"150 billion yuan in intention orders." That's the number plastered across headlines from a blockchain-centric news outlet this week, touting Yuegangwan Intelligent Computing's H1 2026 performance. But after spending years mapping liquidity flows—from 2017 ICOs to DeFi summer to the LUNA collapse—I've learned that numbers this round demand a second look. Intention orders are not contracts. And in a market where capital chases narrative harder than fundamentals, this announcement smells more like a liquidity trap than a breakthrough.
Let's start with the basics. Yuegangwan Intelligent Computing, an opaque player in the AI cloud services space, released data claiming aggregate computing power intention orders of 35,000 PFLOPS (FP16), valued at over 150 billion yuan. They report 20 billion yuan already delivered—roughly 6,000 PFLOPS. That leaves 130 billion yuan and 29,000 PFLOPS still in the pipeline. The source? A Web3/blockchain news site, not a traditional financial wire. Based on my experience auditing token distribution patterns during the ICO craze, I know that when companies announce such figures through fringe channels, there's often a deliberate reason—easier narrative control, less scrutiny.
The company's background is murky. No mention of GPU vendor, chip supply agreements, or even a clear business model. The implied price per PFLOPS sits at roughly 42.9 million yuan per PFLOPS—a number that demands cross-referencing. Public cloud providers like Alibaba Cloud charge roughly 300,000–500,000 yuan per PFLOPS per year for H100-equivalent instances. If Yuegangwan's contracts span one year, this pricing is at the high end; if they span three years, it's remarkably cheap. Without service term disclosure, we're flying blind. But the blockchain context hints at a possible Web3 cross—perhaps the compute is intended for crypto mining or AI-enhanced on-chain oracles, a sector I've tracked since 2020 when I reverse-engineered Curve's liquidity pools.
But the real story isn't the headline—it's the gap. The conversion rate from intention to delivery is a paltry 13.3%. In crypto terms, that's worse than most DeFi protocol's total value locked (TVL) growth during a bull run. It's a classic signal: the company is using future promises to prop up current valuation. I built this analysis tool during the 2017 ICO gold rush—a Python script that tracked token distribution and vested schedules across 50+ projects. The lesson then was the same now: when delivery lags promise by 87%, you're not looking at growth; you're looking at a liquidity deferral machine.
Now let's get into the hardware. 35,000 PFLOPS FP16 is roughly 17,500 NVIDIA H100 GPUs (at 2 PFLOPS each) or about 8,750 B200s. Given U.S. export controls on advanced chips to China, sourcing that volume is almost impossible without a direct government-level deal or reliance on domestic alternatives like Huawei Ascend 910B. But Huawei's ecosystem is immature—I've seen the compatibility issues firsthand during my project integrating on-chain settlement for cross-border payments. A switch to Ascend would degrade training performance and force customers to migrate frameworks, a non-trivial cost that kills the “compute-on-demand” value proposition.
Consider the capital expenditure. At a conservative $30,000 per H100 (black market prices are higher), 17,500 units cost $525 million (roughly 3.8 billion yuan). But that's just the GPUs. Add networking (InfiniBand switches ~$50M), data center construction (I priced similar builds in Warsaw: ~$100M for 20 MW), cooling, power—total upfront could exceed 7 billion yuan. Yet the delivered portion is only 20 billion yuan revenue? That implies a high margin if they collected full payment, but more likely it's pre-payments or short-term rentals. The unit economics don't add up.
If the average service contract is three years, annual revenue from the full 150 billion yuan would be 50 billion yuan. For a company with 7 billion in upfront capex, that's a 7x revenue multiple on the cost—plausible if they have low operating expenses. But the risk is all front-loaded: they need to secure the chips, build the data center, and hope the intention orders convert. If conversion stays at 13%, all that capex becomes idle capacity. That's a liquidity crisis waiting to happen. I saw the same pattern during the LUNA collapse—a thesis I published in 20 pages arguing that algorithmic stablecoins were liquidity crises masquerading as tech failures. The same dynamic applies here: a stack of future obligations that vaporize when the music stops.
Here's the contrarian take: This announcement is not about delivering compute; it's about raising capital. The company is using these 150 billion intention orders as collateral for investors. They're selling a vision of AI compute scarcity—a bull market narrative. But anyone who remembers the 2022 contagion—Celsius, Three Arrows Capital—knows that quasi-collateralized debt instruments blow up when liquidity dries up and counterparty risk materializes. Compare this to the stablecoin yield products I've warned about: sUSDe at Ethena, which is built on maturity mismatch. Intention orders are the same—they promise future compute delivery backed by customer commitments that may not materialize. When the bull market pauses, these “commitments” will be the first to evaporate.
Another rug? No, just a liquidity trap. The 20 billion yuan delivered is the anchor, but the remaining 130 billion is a risk stack that only works if capital keeps flowing. The blockchain news source is another clue. In my experience auditing crypto projects, such distribution is a red flag. Legitimate infrastructure companies use Bloomberg, Reuters, or at least TechCrunch. The Web3 channel suggests a targeted audience: crypto-native investors who are more tolerant of high-risk narratives. This is a sophisticated marketing play to attract speculative capital—not a signal of operational reality.
Beyond the numbers, there's a regulatory dimension. China's export controls on GPUs and its strict data localization laws will shape whether Yuegangwan can even import the necessary chips. If they rely on domestic alternatives, they face a performance and adoption gap. If they rely on smuggled or grey-market hardware, they risk seizure. The compliance integration in this case is non-trivial—and the company's silence on it speaks volumes.
Liquidity doesn't care about your intention orders. In the current bull market, euphoria masks technical flaws. Yuegangwan's announcement is a test: will investors see through the numbers, or will they chase the narrative? Based on my macro-causal framework, the risk-reward is skewed to the downside. The only way this plays out positively is if the company has ironclad supply chain agreements and a conversion rate exceeding 50%—both unlikely given today's geopolitical constraints.
My advice: treat this as a liquidity trap. The real opportunity lies in verifying on-chain infrastructure—not press releases. When the next downturn hits, the 130 billion yuan gap will become everyone's problem. For now, watch the delivery metrics and ignore the headlines. The market will eventually ask: where's the compute?