Hook
Tonghuashun (300033.SZ) just reported a 75%-95% net profit surge for the first half of 2026. The market cheered. Analysts praised its AI integration. But the front-runner didn't celebrate — it read the fine print. The entire increase came from a surge in A-share trading volume, not from any structural improvement in its product. This is not innovation. This is a cyclical fluke dressed up as a tech breakout. The same pattern will repeat — until the market turns, and the AI narrative will evaporate faster than a failed layer-2 token.
Context
Tonghuashun is China’s dominant retail financial information platform — think of it as a Bloomberg Terminal for the masses, heavily focused on stock trading tools, feeds, and fund distribution. In 2025, it launched its own large language model (LLM) for financial scenarios, promising AI-powered stock screening, smart assistants, and personalized analytics. The press ate it up. Investors bought the story. But the underlying economics remain unchanged: the company earns the bulk of its revenue from subscriptions and advertising that scale directly with retail trading activity. The 2026 H1 boom was driven by a loose monetary policy in China and a market rally, not by AI adoption. The second quarter net profit exploded sequentially — a textbook sign of cyclical leverage, not sustainable growth.
Core: The Systematic Teardown
Let me be blunt: Tonghuashun’s AI pivot is a distraction from its fundamental fragility. I’ve spent 29 years in crypto due diligence, and I’ve seen this script before — projects that wrap themselves in technical buzzwords to mask a broken incentive structure. Here’s what the market misses:
1. The AI model is a black box built on biased data.
The company’s LLM is trained on years of retail trading data. But retail data is inherently noisy and full of behavioral biases — chasing momentum, panic selling, herding. Training an AI on such data means it will amplify those biases, not correct them. In my 2017 audit of the EOS smart contract, I found a race condition that could mint infinite tokens. The design flaw was hidden by hype. Here, the flaw is hidden by data volume. The AI cannot differentiate between signal and noise because it was trained on noise. It will produce recommendations that look smart in a bull market and devastating in a bear market.
2. The revenue model has zero alignment with user outcomes.
Tonghuashun gets paid when users trade more — it earns from commissions on fund distribution and advertising. Its AI tools are designed to increase user engagement and trading frequency, not to improve investment returns. That’s a misaligned incentive structure. A bug is just a feature that hasn't been exploited yet. In this case, the bug is the business model: the company profits from churn, not from long-term value creation. Compare this to a well-designed DeFi protocol where fees are tied to protocol health, not user hyperactivity. Tonghuashun is the equivalent of a casino giving away “free” analytics.
3. Regulatory headwinds are coming for AI in finance.
China’s regulators have already signaled stricter oversight of AI-based investment advice. The same SEC-style regulation-by-enforcement pattern is unfolding in Beijing. Tonghuashun’s AI currently operates in a gray zone — it provides “intelligence” but claims it’s not giving investment advice. Once the rules land (and they will, likely within 12 months), the company will have to either remove key AI features or apply for an expensive license. This is identical to the regulatory ambiguity that killed many DeFi projects in 2022-2023. The market is pricing in zero regulatory risk.
4. The moat is an illusion.
The company’s only true moat is its user base — but that user base is tied to the A-share market, not to the platform itself. Switching costs are low; users can easily migrate to East Money (300059.SZ) or Alipay’s wealth management section. The AI features do not create a network effect because they don’t improve with more users — they only improve with more biased data. In crypto, we call this a “pseudo-network effect.” Real network effect requires users to add value to each other, like liquidity providers in a DEX. Tonghuashun is just a tool. Tools get replaced.
Contrarian: What the Bulls Got Right
I will concede one point: Tonghuashun has an unparalleled data asset. Its decades of tick-level trading data, emotional sentiment snapshots, and user behavior logs are valuable for market microstructure analysis. If the company were to sell this anonymized data to institutions or use it to build a truly transparent, auditable AI model, it could become a foundational layer for financial analytics — akin to The Graph for finance. The bulls are right that data is an asset. But data alone is not a business. The company needs to convert that data into a subscription product that is uncorrelated with market cycles. They haven’t done that yet. The current AI product is still a marketing gimmick.
Takeaway
Tonghuashun’s 2026 H1 results are a classic bull-market mirage. The AI narrative is a smoke screen for a structurally fragile, cyclical business. When the A-share market corrects, the net profit will halve, the AI will be blamed for bad recommendations, and regulators will tighten the noose. The question is not if, but when. Meanwhile, the company’s stock trades at 80x trailing earnings — a valuation that assumes the bull market lasts forever. In crypto, we call that a diamond hand’s hope. In traditional finance, we call it a short signal. The front-runner didn't bet on AI — it bet on the cycle. And the cycle always turns.
