The ghost in the machine is not a codebase, but a balance sheet. On August 14, 2025, Zhiyang Innovation, a Chinese power sector informatics firm, filed a prospectus to raise up to 904 million yuan (approximately $126 million) for multi-domain embodied intelligence, AI development, smart perception terminals, and energy facility support. The announcement, buried in a sea of A-share capital market filings, would normally escape the attention of the crypto observer. But under the lens of macro-liquidity and the battle for computational primacy, this single event reveals a structural shift: traditional industrial capital is now aggressively crowding into the same territory that crypto-native AI and DePIN projects have been quietly cultivating for years.
Tracing the liquidity ghost in the machine, one finds a pattern. Zhiyang is not a blockchain company. Its core business, inferred from the filing's language, lies in power grid intelligent operation and transmission line monitoring—a classic “digital twin” of physical infrastructure. Yet the announced capital allocation breaks down into four pillars: embodied AI (long-term exploration), general-purpose AI perception terminals (mid-term monetization), energy facility support (short-term fundamentals), and debt repayment. This tripartite structure mirrors the very stack that DePIN projects like Render Network, Bittensor, and Akash Network are trying to build: a decentralized, verifiable layer for AI inference, edge sensing, and energy-efficient compute. The difference is that Zhiyang is doing it within the Chinese regulatory framework, using A-share equity, while crypto projects rely on token incentives and global node operators. The convergence of destination is striking.
Context: The Protocol That Is Not a Protocol
To understand the significance, we must map the current landscape of AI capital deployment. Since 2023, the narrative around AI has bifurcated. On one side, centralized AI labs (OpenAI, Google, Baidu) continue to consume billions in GPU clusters and data centers. On the other side, a growing cohort of crypto-native projects argues that the future of AI must be permissionless, privacy-preserving, and verifiable—hence the rise of on-chain inference markets, decentralized training protocols, and zero-knowledge machine learning. The conflict between these two paradigms is not merely technical; it is a war for capital allocation. Zhiyang's raise represents a third vector: traditional industrial companies that see AI not as a standalone product but as an upgrade to their existing physical asset management systems. These companies bring deep domain knowledge, existing customer relationships, and—crucially—access to cheap equity capital. They do not need to issue tokens or bootstrap a community. They simply pivot.
Core: The DePIN Parallel and the Liquidity Drain
Let us quantify the overlap. Zhiyang's “general-purpose AI perception terminal” is, in essence, a hardware device that collects multi-modal data (visual, acoustic, thermal) from power substations, transmission lines, and other industrial assets. This data is then processed by an AI inference engine—likely running on their own edge compute—and fed into a management platform. The terminal is general-purpose, meaning it can be repurposed for smart cities, traffic management, or even agricultural monitoring. Now, consider the DePIN thesis: a network of physical devices (cameras, sensors, IoT gateways) that collectively provide data to a decentralized protocol, with token rewards for operators. The technical architecture is nearly identical. The only difference is the trust layer. Zhiyang relies on a centralized backend and corporate reputation; DePIN projects rely on cryptographic verification and token economics. The capital raise of $126 million, if deployed over 12-18 months, could fund the deployment of tens of thousands of such terminals. In the DePIN world, achieving that scale would require a token market cap in the billions and years of community building. Zhiyang can do it through a single board resolution.
This is where the “liquidity fragmentation” narrative becomes dangerous. As I argued in 2023 during my work on CBDC privacy layers, liquidity fragmentation is not a technical problem but a manufactured narrative used by VCs to push new products. The real fragmentation is between the capital markets that fund centralized AI and the token markets that fund decentralized AI. Zhiyang's raise is a canary in the coal mine: if traditional industrial companies begin to absorb the hardware and compute capacity that DePIN projects rely on, the cost basis for decentralized nodes will rise. The ETF wave washed away the retail tide, but what happens when institutional capital bypasses crypto entirely and builds the same infrastructure in the regulated world?
Contrarian: The Decoupling Thesis That Fails the Stress Test
A common defense among crypto-AI maximalists is that “decentralized AI will win because it respects user sovereignty and data privacy.” This is a noble sentiment, but it ignores the reality of capital efficiency. Zhiyang's terminals, for example, will be deployed in Chinese state-owned power grids. The data will be processed on-premises or in government-approved clouds. Privacy is not a feature; it is a regulatory requirement. The EU's MiCA and China's data security laws are already pushing centralized AI solutions toward on-device processing and federated learning—precisely the same architectural patterns that crypto projects advocate, but without the token layer. The irony is that the most privacy-preserving AI deployments may come from traditional industrial firms, not from crypto startups, because they have the resources to implement zero-knowledge proofs at scale without the overhead of a blockchain consensus layer. Privacy eroded not by code, but by consensus—and the consensus here is that capital, not ideology, determines the deployment speed.
We sleepwalk into a digital panopticon, but the panopticon is built with corners cut by efficiency. The merge was a fever dream for liquidity, but the reality is that the physical world is being wired with AI sensors by companies like Zhiyang, not by decentralized node operators. The crypto narrative of “everyone can be a validator” is beautiful, but it cannot compete with the capital velocity of a $126 million equity raise from a single board meeting. The contrarian angle is not that crypto-AI will fail, but that it will be marginalized to niches where decentralization is a regulatory necessity (e.g., cross-border payments, censorship-resistant inference) while the bulk of industrial AI will be captured by traditional firms that have simply upgraded their balance sheets.
Takeaway: Positioning for the Next Cycle
For the macro watcher, Zhiyang's raise is a signal to revisit the correlation between traditional industrial capital expenditure and crypto-native AI token valuations. If the S&P 500's AI CapEx continues to grow at 30%+ YoY, the pool of available compute and talent for decentralized projects may shrink. The smart play is not to bet against centralized AI, but to identify the plumbing that both sides must share—specifically, the oracle networks, verifiable random functions, and zero-knowledge coprocessors that will bridge the gap between Zhiyang's terminals and the on-chain world. History rhymes in the ledger; the ledger is being written by those who control the terminals. The question is not whether crypto will survive, but whether it will be the operating system or just a dusty archive.
As I wrote in my 2024 white paper on G20 liquidity correlation, the next cycle will be defined not by price action but by infrastructure ownership. Zhiyang's $126 million is a small drop in the ocean of global AI investment, but it is a drop that lands squarely in the territory where DePIN dreams live. The liquidity ghost is now wearing a corporate suit. The question is whether the crypto ecosystem can build a better suit—or if it will be left naked when the physical world turns on its sensors.
