On August 17, the established AI company SenseTime experienced a long-awaited strong surge, with its stock price rising over 13% intraday and closing up more than 8%, bringing its total market capitalization back to approximately HK$64.6 billion. This surge in investor enthusiasm was triggered by SenseTime’s “positive profit warning” released on the evening of the 16th, in which the company forecasted it would achieve profitability for the first half of 2026—marking its first consolidated profit since listing on the Hong Kong Stock Exchange in 2022.
However, while the market celebrates this historic turning point, SenseTime’s stock price remains far from its peak. Looking back to the end of 2021 and early 2022, shortly after its listing, SenseTime was heavily favored by investors, with its stock price briefly nearing HK$9.70 per share and its market capitalization surpassing HK$270 billion. Today, even after today’s sharp rally, its stock price hovers around HK$1.53, with its market cap having declined by more than 80% from its peak.
From being the most sought-after among the "Four Great AI Giants" to facing its darkest hour amid multiple storms, this once-dominant company at the forefront of the trend is now striving to catch up with the times by fully transitioning into generative AI.
Why did it fall behind when AI was at its peak?
In 2018, if you asked someone in the tech industry: Which Chinese company is strongest in AI?
The answer likely includes SenseTime.
Indeed, during those years, SenseTime was the undisputed leader. Wherever there was a hot AI-related field—facial recognition, smart cities, autonomous driving—you could almost always see its presence. Along with Megvii, Yitu, and CloudWalk, it was known as one of the “Four Little Dragons” of AI, a constant fixture in tech media coverage, raising round after round of funding as its valuation soared. Back then, SenseTime was like the top student in school who aced every exam—everyone believed it had an incredible future.
However, fast-forward to 2023, and the situation suddenly changed.
When ChatGPT, DeepSeek Claude and Gemini have ignited global attention; as AI becomes the hottest concept in capital markets and OpenAI and Anthropic become the focus of the tech world, SenseTime has gradually faded from public view.
Why is this happening? The answer may be harsh: SenseTime embraced AI 1.0 but missed AI 2.0.
To understand this, you first need to clarify what distinguishes AI in these two eras.
In the era when SenseTime started, AI primarily performed "recognition" tasks—given a face, it would identify who it was; given a photo of an intersection, it would mark which vehicle ran a red light; given a surveillance video, it would count how many people passed by.
This technology has a professional term: computer vision—essentially teaching machines to understand the world. SenseTime has become a leader in this field, earning substantial profits by providing enterprises with facial recognition and smart city solutions.
But the emergence of ChatGPT completely changed the rules of the game. The new generation of AI no longer just understands—it also creates, writing articles, producing videos, and even generating code.
It’s like Shang Tang spent ten years mastering the art of recognizing punches, finally becoming the best in the world—only for everyone to suddenly start practicing generative sword techniques. The old skills haven’t become useless, but they’re no longer the main focus.
If it were only a matter of falling behind on the technological front, SenseTime might still have had a chance to turn things around. What was truly devastating was that, from 2019 to 2023, bad news came one after another, as if prearranged.
Starting in 2019, the United States added SenseTime to its sanctions list. Then, at the end of 2022, ChatGPT emerged unexpectedly, fundamentally disrupting the industry’s technological paradigm. SenseTime’s core competency in computer vision, once a future-oriented field, suddenly became a traditional business, while its efforts in large models lagged behind.
When it rains, it pours. The once cash-generating smart city business has also begun to shrink, causing revenue growth to stall.
2023 was also SenseTime’s “darkest hour.” That year, founder Tang Xiao’ou passed away, sparking widespread sorrow across the industry and delivering a severe blow to internal morale and investor confidence; almost simultaneously, short-seller Grizzly Research released a report questioning SenseTime, further dragging down the stock price.
A series of blows has led to startling figures: in 2021, smart city services accounted for 45.6% of SenseTime’s total revenue; by 2022, this share had dropped to 28.8%; and in 2023, it fell further to approximately 12%. SenseTime has even proactively reduced its reliance on smart city services.
Meanwhile, the market capitalization dropped from a peak of approximately HK$300 billion to just a few billion HKD; the workforce was drastically reduced from 6,114 employees to 2,472, a cut of nearly 60% over three and a half years—equivalent to six out of every ten employees leaving.
Almost every crisis imaginable has occurred over the past few years.
This is less a story of falling behind and more a story of a star AI company struck by multiple storms at a turning point in history.
Can the company that once made money through "facial recognition" make a comeback?
If you think SenseTime has completely disappeared and will never reappear, you might be overlooking a very interesting fact: SenseTime has transformed from a company focused on facial recognition into a technology enterprise driven by both generative AI and computer vision.
To put it simply: SenseTime actively "killed" its past self.
Based on the latest earnings forecast, the transformation has already shown initial results: for the first half of 2026, the company expects to record a net profit of approximately RMB 500 million to 700 million, compared to a net loss of approximately RMB 1.489 billion in the same period last year; the net loss has decreased by 60% to 70% year-over-year.
Notably, SenseTime explicitly stated in its announcement: “This will be the company’s first-ever consolidated profit since its listing.”
This is a very important signal, because over the past few years, the biggest question outsiders have had about SenseTime is whether it can actually make money.
As one of China’s most prominent stars in the AI industry, SenseTime told a story of technology transforming the world between 2018 and 2021. At that time, it was widely accepted that whoever possessed the most advanced algorithms, the most AI research papers, and the highest facial recognition accuracy would likely become the winner of the next era.
But later, people realized that technological leadership does not guarantee commercial success.
In the past, SenseTime's core business was smart cities, smart commerce, and smart living. Put more simply, this meant selling artificial intelligence solutions to local governments and enterprises—such as facial recognition access systems, urban security systems, traffic recognition, and campus management.
This business model has an inherent flaw: once the project is completed, the revenue ends. This makes it more like a technology services company than an internet platform.
An internet platform can generate continuous revenue from its users, but project-based businesses must continually seek new orders.
This is also why SenseTime has struggled to overcome its losses over the past several years.
The emergence of generative AI has changed all of this. Previously, SenseTime sold projects; now, it sells models.
A large model can simultaneously serve multiple industries, including finance, education, office productivity, robotics, and content creation. It is no longer a one-time delivery business, but rather a sustainable subscription model. The greatest advantage of this approach is that, in theory, it offers greater economies of scale.
From a business structure perspective, generative AI has become SenseTime’s new growth engine. For example, in full-year 2025, generative AI revenue reached RMB 3.63 billion, accounting for 72.4% of total revenue; meanwhile, revenue from the company’s other core business, visual AI, was RMB 1.08 billion, representing 21.6%.
In other words, SenseTime, once known for generating revenue through facial recognition, now derives more than seventy percent of its income from large models.
Meanwhile, SenseTime has also begun proactively controlling costs.
Training large models is essentially a game of burning money. Over the past few years, SenseTime has been labeled a "money-burning machine." From its founding to its IPO, cumulative R&D investment exceeded ten billion yuan, and ongoing losses have become the biggest concern for investors.
However, over the past two years, SenseTime has clearly adjusted its strategy. On one hand, it has begun emphasizing model training efficiency. On the other hand, it has been working to reduce its reliance on high-end computing power by optimizing model architectures and adapting to domestic chips, thereby lowering training costs.
From the operating results, this strategy has indeed begun to take effect. However, the issue is that profitability does not mean the company has turned things around. At this point, if asked whether SenseTime has turned things around, the answer is likely no—because it is now facing a larger question: Is there still an opportunity for independent AI companies?
Because the current competition among large models has essentially become a battle of ecosystems.
Look at the players standing at the center of the table: OpenAI is backed by Microsoft, Google owns search, Android, and cloud computing, Alibaba has Alibaba Cloud and its e-commerce ecosystem, Tencent owns WeChat, and ByteDance owns TikTok... These companies share a common trait: they don’t just have models—they also possess computing power, users, traffic, and application entry points.
The model is just one part; what truly determines victory is the entire ecosystem—and this is precisely where SenseTime is weakest.
In fact, this is not just an issue facing SenseTime, but a challenge shared by all independent AI companies worldwide.
In the past, SenseTime needed to prove whether AI technology could become a viable business. Today, it must demonstrate that, in the era of large models dominated by giants, an AI company without a super traffic gateway can still build a new commercial moat through its model capabilities and industry implementation expertise.
This article is from the WeChat official account "Phoenix Tech," authored by Phoenix Tech.