The rumor hit my Telegram feed at 3 AM Vancouver time. Zhipu AI—the Beijing-based darling of China’s LLM race—had just completed a $40B secondary placement in Hong Kong. The crypto-native analysts I follow were silent. The equity-watchers shrugged. Then I read the filings. The new shares barely moved the needle on tradable float. The market absorbed the issuance like a sponge that was already saturated. As a governance architect who has watched DAO treasuries implode over similar misalignment of supply and demand, I recognized the pattern instantly: this was a liquidity trap, not a funding victory.
Zhipu is no small player. Backed by Alibaba, Tencent, and state-linked funds, it sits atop the “Big Six” of Chinese generative AI. Its GLM series claims parity with GPT-4 on several benchmarks. The company was last valued near $20B pre-money. A $40B raise—if fully subscribed at a premium—would have been a staggering endorsement. But the details tell a different story. The placement was structured as a top-up issuance to existing shares, priced at a discount to the last secondary trade. And the aftermarket volume? Thin as a DAO proposal quorum in a bear market. The primary dealers moved the paper, but the real buyers—institutional long-only funds, sovereign wealth—sat out.
Let’s dissect the mechanics. A $40B placement against a $200B implied market cap would normally add 20% dilution. But Zhipu’s tradable float before the deal was estimated at only $12B—meaning the new shares represented 3.3x the existing liquidity. “Barely moves the needle” in this context means the stock’s average daily volume over the past month was roughly $200M. To absorb $40B without price disruption, you need institutional accumulation over weeks, not days. The fact that the price barely budged signals that the new shares were placed almost entirely with “sticky” holders—perhaps strategic investors or family offices with lockup arrangements—rather than into the open market. This is the opposite of a liquid capital raise.
From where I sit—having co-founded a DAO that saw its treasury drained by a flawed multisig in 2017—this smells of governance malpractice. Zhipu’s board probably saw the placement as a necessary step to fund compute purchases for next-gen training. But the structure reveals a deeper misjudgment: they assumed demand for their equity would match the hype around their technology. The market proved otherwise. Code is law, but people are the soul. A technical model that performs beautifully in benchmarks doesn’t guarantee investor appetite for shares that can’t be easily traded.
The irony is thick. Crypto markets have been ridiculed for their volatility, yet they often provide more liquidity for assets of comparable size. A token with a $40B market cap on a DEX like Uniswap might see $500M in daily volume—2.5x the turnover of Zhipu’s float. Yes, the volatility is higher. But at least the exit path is clear. Traditional equity placements, especially in regions with less liquid secondary structures (Hong Kong is not New York), suffer from what I call “phantom liquidity”—the illusion of depth created by a few market makers that vanishes when real size hits the tape. This is why many Chinese tech companies have explored tokenization. A tokenized share can be traded 24/7 across borders with composable liquidity.
During my time building EquiSwap in the DeFi Summer of 2020, I learned a brutal lesson about liquidity assumptions. We designed a perfectly balanced pool, but when a flash loan attack drained one side, the entire AMM seized up. The code was sound; the economic assumptions were fragile. Zhipu’s placement is a similar failure of economic design. They assumed that a large issuance at a discount would attract eager buyers, but they ignored the fact that existing holders are often the only ones willing to absorb new supply—and only if they have confidence in the narrative. Right now, the narrative around Chinese AI is bifurcated. Domestically, it’s a patriotic imperative. Internationally, it’s fraught with regulatory risk and export controls. The global allocator community is cautious.
Decentralization is a verb, not a noun. It requires active participation. Zhipu’s attempted capital raise was a centralized decision made behind closed doors, and the market punished that opacity. In contrast, a DAO governance vote on a treasury rebalancing—even if contentious—at least surfaces the true demand signal. On-chain, you can see exactly who is buying and selling. Zhipu’s placement details are hidden in term sheets and NDAs. The market absorbs the news with a shrug because it has no way to price the underlying risk.
Now, the contrarian take: perhaps this is a nothingburger. Maybe the placement was never intended to be a true secondary offering but a strategic transfer of shares to a consortium of state-aligned entities that agreed to hold indefinitely. In that case, the “barely moves the needle” is a feature, not a bug. The liquidity is deliberately suppressed to prevent value extraction by foreign funds. China’s capital controls make free flow difficult anyway. But that interpretation only reinforces the centralization thesis. A company that can’t offer a credible exit to early investors will struggle to attract top talent. I’ve seen this in crypto projects that lock tokens for four years with no vesting cliffs—the best engineers leave for projects with better incentives.
I experienced this firsthand during the Winter of Value in 2022. My Canvas of Consensus NFT experiment lost funding when its treasury tokens collapsed. The community’s trust evaporated because we had no mechanism to unlock value without causing a price crash. Zhipu’s placement is the same story at a corporate scale. The liquidity is trapped, and so are the stakeholders.
What does this mean for the broader AI/crypto intersection? It reinforces my belief that future capital formation for AI projects will happen on-chain. The combination of tokenized equity, programmable vesting, and decentralized governance creates a feedback loop that aligns incentives better than any investment bank can. Zhipu could have issued a security token with automated market making and a bonding curve. Instead, they chose the old way—and got the old result: a lot of money, but no real liquidity.
The takeaway for founders and investors is simple: trust isn’t verified on-chain; it’s built off-chain. But once built, on-chain execution provides transparency that secondary markets crave. Zhipu’s next raise—if there is one—should be on Ethereum. Until then, consider this a cautionary tale. The hype cycle of AI has met the reality of capital constraints. And in the battle between narrative and liquidity, liquidity always wins.
As I write this, Zhipu’s stock is trading sideways. The order book is thin. The secondaries desks I called have no interest. The placement was a success by the numbers but a failure by the spirit. Decentralization is not just about technology; it’s about creating a market that works for everyone. Zhipu has a great model. But a great model without liquidity is just a beautiful prison.
