SenseTime Says It Turned a Profit. The Math Says Something More Complicated.

Chinese AI company SenseTime announced on August 16 that it will report its first-ever consolidated profit — RMB 500 million to 700 million — for the first half of 2026. The stock rose 8 percent the next day. It would be the first profitable half-year since the company listed in Hong Kong in late 2021, swinging from a loss of RMB 1.49 billion in the same period a year earlier.

That's the headline. Here's what the headline leaves out.

The profit exists almost entirely because the companies SenseTime invested in — fair value gains from investments in its AI ecosystem — have appreciated in value. These fair value gains are real under IFRS accounting. They are not revenue from SenseTime's own products, and they are not guaranteed to recur.

On an adjusted basis, which strips out these mark-to-market investment swings, SenseTime is still losing money. The adjusted net loss will narrow by 60 percent to 70 percent year-over-year, but narrow is not eliminated. The core business — selling AI software, large model services, and computer vision systems to enterprise customers — has not yet crossed into profit territory.

That distinction matters. It tells you whether the company is actually solving its profitability problem or simply riding a bull market in the startups it funded.

The core story underneath is different from the profit headline, and it's more interesting. SenseTime has spent four years losing money at a massive scale. Since its listing in December 2021 at roughly HK$9 per share, the company burned through billions while trying to build the largest AI software platform in China. Its workforce shrank from 6,114 to 2,472 employees between 2022 and 2025. The old smart-city business, which once made up 45.6 percent of revenue in 2021, falling to about 12 percent by 2023, withered as Chinese government spending on surveillance projects dried up.

Then came the pivot. Generative AI replaced smart cities as the revenue engine. In 2025, the large model segment generated RMB 3.63 billion — 72.4 percent of total revenue — up from virtually nothing three years prior. Revenue for the full year rose 33 percent to more than RMB 5 billion. Adjusted losses narrowed for the fourth consecutive half-year. In the second half of 2025, operating cash flow turned positive for the first time since listing. Trade receivable collections reached a record RMB 4.87 billion in a single half-year.

Those are not vague trend lines. They're the kind of sequential improvement that suggests a company is actually reaching its breakeven point.

The question is whether the operating improvement is structural or just a cost-cutting exercise dressed up as growth. SenseTime's strategy hinges on two things: efficiency and ecosystem.

On efficiency, the company is betting that it can win on price even if it doesn't lead on raw model quality. Cofounder Lin Dahua told CNBC that SenseNova U1 costs ten times less than OpenAI's ChatGPT Images 2.0, acknowledging a quality gap but arguing that customers will choose the cheaper model that "does the job." The company's proprietary NEO architecture requires one-tenth of the training data and computing power of peer models, and SenseTime is actively adapting its systems to run on domestic Chinese chips to reduce inference costs. This isn't a claim of technological supremacy — it's a claim of commercial pragmatism.

On ecosystem, SenseTime uses what it calls a "1 + X" strategy. The "1" is its dual core: generative AI and computer vision. The "X" is an incubation layer of vertical-specific startups in areas like autonomous driving, healthcare, retail, and robotics. These X businesses are where SenseTime makes its investments — and where the fair value gains in H1 2026 came from.

That's the key tension in this earnings story. The ecosystem investments validated SenseTime's thesis in the market, but they also created a dependency. If the AI startup market cools — as it could if funding slows or valuation multiples compress — those fair value gains reverse. The H1 2026 profit, in that scenario, turns back into a loss.

Meanwhile, the competitive moat remains an open question. SenseTime is an independent AI company in a market dominated by platform giants — Baidu, Alibaba, and Tencent — that have their own cloud infrastructure, billions of users, and deep cash reserves to subsidize loss-making AI development. A strategist at UBP noted that standalone players are "continuing to be loss-making", and SenseTime is only just now approaching profitability on an adjusted basis. The company is also under U.S. sanctions, which blocks it from accessing the most advanced semiconductors and limits its ability to expand into Western markets.

That said, the balance sheet gives SenseTime room to breathe. Cash reserves stood at RMB 13.2 billion as of mid-2025. Total debt is roughly $850 million. At those levels, the company isn't running out of runway anytime soon, even if profitability takes longer than expected.

Here's the investor frame. SenseTime is trading at a market cap of approximately HK$64 billion — roughly 80 percent below its peak of approximately HK$270 billion in late 2021. The market is not pricing this as a profitable company. It's pricing it as a company that might one day become one, with a significant discount for the execution risk and competitive uncertainty.

The H1 2026 result moves the needle on that question. The adjusted loss narrowing by 60-70 percent year-over-year is real progress. Revenue is growing at roughly 30 percent. Operating cash flow turned positive in H2 2025. The fair value gains, while volatile, confirm that the AI ecosystem SenseTime helped build has real market value.

But the profit headline itself is noise if you take it at face value. The adjusted number tells you the actual state of the business: improving, approaching, not arrived. The next earnings report will be the test. If the adjusted loss narrows again — or disappears — the thesis that SenseTime has crossed a threshold gains real traction. If it doesn't, the fair value gains alone won't carry the stock.

The stock is still trading far below its peak, which means the market hasn't declared the turnaround story over. That gives a patient investor an option: wait for the adjusted loss to actually vanish, confirm the revenue growth holds, and only then decide whether SenseTime has earned a re-rating. Or buy the improvement trajectory now and accept that the path to sustained profitability is not yet proven.

Either way, don't mistake accounting profit for business profit. The former is a moment. The latter is a pattern. SenseTime is building the pattern, but it isn't finished yet.