Multiple AI unicorns, including OpenAI, Anthropic, and SpaceX, are pushing for IPOs with valuations reaching trillions of dollars, despite their profitability remaining unproven. Their valuations significantly diverge from traditional cash flow models, leaning more towards a "faith-based pricing" model that bets on AI's potential to reshape software, research and development, and space infrastructure. However, upcoming share lock-up expiries and massive IPO fundraising in July will test market liquidity and valuations, marking the first major stress test for AI assets.
The first half of 2026 is witnessing a rare wave of public listings for global tech companies. SpaceX is set to list on Nasdaq on June 12 with a valuation of $1.77 trillion, potentially the largest stock offering in history. Anthropic, valued at $965 billion, has confidentially filed for an IPO. On the Hong Kong exchange, Zhipu AI and MiniMax have seen their share prices surge post-listing, with Zhipu's market cap at one point exceeding twice that of JD.com, far surpassing Baidu and Meituan.
Most of these companies have yet to demonstrate profitability in the traditional sense. Zhipu's 2025 revenue was only 724 million yuan, and it remains unprofitable. SpaceX reported 2025 revenue of $18.7 billion with a net loss of $4.9 billion. While Anthropic's Annual Recurring Revenue (ARR) has surpassed $10 billion, its valuation is nearing $1 trillion.
What Drives These Lofty Valuations?
First, AI is seen as a new technological revolution, the largest platform shift since the internet. The consensus among investors is that trillion-dollar giants will emerge from these companies, with the potential for industry disruption appearing limitless. Second, there is a scarcity of investable assets. "Pure-play AI unicorns" represent the most compelling capital narrative globally in 2026. Zhipu issued only about 37 million H-shares, resulting in a tiny float. SpaceX's offering represents just 4.2% of its total shares. Extremely limited share supply, coupled with surging investor demand, has pushed prices to levels that defy traditional valuation models.
Basis for the Valuations
Zhipu AI and MiniMax both listed on the Hong Kong Stock Exchange in January 2026. At their peak, the market cap gap between them exceeded HK$400 billion. For Zhipu, with a market cap around HK$500 billion, what market is it anchoring to?
If measured by AI coding, China's total corporate R&D expenditure is about 3.4 trillion yuan. Even optimistically assuming 30%-40% is software-related and 70% of that will be handled by AI in the future, the maximum addressable market is only about 700 billion to 1 trillion yuan. At approximately 460 billion yuan, Zhipu's market cap alone represents 46%-66% of this hypothetical maximum for the entire Chinese AI coding replacement market.
If measured by SaaS replacement, the annual revenue of the US SaaS application layer is about $200 billion, roughly 1.4 to 1.5 trillion yuan. Zhipu's current market cap is about $64 billion, equivalent to about 30% of the US SaaS annual revenue. Moreover, China's SaaS market lacks the scale and payment habits of the US market.
For multimodal content creation, AI video generation primarily replaces production costs rather than creating a new market of equivalent size. A more reasonable estimate might focus only on the portion of content production costs compressible by AI, likely in the range of 200-300 billion yuan.
Combining these three optimistic scenarios—China's AI coding replacement market (700bn-1trn yuan), the theoretical US SaaS replacement ceiling (1.5trn yuan), and content production cost savings (200-300bn yuan)—yields a total pool of about 2.4 to 2.8 trillion yuan. Even under this maximum assumption, Zhipu's market cap would still account for 16%-19% of the total pool, not considering that many AI companies are competing for the same budget.
Thus, the market is pricing in a complete rewrite of R&D, software, and content production budgets by the AI application layer in advance. This is where the term "AI faith" is intriguing. From the perspective of replacing existing markets, this valuation is hard to explain with arithmetic alone. But valuation itself is not static arithmetic; it's a bet on the future. It only indicates it's expensive now, not that it will necessarily be expensive in the future.
Across the ocean, SpaceX, the earliest to rush for an IPO, has Starlink as its foundation with 10.3 million users, $11.3 billion in 2025 revenue, and a 39% profit margin. Yet, its $1.77 trillion valuation implies a price-to-sales ratio of nearly 100x. An aerospace analyst said, "The market's expectation for SpaceX is not just about providing computing power from space. It's about providing the infrastructure for the future of space. How to price infrastructure is a matter of imagination, with three core elements: the value of logistics, the value of infrastructure, and the value of resources."
Anthropic's faith is based on its rapidly growing ARR, which surged from $9 billion at the start of the year to $44 billion by May, doubling in two months. This is because it has captured the AI coding segment, with investors viewing it as an infrastructure company seizing the developer entry point and the future high ground of safe AI, warranting platform-level valuation multiples.
OpenAI is the giant unicorn at the center of the generative AI wave, with an $852 billion valuation, 900 million weekly active ChatGPT users, and the most complete product portfolio. However, lacking further imagination for growth at the moment makes it difficult for the capital market to grant it even higher "faith."
The Formula Behind the "AI Faith"
The term "faith" might sound derisive, but market faith is never blind. It has a formula, a transmission path, and internal logic.
Just two days before SpaceX's roadshow began, on June 2, the Forbes Real-Time Billionaires List showed Softbank Group Corp founder Masayoshi Son's wealth climbing to $100.7 billion, reclaiming the title of Asia's richest person after over two decades. On the same day, SoftBank's market cap soared to 49.3 trillion yen, officially surpassing Toyota Motor, ending Toyota's two-decade reign as Japan's most valuable company.
Son's "revenge" path is the best lens to understand this AI faith-based pricing. At the end of 2022, he was seen as a "loser." The bleeding from Vision Fund I's failures in WeWork, Uber, and other sharing economy companies continued, and SoftBank's market cap had shrunk to just over $60 billion. However, during two years of silence, he executed a complete strategic pivot.
In 2016, Son invested $31 billion to acquire Arm. In 2023, he spent another $16.1 billion to buy back a 25% stake from the Vision Fund, bringing the total investment to $47.1 billion. By 2026, the explosion in server CPU demand from AI Agents, combined with the energy efficiency advantages of the Arm architecture, sent Arm's stock price soaring 268% year-to-date. Based on SoftBank's nearly 90% stake, the paper profit exceeded $320 billion by early June.
Above the foundational layer is the application layer. SoftBank has invested over $64 billion cumulatively in OpenAI, securing about a 13% stake, making it the second-largest external investor after Microsoft. As of March 31, 2026, this investment had generated approximately $45 billion in recognized investment gains.
Further up is computing infrastructure. For the US Stargate project, SoftBank partnered with OpenAI and Oracle on a total investment of $500 billion. For a French data center plan, the commitment is 75 billion euros, with the first phase of 45 billion euros to build 3.1 gigawatts of computing power by 2031. In between, SoftBank acquired Ampere Computing in 2025 for $6.5 billion, completing the final piece of the puzzle for Arm-architecture data center CPUs.
The chain is: Chip design (Arm) → General-purpose computing chips (Ampere) → Data center infrastructure (Stargate + France) → Model layer application (13% stake in OpenAI). From the silicon at the bottom to intelligence at the top, Son has drawn a causal chain for vertically integrated AI infrastructure. The market has given him the corresponding valuation. SoftBank's stock price has risen over 140% from its 2026 low, with annual net profit reaching a record 550.8 billion yen.
Son stated at the shareholders' meeting, "The scale of this AI wave is at least 10 times that of the internet, possibly even 50 times." He drew an analogy to the 1929 Great Depression, where industries related to electrification and industrial mechanization suffered heavy setbacks but then grew for nearly a century.
The problem is that not every investor is making Son's kind of systematic bet. Many simply see his results and replicate his confidence with far cruder logic.
The July Test: The First Challenge to the Faith
In July 2026, a liquidity shock scheduled half a year in advance will arrive on time. The lock-up period for Zhipu's cornerstone investors, holding about 25.86 million shares (5.76% of total shares), will expire. This number may not seem large, but it must be divided by the current float: Zhipu's true free float is only about 2.67%, with just over 12 million shares tradable. The cornerstone unlock will more than triple the available tradable shares overnight.
The impact on MiniMax is more severe. On July 8, approximately 44.29% of shares held by old shareholders (subject to a six-month lock-up) and 5.34% held by cornerstone investors—nearly 50% of total shares combined—will be unlocked simultaneously. Previously, MiniMax's free float was about 5.44%. This means post-unlock, the number of tradable shares will explode from less than 6% of total shares to nearly 56%, a tenfold expansion of the float overnight.
Historical precedents are not comforting. In August 2021, Kuaishou's stock plunged 15.3% on the day of its large-scale post-listing lock-up expiry. Xiaomi fell 7.5% on its first unlock day in January 2019 and continued to drop to new post-IPO lows. However, the situation facing Zhipu and MiniMax is more complex. When Kuaishou's lock-up expired, its stock price had already fallen significantly from its peak, greatly reducing the motivation to sell. In contrast, Zhipu's price remains about 10 times its IPO price, and MiniMax is over 3 times its IPO price. Cornerstone and pre-IPO investors are sitting on paper gains of several times to tenfold.
The timing is also crucial. On June 8, both Zhipu and MiniMax were included in the Hang Seng Tech Index, with a combined weighting of about 5%-7%. Passive tracking funds amount to about $25 billion, implying forced buying of $1.25 to $1.75 billion. Zhipu was simultaneously added to the Stock Connect program, with Southbound funds expected to contribute about HK$18.6 billion in net inflows under a base scenario. MiniMax, due to its weighted voting rights structure, will not be eligible for Stock Connect until August 6.
This liquidity support might alleviate some selling pressure. However, when the free float expands from 3% to over 50%, the magic of the "scarcity premium" pricing factor will diminish. Of course, unlocking does not equal selling. Not all shareholders who gain liquidity will sell on day one. For example, strategic investors may have limited motivation for large-scale, immediate selling. Nevertheless, significant volatility is inevitable.
Shifting the focus back across the ocean, Anthropic, OpenAI, and SpaceX are set to raise a combined total of over $200 billion from their public IPOs. This figure is equivalent to the sum of the US IPO market over the past three to four years and could be concentrated within the next 4-5 months. Specifically, SpaceX aims to raise $75 billion in its June 12 listing. OpenAI and Anthropic confidentially filed S-1 forms on June 8 and June 1, respectively, with listing dates yet to be determined but expected as early as this autumn, each planning to raise $60-90 billion.
The real pain point begins during the roadshow phase. Large asset managers need to sell other tech stocks to free up capital for new subscriptions, starting with SpaceX's June roadshow. Historically, however, fears of liquidity drain from mega-IPOs have proven to be short-term digestion issues, not structural disasters.
The real concern is whether these trillion-dollar valuations are truly justified. If any of these companies breaks issue price, the shock to the valuation anchor for the entire AI sector would be far more damaging than the loss of $200 billion in liquidity. Yet, several secondary market investors stated, "Even if primary market pricing is high, there's still significant room in the secondary market. This round of cash burn is far less brutal than the internet wave, but the growth imagination is greater."
At the peak of the 2000 internet bubble, total annual US venture capital investment in internet companies was about $105 billion (approximately $190 billion in today's dollars adjusted for inflation). However, individual company burn rates were limited. Amazon was considered the biggest "cash-burning machine" of that era, with 1999 revenue of $1.64 billion and a net loss of $720 million. Pets.com burned less than $300 million from inception to collapse; Webvan consumed about $800 million. During the entire dot-com era, no single company lost over $1.5 billion in a single year. But many dot-com companies burned cash with almost no revenue before dying outright, leading to a brutal scene where the Nasdaq fell 78% from 5048 to 1114 points, with hundreds of listed companies going to zero.
In contrast, current AI companies have real and rapidly growing revenue, backed by giants with immense financial strength like Microsoft, Google, and Amazon, making a dot-com-style wave of sudden deaths unlikely in the short term. However, a Morgan Stanley analyst also pointed out that the capital expenditure-to-revenue ratios for Meta, Microsoft, and Alphabet (21-35%) have already exceeded the levels seen at AT&T during the peak of the telecom bubble. Optimism exists, but risks are accumulating.
The Race Against Time
This is a race against time. Domestically, DeepSeek's rumored valuation in May was still in the $10-20 billion range, then quickly revised up to $45-50 billion, and is now approaching $60 billion. According to media reports, StepFun submitted its listing application to the Hong Kong Stock Exchange on June 8, having just completed a nearly $2.5 billion Pre-IPO round. Moonshot AI was reported on June 8 to be seeking a new $2 billion funding round at a valuation targeting $30 billion while preparing for a Hong Kong IPO. 01.AI has pivoted to enterprise AI applications, with founder Kai-Fu Lee explicitly targeting a 2027 Hong Kong listing.
"Nobody knows how long the window will last, but seizing the window is crucial," one investor said. During this window, understanding the logic behind AI faith-based pricing means seizing the opportunities of this golden age. Otherwise, it's just the illusion of a bubble. It is the best of times; it is the worst of times.
Comments