June 23, 2026, saw a sharp decline in the share prices of the two leading Hong Kong-listed AI large model companies, Zhipu (02513.HK) and MiniMax, which fell by over 9% and 13% respectively, dragging down the entire Hong Kong tech sector. Just one trading day prior, Zhipu’s intraday price had surged more than 42%, reaching a record high of HKD 2,980 per share and pushing its total market capitalization past HKD 1 trillion. This dramatic reversal within 24 hours was not an isolated price fluctuation but a concentrated global repricing of the sustainability of capital expenditures in the AI industry. The ripple effects of this turmoil are spreading from US tech giants to the Asia-Pacific markets and are ultimately impacting risk pricing logic in the crypto asset space.
Immediate Trigger for the Crash: How Nasdaq’s Drop and Talent Exodus Hit Hong Kong Tech Stocks
The sharp declines in Zhipu and MiniMax on June 23 were directly triggered by the overnight US market. On June 22 (US Eastern Time), the Nasdaq Index dropped 1.32%, with Google falling nearly 6% and Microsoft and Meta also weakening. The immediate cause for the plunge among US tech giants was Google’s loss of its second top AI researcher in a week—Nobel laureate John Jumper left to join Anthropic, following Gemini AI model co-lead Noam Shazeer’s earlier move to OpenAI. The ongoing exodus of core AI talent has directly fueled market concerns over Google’s AI strategy.
But talent loss is only the surface. The market’s deeper worry centers on a structural issue: just how high is the return on massive AI investments by these tech giants? Alphabet’s Q1 free cash flow dropped 47% year-on-year to $1.012 billion, while Amazon’s free cash flow plunged 95% over the past 12 months to $1.2 billion. As the world’s leading tech companies burn cash on data centers while seeing free cash flow deteriorate sharply, investors are re-examining a fundamental question—are these capital expenditures truly sustainable?
Panic quickly spread from the US to Asia-Pacific. On June 23, South Korea’s KOSPI index plummeted over 4%, triggering a programmatic circuit breaker; the Nikkei 225 fell, and SoftBank dropped more than 8%. Hong Kong’s AI large model sector came under simultaneous pressure, with growth tech stocks broadly sold off. As the "dual giants" of Hong Kong’s large model sector, Zhipu and MiniMax became the focal points for selling pressure.
HKD 1 Trillion Market Cap vs. RMB 724 Million Revenue: The Valuation-Fundamentals Gap
To understand why Zhipu collapsed so quickly after reaching a HKD 1 trillion market cap, we first need to look at its fundamentals. According to Zhipu’s 2025 annual report, the company posted annual revenue of RMB 724 million, up 131.9% year-on-year, but reported a net loss of RMB 4.718 billion, with an adjusted net loss of RMB 3.182 billion. With a market cap of HKD 1.07 trillion against RMB 724 million in annual revenue, its price-to-sales ratio stands between 1,300 and 1,600 times.
How does this stack up? For comparison, Anthropic’s primary market valuation is around $96.5 billion, with a price-to-sales ratio of about 20x; OpenAI is valued at $85.2 billion, with a price-to-sales ratio around 34x. A company with RMB 700 million in revenue and RMB 4.7 billion in losses boasting a trillion-yuan market cap clearly isn’t being valued on its current income statement, but rather on its perceived "rare ticket" status as "China’s AI infrastructure." The problem is, when the gap between valuation and fundamentals becomes extreme, any marginal change can trigger a violent repricing. The June 23 selloff was the market collectively reassessing this gap.
MiniMax faces similar structural contradictions. In 2025, MiniMax’s revenue was about $79.04 million, up 159% year-on-year, while its annual loss increased 302% to $1.87 billion. Both companies are in the classic early-stage high-growth phase of "trading losses for scale, trading time for market space." Yet capital market patience is not infinite—when global investors start to question the overall return efficiency of AI capital expenditures, these high-valuation, unprofitable companies are often the first to be hit.
Why Scarcity Premium Is Rapidly Fading
The core logic behind the previously sky-high valuations of Zhipu and MiniMax was the "scarcity premium." UBS analysts break this down into three elements: scarcity of listed global model companies, shares still under lockup, and low liquidity. But this logic is now unraveling on multiple fronts.
First, global AI IPO supply is expanding—Anthropic and OpenAI have both confidentially filed for IPOs, planning to go public within the year. Second, a major lockup expiration is looming in July. According to HKEX filings, Zhipu will see its first group of cornerstone investor shares unlocked on July 8, totaling 25.68 million shares, about 11.9% of its H-shares; currently, there are only about 11.74 million freely tradable shares on the market, meaning the float will instantly expand 2.2 times after the lockup ends. MiniMax faces an even larger unlock on July 9—HSBC estimates about 65% of its shares will become tradable in July.
If the market at the start of this year was chasing the scarcity of Chinese AI companies, the recent correction reflects a shift in thinking: investors are recalculating just how much time and capital these companies need to turn technological advantage into sustainable commercial returns. The scarcity premium is fading, replaced by a more cautious assessment of commercialization efficiency and profitability pathways.
Global AI Capital Expenditure: How Long Can the $700 Billion Spending Spree Last?
The crashes of Zhipu and MiniMax are not isolated incidents—they reflect growing doubts about the sustainability of global AI capital expenditures. Data shows that in 2026, Microsoft, Google, Amazon, and Meta are expected to spend a combined $700 billion on capital expenditures, up nearly 80% year-on-year, with the vast majority going to AI data centers, GPU clusters, networking equipment, and power infrastructure. Goldman Sachs forecasts that from 2025 to 2030, hyperscale cloud companies will spend a cumulative $5.3 trillion on AI and data center capex. Vanguard’s 2026 outlook projects that from early 2025 to the end of 2027, cumulative AI capex will reach $2.1 trillion.
This expansion is no longer solely funded by operating cash flow, but increasingly relies on external financing. Amazon recently secured a $17.5 billion syndicated loan, while Google launched its largest-ever equity financing plan, totaling $80 billion. Morgan Stanley predicts that AI-related bond issuance will exceed $570 billion in 2026. The four US hyperscalers are guiding for combined capex of $700–725 billion in 2026, likely surpassing $1.1 trillion in 2027; even if operating cash flow still exceeds $900 billion, free cash flow will systematically turn negative.
"AI is not the internet, it’s real estate"—this analogy is resonating with more and more investors. Tech giants are becoming the world’s most aggressive infrastructure builders, frantically raising funds, acquiring land, building data centers, and buying GPUs. The beauty of the internet business lies in near-zero marginal costs, but every additional AI user consumes more tokens. When scale effects in AI become a reverse force, market skepticism about the sustainability of capital expenditures gains a solid logical foundation.
From the AI Narrative Cycle to Crypto: The Common Logic of Capital Cycles
The boom and bust of Zhipu and MiniMax offer a valuable reference point for the crypto industry. The capital narrative around AI large models shares striking structural similarities with crypto’s bull and bear cycles—both rely on "future imagination" to drive valuations, and both experience the full cycle from "scarcity premium" to "liquidity release" to "value reassessment."
This is nothing new for crypto. The 2021 NFT boom, the 2022 DeFi valuation bubble, and the 2024 Bitcoin ETF anticipation rally all followed the same capital cycle: narrative-driven capital inflows → valuations detached from fundamentals → marginal changes trigger sell-offs → market repricing. Zhipu’s rise to a trillion-dollar market cap at a 1,300x price-to-sales ratio mirrors how crypto assets discover prices untethered from fundamentals during bull markets; the impending lockup expirations triggering sell-off expectations closely resemble the "whale unlock" price pressure in crypto markets.
A deeper connection lies in liquidity competition. Former BitMEX CEO Arthur Hayes once warned that the AI bubble is siphoning off the liquidity needed for Bitcoin’s next leg up. As global institutional investors allocate an ever-larger share of risk budgets to AI assets, the space left for crypto is being passively squeezed. In Q1 2026, global venture investment approached $300 billion, with AI companies alone attracting about $242 billion—roughly 80% of global VC flows. Capital is exiting crypto en masse and pouring into the AI narrative—should the AI narrative itself face sustainability doubts, the direction of capital reallocation will become a key variable for the crypto market to watch.
The Market Is Relearning How to Price AI
The crash in Zhipu and MiniMax is, at its core, the market establishing a new pricing anchor for AI assets. Over the past two years, AI sector valuations have hinged on an implicit assumption: that expanding capital expenditures would automatically translate into technological leadership and commercial returns. But the series of events in June 2026—from Google’s talent exodus to deteriorating free cash flow, from lockup pressure to synchronized global sell-offs—is forcing the market to reassess this assumption.
Microsoft CEO Satya Nadella recently remarked, "If models become cheaper and more interchangeable, investors may question whether these expenditures are building lasting advantage or simply adding profit pressure." The AI market is moving toward commoditization—when model capabilities converge and open-source models continue to catch up, the moats built by massive capex may not be as strong as imagined.
For the crypto industry, the doubts over the sustainability of AI capital expenditures offer a profound lesson: any asset class relying on "future imagination" must eventually shift from narrative-driven to fundamentals-driven valuation paradigms. When the tide goes out, the market always cares about the same question—can these investments deliver sustainable returns?
Conclusion
On June 23, 2026, Zhipu and MiniMax dropped over 9% and 13% respectively, directly triggered by a chain reaction of sell-offs among US tech giants due to talent loss and capex concerns. The deeper cause, however, is that the "scarcity premium" logic underpinning their valuations—hundreds or even thousands of times sales—is unraveling, with large-scale lockup expirations in July further intensifying selling pressure. Global AI capital expenditures have reached the $700 billion level, while free cash flow at leading tech firms is simultaneously deteriorating—the market is conducting a systemic reassessment of the sustainability of the "burn cash for growth" model. The impact of this reassessment extends beyond the AI sector, indirectly influencing the risk pricing logic of crypto assets through liquidity competition and shifting risk appetites.
FAQ
Q1: What is the core reason behind the sharp declines in Zhipu and MiniMax?
The direct cause is the overnight drop in the US Nasdaq and weakness in tech giants like Google due to talent loss, with panic spreading to Asia-Pacific markets. The deeper reason is a collective market reassessment of the sustainability of global AI capital expenditures, coupled with the looming pressure of large-scale lockup expirations for both companies in July.
Q2: How large is AI capital expenditure?
In 2026, Microsoft, Google, Amazon, and Meta are expected to spend a combined $700 billion on capex. Goldman Sachs forecasts that from 2025 to 2030, hyperscale companies will cumulatively spend $5.3 trillion on AI capex.
Q3: Why is Zhipu’s valuation raising concerns?
Zhipu’s 2025 revenue was RMB 724 million, with a net loss of RMB 4.718 billion. With a trillion-yuan market cap, its price-to-sales ratio is as high as 1,300–1,600x. When the gap between valuation and fundamentals is too wide, any marginal change can trigger a violent repricing.
Q4: What does this AI valuation adjustment mean for the crypto industry?
AI is absorbing about 80% of global venture capital, crowding out liquidity from the crypto market. If the AI narrative itself faces sustainability doubts, capital may seek new allocation targets, potentially benefiting crypto assets from this rebalancing.
Q5: What is the focus of the debate over the sustainability of AI capital expenditures?
The core issue is whether massive investments can generate commensurate returns. Leading tech companies’ free cash flow is deteriorating, and the commoditization trend in AI may further compress profit margins. The market is shifting from a "compute expansion" narrative to an "application and cash flow realization" evaluation framework.
