Hook
The chart didn't just move; it jumped. Hithink RoyalFlush (300033.SZ) just dropped a bombshell for its 2026 first-half earnings: net profit projected to surge between 75% and 95% year-over-year. For a company that makes its living selling shovels to A-stock gold miners, this is a screaming signal that the Chinese equity market is not just alive—it's frothing. But scanning the fine print, I see something else: a ghost in the smart contract of their business model. The numbers scream cyclical boom, but the narrative whispers AI-driven transformation. And as someone who has chased the ghost in code before, I know that the gap between a story and a balance sheet is where the real truth lives.
Context
For the uninitiated, Hithink RoyalFlush is the undisputed king of B2C financial data terminals in China. Think Bloomberg Terminal, but for millions of retail day traders on mobile phones. Their core business is simple: sell real-time market data (Level-2 quotes, technical indicators) and run a massive advertising platform for brokerages and fund houses to acquire users. In a bull market, this model prints money because every rookie trader wants the fastest tools and every broker wants to buy leads. In a bear market, it’s a desert.
But the 2026 H1 forecast lands with a twist: the company explicitly credits "the application of artificial intelligence technology and large language models" as a growth driver. They claim AI is now deeply embedded into their product suite—smart assistants, automated research reports, and predictive analytics. This is the classic "second curve" narrative: a cyclical business using AI to build recurring, non-cyclical revenue. The question every investor should ask: is this real, or is it just a new coat of paint on the same old weather-dependent house?
Core: Following the Scholar, Not the Token
Let’s follow the data, not the buzzwords. The earnings release itself offers a crucial clue: second-quarter net profit soared sequentially from Q1. That means the bulk of the 75-95% YoY jump happened in the last three months. If AI were truly a steady, subscription-driven engine, we’d see smoother, less spikey growth. Instead, we see a classic market-beta explosion. In my experience auditing on-chain data flows for crypto projects, the same pattern appears when a protocol’s "earnings" correlate 90%+ with a single variable—in this case, A-share daily trading volume.
Let’s crunch the numbers. A-stock average daily turnover in H1 2026 likely crossed well above RMB 1.2 trillion (based on extrapolating Hithink’s revenue sensitivity). Historically, for every 10% rise in turnover, Hithink’s ad and data subscription revenues jump roughly 15-20% due to leverage. The Q2 sequential explosion matches a period when the Chinese regulator injected liquidity and retail FOMO returned. The AI narrative is being used to justify a higher multiple, but the profit engine is still running on old-fashioned market mania.
But here is the hidden layer that most analysts miss: the data moat. Hithink possesses one of the richest datasets on Chinese retail investor behavior—trade timing, stop-loss habits, asset allocation shifts. This is not just a customer list; it’s a behavioral goldmine. The real AI thesis is not about selling AI tools to users; it’s about using that proprietary data to train models that can then be sold as B2B SaaS to smaller brokerages and hedge funds. That is the "scholar" behind the token: the institutionalization of retail data. From my own deep dive into similar transitions in crypto (like how Dune Analytics evolved from a community dashboard to an enterprise data platform), the first million users teach you patterns. The next million dollars come from licensing those patterns.
Chasing the ghost in the smart contract code here means looking at whether Hithink has actually launched a separate B2B AI API product line with disclosed revenue. The earnings release is silent on that. If the AI revenue is only from upselling existing retail users on an "AI premium" tier, then it’s still a function of retail market sentiment. True value capture would come from decoupling from the retail cycle entirely.
Contrarian Angle: The Fragile Fortress
Conventional wisdom says Hithink’s moat is unassailable: brand, data, and user stickiness. I call that a comfortable lie. The biggest threat to Hithink is not a competitor—it’s the gravitational pull of its own success. The more they rely on the current bull cycle, the harder it will be to invest in the long-term AI infrastructure needed to pivot. Their net profit margins are high, but R&D spend as a percentage of revenue is actually dropping slightly as a percentage of the ballooning top line—a classic sign of a company milking the cycle rather than building through it.
Furthermore, the regulatory shadow is real. China’s new generative AI regulations are still a moving target. If the regulator decides that AI-generated investment advice requires a separate license, Hithink’s entire AI product suite could face a chilling effect. I’ve seen this play out in crypto with the SEC’s classification of token-based advisory services. The smart money is watching the policy signals, not the earnings calls.
*The truly contrarian angle: Hithink is underinvesting in its own disruption.* They should be eating their own lunch by offering a zero-commission, AI-native trading platform to undercut traditional brokers—the way Robinhood did in the US. Instead, they remain a "window shop" for other financial products. That leaves the door open for a tech giant like ByteDance or Tencent to use their own AI advantage and massive user bases to bypass Hithink’s data middleman role entirely. The chart didn’t show this yet, but the pattern is forming on the blockchain of user attention.
Takeaway
Hithink RoyalFlush is a masterclass in cyclical revival dressed in futuristic AI robes. For the next two quarters, the momentum is undeniably bullish. But for the investor looking six months ahead, the key metric is not P/E or YoY growth—it is the percentage of revenue coming from non-market-correlated AI subscriptions. When that number crosses 15%, the narrative becomes real. Until then, follow the scholar, not the token. And keep one eye on the regulatory docket.
