in AI: most companies will build their own stacks until they can’t. The contrarian bet is on the middleware that solves the integration problem once the vertical silos become too many to manage.

Takeaway

Zhiyang’s 904M yuan raise is a microcosm of the next market cycle: traditional industrial capital is flowing into AI infrastructure, and crypto’s role is to provide the settlement layer for the inevitable interoperability crisis. The narrative isn’t about whether AI will be decentralized—it’s about when the centralized stacks will hit a coordination wall. When that happens, the projects that have built the bridges (DePIN middleware, tokenized compute, agent-to-agent settlement protocols) will capture the value.

Follow the structure, not the hype. The capital is moving, and the infrastructure for that movement is being built—both in Shenzhen and on-chain. The question is which side you’re betting on when the two meet.