The "Price-to-Dream Ratio" Era of AI Investment: Revelations from the Global Frenzy to China's Twin Stars
- Dr Frederick Wong
- Apr 8
- 6 min read
Based on the deep integration of Tencent Tech's "Global Investors Are Pushing Large Model Companies to a 'Price-to-Dream Ratio'" and Silicon-Based Lab's "Are the 400-Billion-Valued Zhipu and MiniMax Relying on Just Two Formulas?"
I. Phenomenon: An Unprecedented Valuation Revolution
In 2026, the global capital market is witnessing a bizarre "valuation revolution"—large model companies that are not yet profitable and continue to incur massive losses are hitting trillion-dollar market capitalizations at an unprecedented speed.
Global Benchmark: OpenAI's "Astronomical Figures"
Valuation: $852 billion (approximately 6.1 trillion RMB)
Financing: $122 billion (one of the largest in Silicon Valley history)
Revenue: Annualized $25 billion (a 4x increase in 14 months)
Losses: Projected loss of $14 billion in 2026, with cash flow not expected to turn positive until 2030
Price-to-Sales (P/S) Ratio: 34x (Microsoft is at 12x, Google at 6x)
China's Twin Stars: The "Hong Kong Stock Myth" of Zhipu and MiniMax
Metric | Zhipu (02513.HK) | MiniMax (00100.HK) |
Listing Date | January 8, 2026 | January 2026 |
Issue Price | 116.2 HKD | Close to half of the current price |
Latest Stock Price | 917 HKD (+689%) | Continuously doubling |
Market Cap | 400 Billion+ HKD | 380 Billion+ HKD |
2025 Revenue | 720 Million RMB (+132%) | 560 Million RMB (+159%) |
Adjusted Loss | 3.18 Billion RMB | 1.72 Billion RMB |
P/S Ratio | Hundreds of times | Hundreds of times |
Key Signal: Individual investors are pouring in with unprecedented enthusiasm—OpenAI sold $3 billion in shares to individual investors for the first time, the Hong Kong public offering was oversubscribed by 1,159 times, and the closed-end fund VCX surged with a premium of over 31 times on its first day of listing.
II. Consensus: Four Common Logics of the Two Reports
1. "Price-to-Dream Ratio" Replaces Traditional Valuation Systems
Both articles point out that traditional PE (Price-to-Earnings) and PS (Price-to-Sales) ratios can no longer explain current valuations:
Tencent Tech proposes the "Price-to-Dream Ratio": Investors are buying into the belief that AGI will change everything.
Silicon-Based Lab reveals that the two companies use "formulas" to price themselves:
Zhipu: AGI Commercial Value = Intelligence Upper Bound × Token Consumption Scale
MiniMax: Platform Value = Intelligence Density × Token Throughput
Essence: Valuations are no longer based on current profitability, but on the monopolistic potential of future intelligence.
2. Token Economy Becomes the Core Metric
Both reports emphasize that the scale of Token consumption is the key to measuring the commercial value of large models:
Dimension | Specific Manifestation |
Revenue Driver | The explosion of AI agents like OpenClaw leads to exponential growth in API calls. |
Cost Challenge | Complex tasks trigger millions of Token calls, becoming a "Token Black Hole." |
Pricing Power | High-complexity, high-reliability Tokens hold sustained pricing power. |
Stratification Trend | Standardized Tokens move towards low prices/free, while high-quality Tokens command a premium. |
3. FOMO Sentiment Among Individual Investors
Both reports document the frenzy and risks of retail investor entry:
US: The VCX fund soared from a net asset value of $19 to $575 (+2926%), before plummeting 50% after being targeted by short-seller Citron Research.
China: Hong Kong stocks were oversubscribed by 1,159 times, with monthly stock price volatility exceeding 80-100%.
Structural Trap: Fund shares are traded on the public market, but the underlying assets are highly illiquid private company shares.
4. Computing Power Bottleneck is a Common Shackle
Both companies face the dilemma of "having demand but lacking computing power":
Without building their own data centers, computing expenditures flow to cloud vendors (Alibaba Cloud, Tencent Cloud, etc.).
In the Agent era, the number of calls for a single task has increased from 1 to dozens or even hundreds.
Zhipu CEO Zhang Peng warned: "The biggest problem in the next 12 months may be computing power."
III. Divergence: Different Perspectives of the Two Reports
Dimension | Tencent Tech (Global Perspective) | Silicon-Based Lab (Industry Depth) |
Focus | Global capital flows, IPO time windows, institutional divergence | Business model differences, human efficiency comparison, technology roadmaps |
Risk Warning | Lock-up expirations, macro environment, tech stock pressure | Computing power bottlenecks, distance to breakeven, Token stratification |
Valuation Benchmark | OpenAI vs Anthropic vs Tech Giants | Zhipu vs MiniMax vs Cloud Vendors |
Investment Channels | Closed-end funds (VCX/DXYZ), Thematic ETFs | Direct listing in Hong Kong, API economy |
Key Milestones | Q4 2026 - Q1 2027 OpenAI IPO | H2 2026 Zhipu/MiniMax lock-up expiration |
Institutional Attitude | NYU Professor Damodaran questions it, JPMorgan is bullish | JPMorgan ratings, founders' efficiency philosophy |
Core Difference: Tencent Tech focuses more on the cyclical risks of the capital market (FOMO, premiums, short-selling), while Silicon-Based Lab focuses more on efficiency competition at the industry level (intelligence density, human efficiency ratio, full-modality).
IV. Revelations: The Wealth Effect and Future Rules of AI Investment
🔴 Five Major Warnings for Investors
The Inevitability of Valuation Reversion
Historical pattern: After the 2021 SaaS high-valuation IPO wave, most companies' stock prices continued to decline.
Current signals: VCX fell from a peak of $575 to $130, and DXYZ has already gone through a cycle of boom and bust.
Key nodes: VCX lock-up expiration in September 2026; Zhipu/MiniMax lock-up expiration in H2 2026.
The Risk of "Internal Circulation Financing"
Of OpenAI's $122 billion financing:
Amazon: $50 billion ($35 billion contingent on IPO or AGI realization)
Nvidia: $30 billion (mainly computing power, not cash)
SoftBank: $30 billion (arriving in three installments)
Essence: Companies investing in and purchasing from each other, artificially inflating revenue and valuation.
Liquidity Illusion
Closed-end funds can be bought and sold at any time on the NYSE/HKEX.
However, the underlying assets are highly illiquid private company shares.
When sentiment reverses, prices will rapidly collapse toward or even below net asset value.
The Cognitive Gap Between Institutions and Retail Investors
Retail investors: Buying the "dream of changing the world."
Institutions: Focusing on customer retention rates, cash flow, and computing costs.
Information gap: OpenAI has never publicly disclosed its enterprise customer retention rate.
Geopolitics and Macro Environment
In Q1 2026, the Magnificent Seven ETF fell by 16% (the worst in history).
US-Iran conflicts pushed up oil prices, exacerbating global uncertainty.
Pressure on tech stocks and doubts about AI investment ROI may arrive simultaneously.
🟢 Three Major Opportunities for AI Entrepreneurs
Formulaic Survival: Find Your "Valuation Narrative"
Zhipu's path: B-end model capabilities → Intelligence Upper Bound × Token Scale.
MiniMax's path: C-end platform efficiency → Intelligence Density × Token Throughput.
Key: Clarify differentiation and give the capital market a quantifiable story.
Efficiency is Paramount: Human Efficiency Determines Survival
MiniMax uses less than half the headcount of Zhipu to achieve similar revenue.
Pursuing training efficiency in 2025, shifting to R&D efficiency and model iteration efficiency in 2026.
Revelation: In a capital winter, highly efficient teams are better equipped to cross the cycle.
Agents are the Next Battlefield
From "teaching Agents to work" to "observing Agents work."
Model architecture + environment design + multi-Agent coordination become core capabilities.
High-value scenarios (programming, in-depth research) master API pricing power.
🟡 Long-term Thinking on Wealth Management
Allocation Rather Than Speculation
Investors truly allocating AI assets should focus on underlying technological progress rather than stock price fluctuations.
Institutions like ARK Invest and JPMorgan reduce volatility through long-term holding + dollar-cost averaging.
Principle of Diversification
The threshold for directly holding OpenAI/Anthropic shares is extremely high ($25k-$100k + half-year wait).
Disperse risks through ETFs (AGIX, ARKK) and Hong Kong stocks (Zhipu, MiniMax).
Avoid: Chasing a single "Price-to-Dream Ratio" target with all your funds.
Time Frame Matching
Short-term (within 1 year): Extremely high risks of lock-up expirations, macro volatility, and sentiment reversal.
Medium-term (3-5 years): Commercialization of model capabilities, breakeven point.
Long-term (10 years): Whether AGI is realized, whether intelligence becomes infrastructure.
Identify "True Innovation" vs. "Fake Narratives"
True innovation: Genuine growth in Token consumption, pricing power in high-value scenarios, continuous decline in inference costs.
Fake narratives: Relying solely on financing to survive, passing on computing costs, opaque user retention rates.
V. Conclusion: Between Faith and Rationality
Both articles collectively depict a contradictory era:
On one hand, OpenAI, Anthropic, Zhipu, and MiniMax represent the absolute forefront of humanity's march toward AGI, and their technological breakthroughs could reshape all industries.
On the other hand, an $852 billion valuation, hundreds of times P/S ratios, and tens of millions in daily losses challenge all traditional financial valuation models.
For investors: This is a gamble of "voting for the future with real money." Remember Citron Research's warning when shorting VCX—when the premium reverts to net asset value, the price could drop to $26 (close to net asset value).
For entrepreneurs: This is a "dual race of efficiency and narrative." You must have Zhipu's pursuit of an "Intelligence Upper Bound," MiniMax's efficiency of "Intelligence Density," and, most importantly, maintain a clear head amidst capital frenzy.
For everyone: Understanding the logic behind these two formulas is more important than remembering the formulas themselves—
AGI Commercial Value = Intelligence Upper Bound × Token Consumption Scale Platform Value = Intelligence Density × Token Throughput
Because in a rapidly changing era, the only constant is change itself.
References:
Global Investors Are Pushing Large Model Companies to a "Price-to-Dream Ratio"
Are the 400-Billion-Valued Zhipu and MiniMax Relying on Just Two Formulas?
Disclaimer: This article does not constitute any investment advice. Investing in large model companies involves extremely high risks, including but not limited to technological iteration risks, computing power bottlenecks, regulatory changes, and market volatility. Please make decisions cautiously based on your own risk tolerance.
[Limited-Time Expert Consultation Invitation]
FOFA sincerely invites visionary entrepreneurs and investors to engage in deep, practical exchanges regarding the aforementioned trends—specifically the implementation of "AI Agent Labor Systems (Agentic AI)."
We will provide you with a complimentary expert planning consultation to help you tailor a specific entrepreneurial path or investment blueprint, allowing technology leverage to serve your asset appreciation.
Book your strategic dialogue NOW:
Comments