Hook
In a market starved for liquidity, one would expect a $4 billion share placement to send ripples across exchanges. But when Zhipu AI, one of China’s leading AI labs, attempted to place new shares in Hong Kong, the market barely blinked. According to a detailed report by Crypto Briefing, the placement barely moved the needle on tradable shares, exposing a stark liquidity challenge that may foreshadow a broader crisis in AI startup valuations. For those of us watching from the macro side, this is not a whisper—it is a warning siren.
Context
Zhipu AI sits among the “Six Little Tigers” of Chinese artificial intelligence, a group that includes Baichuan, Moonshot AI, Minimax, 01.AI, and Stepfun. These companies have collectively raised billions from sovereign funds, tech giants, and venture capital, riding the wave of China’s push for AI sovereignty. Zhipu, in particular, has been lauded for its GLM series of models, which compete directly with OpenAI’s GPT. The company’s valuation has been pegged at over $10 billion in private markets. However, the secondary market tells a different story. The attempted placement—a new share issuance designed to raise capital from institutional investors—was meant to provide a liquidity window for early backers and employees. Instead, it revealed that demand for Zhipu’s equity is far weaker than its headline valuation suggests.
The Crypto Briefing report highlights that the placement “barely moved the needle” on the total tradable shares. This is a technical term worth dissecting. In secondary markets, the ability to absorb a large block of shares without significant price impact is a measure of depth. When a $4 billion block fails to shift the bid-ask spread or volume profile, it implies one of two things: either the placement was done at a steep discount to the prevailing market price, or the buyers were insiders and the shares will not trade freely. Both scenarios point to a lack of genuine external demand.
Core
Let’s go deeper into the numbers. Assume Zhipu’s total tradable shares—the float—are equivalent to roughly 10% of the fully diluted valuation, a common structure for late-stage private unicorns. At a $10 billion valuation, the float would be worth $1 billion. A $4 billion placement would then represent four times the entire float. In normal market conditions, such an issuance would crush the price, unless the placement is pre-sold to a small group of buyers who agree not to dump immediately. The fact that the market barely reacted suggests the placement was likely executed at a significant discount—perhaps 30–50% below the last private round valuation—and the buyers are long-term holders or related parties.
Based on my experience as a digital asset fund manager, I’ve seen this pattern repeatedly in crypto token unlocks. Projects announce large sales to “strategic investors,” but the real signal is the absence of natural buyers. When I modeled liquidity for Bitcoin ETF anticipation in 2024, I learned that volume depth is more important than price. A stock that trades $10 million a day cannot absorb a $4 billion block without cascading sell pressure. Zhipu’s placement is essentially a hidden down round—the company is raising capital at a lower implied valuation, but it’s masked by the opaque nature of private placements.
The implication for investors is clear: Zhipu’s fair value is likely 30–50% lower than its last private round. This is not just a company-level issue. It reflects a systemic problem across Chinese AI startups. The “liquidity illusion” arises when private markets sustain high valuations based on narrative and fear of missing out, while secondary markets—where real money changes hands—demonstrate a refusal to pay those prices. The bust I witnessed in 2022 during the crypto winter taught me that pruning is necessary. The same mechanism is at work here.
Furthermore, the timing is critical. China’s AI companies face dual pressures: the need to purchase expensive domestic chips (like Huawei’s Ascend) due to export controls, and the requirement to comply with increasing regulatory oversight on model safety. These costs erode margins and push cash burn rates higher. A failed or weak placement signals that the public market is unwilling to fund this burn. The result? Talent exodus. When employee stock options cannot be liquidated, the best engineers leave for better liquidity elsewhere—often at hyperscalers like ByteDance or even at US firms like OpenAI, who offer instant liquidity through secondary programs.
Contrarian
Now the contrarian angle. Perhaps the market’s silence is not weakness but wisdom. A placement that barely moves the needle could be interpreted as a disciplined capital raise—Zhipu may have deliberately targeted a small group of blue-chip investors who share a long-term vision, avoiding the noise of a public offering. In this view, the low liquidity is a feature, not a bug. The company is preserving its valuation by not forcing shares onto a shallow market. Additionally, China’s capital controls and the geopolitical chill with the West mean many large funds are simply not allowed to buy Chinese AI equities. The muted reaction may be more about structural barriers than lack of confidence.
But I disagree. Having analyzed capital flows across crypto and traditional markets, I’ve learned that structural excuses often mask fundamental disinterest. A true long-term investor would create market depth, not absorb a discounted block. The fact that the placement didn’t move the needle suggests the buyers are passive holders, not price-makers. This is a bad signal for future rounds. My eye is on the horizon, not the hourly candle. The bust was not an end, but a necessary pruning.
Takeaway
Zhipu AI’s $4 billion placement is more than a single data point—it is a macro signal for every late-stage AI unicorn. When liquidity dries up at the top, the entire pyramid shifts. For those of us navigating this cycle, the lesson is simple: the ability to raise capital quietly may feel like a victory, but it is often a prelude to a down round. Winter clears the weak hands, and the pruning necessary for the next cycle is underway. Watch the volume, ignore the noise.
