China's Public AI Labs Offer Sobering Preview of a Potential OpenAI IPO

The performance of two Hong Kong-listed artificial intelligence labs is providing the first public market blueprint for what a potential IPO from OpenAI or Anthropic might look like, revealing a landscape of staggering valuations, immense cash burn, and extreme volatility.

ZAI Co Ltd (2513.HK) and MiniMax Group Inc (0100.HK) are the closest public analogues to the “pure AI frontier lab” business model, and their financial disclosures paint a cautionary picture for investors.

Sky-High Valuations on Nascent Revenue

Despite rapid, triple-digit revenue growth, the absolute sales figures for both companies remain modest relative to their market capitalizations. According to their fiscal 2025 results, ZAI and MiniMax are trading at exceptionally high multiples.

• ZAI Co Ltd (2513.HK): Generated $103.6 million in revenue but commands a market cap of $47.1 billion, implying a price-to-revenue multiple of approximately 455x.
• MiniMax Group Inc (0100.HK): Reported $79.0 million in revenue against a $13.3 billion market cap, for a price-to-revenue multiple of about 168x.

These valuations suggest investors are pricing in long-term market dominance rather than current financial performance, a sentiment that would likely carry over to a high-profile U.S. AI listing.

The Sobering Reality of Unit Economics

Ad

The most instructive parallel for potential Western IPOs lies in the companies' profitability metrics. Both firms are burning through cash at a substantial rate, with ZAI reporting a net loss of -$274 million and MiniMax a loss of -$1.87 billion in fiscal 2025.

ZAI’s gross margin compression, falling from 66% in 2023 to 41% in 2025, signals that high compute costs are outpacing its pricing power. This mirrors the structural headwind OpenAI has publicly acknowledged. In contrast, MiniMax’s expanding gross margin (to 25.4% in 2025) suggests efficiency gains are possible as model infrastructure matures, though it remains deeply unprofitable.

A Blueprint for a US AI Listing

Synthesizing the performance of these two labs suggests that a publicly traded OpenAI or Anthropic would likely exhibit several key traits:

• Extreme initial valuations, with revenue multiples potentially ranging from 200x to 500x.
• Deepening net losses for several years as spending on compute and research continues to outpace revenue growth.
• Substantial volatility, as seen with ZAI and MiniMax, which are down 73% and 78% from their respective 52-week highs.
• New layers of risk, including regulatory and reputational challenges. A recent data-privacy controversy surrounding ZAI's coding assistant has already created sector-wide pressure in Hong Kong, a preview of the scrutiny public AI companies would face.