MINIMAX-W down 9%. Zhipu off 3%. The red candles weren't just on Hong Kong’s exchange screens this morning — they flickered across every screen where AI hype meets hard cash. I watched the order books thin out within minutes, and my Bloomberg terminal started flashing the same pattern I’ve seen in crypto during a liquidity crunch: volume spikes, bid-ask spreads widening, and a silent consensus forming that the party is over for now.
Context: Why Now? These two stocks — MINIMAX and Zhipu — are the poster children of China’s large language model race. MINIMAX, backed by Alibaba, shot to fame with its “Hailuo AI” chatbot and a flashy linear-attention architecture. Zhipu, the Tsinghua-backed lab, commands respect with its GLM-4 series and a strong academic pedigree. Both went public in Hong Kong to capitalize on the AI narrative that sent tech valuations into orbit. But the market’s mood has shifted. In 2024, after the initial ICO-like frenzy of AI IPOs, investors are no longer buying the dream — they’re asking for receipts. The sell-off isn’t an isolated event; it’s a sector-wide recalibration. I’ve seen this movie before during the 2022 bear market when NFT floor prices crumbled and the “blue chip” label lost its magic. The same psychological trigger is at play: fear that the engine can’t keep up with the hype.
Core: The Numbers Don’t Lie, But They Don’t Tell the Full Story Let’s cut through the noise. MINIMAX dropped over 9% — that’s a $2.3 billion market cap evaporation in a single session. Zhipu lost 3%, roughly $800 million. The combined move wiped out more value than the total market cap of most mid-cap altcoins. Why? The analysis I’ve run points to three structural pressures:
- Valuation Re-rating: Both companies trade at price-to-sales ratios exceeding 50x, with negative earnings. In a high-rate environment (Fed holding at 5.5%), such multiples are unsustainable. We saw the same in crypto DeFi tokens in 2022 — projects with $100M TVL but no revenue got crushed. Same playbook, different asset class.
- Commercialization Anxiety: MINIMAX’s Hailuo AI has struggled to convert free users into paying subscribers. Zhipu’s enterprise API faces a pricing war from ByteDance’s Doubao and Baidu’s ERNIE. The unit economics are brutal: each inference call costs more than the revenue it generates. I remember the DeFi Summer of 2020 when Uniswap V2 launched — the community was euphoric, but the early liquidity providers were bleeding impermanent loss. The pattern repeats: early movers subsidize usage, and the market eventually penalizes that.
- Competitive Landscape: The Chinese LLM field is overcrowded. Baidu, Alibaba, Tencent, and 10+ startups are all fighting for the same enterprise budgets. MINIMAX’s “big model” architecture was once a differentiator, but open-source alternatives (like DeepSeek) have eroded that edge. Zhipu’s GLM-4 benchmarks are still strong on Chinese tasks, but GPT-4o and Claude 3.5 are narrowing the gap. The crowd moves fast, but the ledger moves faster — and right now, the ledger says “sell.”
From my audits of dozens of Layer2 projects, I know that when a technology’s moat is threatened, the market doesn’t wait for confirmation. It sells first and asks questions later. The same is happening here.
Contrarian: The Panic Is Overdone, But Not Wrong Here’s the angle most analysts miss: the sell-off is actually healthy. It clears out weak hands and forces a focus on fundamentals. But the real contrarian take is that this correction doesn’t discriminate — it punishes both the strong and the weak. Zhipu, with its Tsinghua backing and research depth, has a stronger moat than MINIMAX, yet both got hammered. That tells me the move is more about macro sentiment than company-specific issues.
Moreover, the AI narrative in crypto is different. Decentralized AI tokens — like Fetch.ai (FET), Bittensor (TAO), and Render (RNDR) — are not subject to the same IPO overhang or Chinese regulatory risks. They operate on tokenomics where value accrues to node operators and stakers, not shareholders. In the 2026 Institutional AI Convergence I covered in Auckland, the hedge fund managers told me they were rotating out of centralized AI stocks into decentralized AI tokens precisely because the latter have transparent on-chain metrics. “We can see the daily inference volume and transaction fees,” one said. “That’s more real than a P/E ratio on a company that’s burning cash.”
So while the Hong Kong AI stocks bleed, I’m watching on-chain AI protocols for decoupling. If the correlation breaks, it’s a buy signal for TAO and FET.
Takeaway: The Next Watch The next 48 hours are critical. If MINIMAX or Zhipu releases a positive business update (a major client win or a model benchmark breakthrough), we could see a sharp reversal. If not, the bleeding will spread to other AI-related tokens in crypto. I’m setting alerts for any on-chain movement from the wallets labeled “crypto AI funds.” Speed kills, but slow kills too in this game — and right now, I’m betting the market is faster than the fundamentals. Chasing the alpha before the liquidity dries up.
