Seven years of losses exceeding 50 billion yuan, with a market value less than a quarter of MiniMax. Does the capital market still believe in SenseTime's AI story?

Ask AI · Why is the valuation gap between SenseTime and Minimax so large in the capital markets?

With revenue hitting record highs, losses narrowing significantly, and EBITDA turning positive for the first time… SenseTime’s 2025 performance has sent many encouraging signals. However, at the same time, this long-established AI company’s awkward predicament still remains: the profit inflection point has not arrived, a second growth curve has not emerged, and its valuation has been comprehensively overtaken by later entrants.

Recently, SenseTime (0020.HK) released its full-year 2025 results report. The data show that, during the reporting period, the company’s total revenue reached RMB 5.015 billion, up 32.9% year over year, setting a historical high. Full-year net loss was RMB 1.782 billion, narrowing by 58.6% year over year. More worth noting is that in the second half of 2025, EBITDA (earnings before interest, taxes, depreciation, and amortization) reached RMB 376 million, turning positive for the first time since listing.

Looking through the impressive figures on the books, after business adjustments and strategic transformation, SenseTime’s revenue growth already shows growing dependence on its generative AI business. Innovative initiatives are still far from becoming a “second growth curve.” Against the backdrop of AI expansion, pressure on gross margin arises from rising compute costs and depreciation. Behind the sharp loss reduction is the fact that the company has divested businesses that were continuously burning cash; however, the core business is still loss-making. And since 2018, SenseTime has never managed to achieve annual profitability.

Even more brutal is the shift in valuation logic across the AI sector. Compared with traditional AI companies, the capital markets have turned to favor the new darlings of general-purpose large models with more imagination. When Yan Junjie, a former SenseTime vice president, led Minimax to a market capitalization of HK$300 billion, SenseTime—once the “No. 1 among the AI Four Little Dragons”—had long already fallen out of the industry’s first echelon, with its market cap now at less than one quarter of Minimax’s.

Generative AI has become the “main leg”

The most striking part of SenseTime’s 2025 annual report is undoubtedly its generative AI business. According to the financials, generative AI revenue increased from RMB 2.404 billion in 2024 to RMB 3.630 billion, a rise of 51.0%. The share of this business in total revenue increased from 63.7% in 2024 to 72.4%, making it the revenue mainstay for SenseTime.

This growth is mainly driven by the continued surge in market demand for generative AI model training, fine-tuning, and inference, as well as the integrated industrial solution approach that promotes joint commercialization across the compute platform, models, and applications.

Behind the breakout of generative AI business is SenseTime’s reconstruction of its core narrative. In 2023, SenseTime “cut” its prior four business lines—smart commerce, smart living, smart automotive, and smart cities—and clearly defined three new business segments: generative AI, traditional AI, and intelligent vehicles. By the end of 2024, it adjusted again to generative AI, intelligent vehicles, and vision AI. In the first half of 2025, SenseTime further changed the basis for its revenue disclosure, adjusting to generative AI, vision AI, and X innovation business.

Benefiting from its forward-looking layout of the “AI infrastructure—large models—applications” chain, over the past three years SenseTime’s generative AI revenue has continued to break through the RMB 1 billion, RMB 2 billion, and RMB 3 billion thresholds, with its revenue share increasing by 37.6 percentage points.

From being a leader in computer vision to becoming a generative AI service provider, concerns are hidden in the growth backdrop of SenseTime’s transition. Compared with the explosive growth rates of 199.9% and 103.1% achieved in 2023 and 2024 respectively, even though the growth rate of the generative AI business remains at a high level, the momentum has started to cool.

At the same time, although the scale is growing, it does not seem to have formed the “scale effect.” In 2025, SenseTime’s cost of sales increased by 37.4% year over year to RMB 2.958 billion, including AIDC compute operating costs surging by 163.5% year over year. With the continued expansion of the generative AI business, server depreciation, hardware procurement, and cloud service expenses remain elevated, causing the company’s overall gross margin to fall further to 41.0%.

Where will the new growth momentum come from

When generative AI suddenly becomes the core engine of growth, SenseTime also starts facing the test of whether its business structure is balanced.

In the first half of 2025, SenseTime’s vision AI segment absorbed the smart cockpit business from its former intelligent automotive business. Meanwhile, under its “1+X” strategy, high-investment, long-cycle unified efforts—such as intelligent automotive, smart healthcare, home robots, and smart retail—were all brought into the X innovation business segment.

As the number one leading company in China’s computer vision field, SenseTime’s vision AI business—once the largest source of revenue—has seen a sharp decline in revenue since 2023. In 2024, revenue fell significantly by 39.5% to RMB 1.112 billion. After adjusting the disclosure basis in 2025, vision AI revenue was RMB 1.047 billion, up 3.4% year over year.

In fact, To G businesses mainly driven by smart cities and security and surveillance have long faced issues such as project-based delivery, long cash collection cycles, and demand nearing saturation. SenseTime’s proactive shrinking of this segment and shifting its focus also indirectly suggests that the growth rate of vision AI is limited and cannot become a second growth curve. Instead, the emphasis is on its profitability. As SenseTime states in its financial report, the vision AI business is moving from a technology investment phase into a scalable harvest phase, becoming a solid pillar driving group revenue growth and improving cash flow.

The X innovation business, which represents the direction of future development, has ended up contracting. In 2025, this segment’s revenue fell 5.9% year over year to RMB 302 million. Its share of revenue dropped from 8.6% to 6.0%. SenseTime attributed the reduction in this revenue to the spinoff of its intelligent driving business from the financial statements in August 2025.

But the deeper problem is that although sectors such as smart healthcare, home robots, and smart retail are considered to have broad prospects, they are still in the investment stage. In some sectors, the commercialization pathways are even unclear, meaning they cannot scale quickly, and in the near term it is also difficult to form stable profits.

SenseTime disclosed in its financial report that under its “mother ship (group) + daughter ship (ecosystem)” synergy system, ecosystem companies incubated by the group have made smooth progress in external financing in the primary market and have received high recognition from outside capital, including internet giants, top-tier venture capital firms, and industry funds. However, the financial report did not disclose specific financial figures or project progress for these “X” innovation businesses.

Behind the “track record” of reducing losses

In SenseTime’s financial report, “a significant reduction in losses” is a standout performance that cannot be overlooked. According to reports, during the earnings call after the report was released, company executives frequently mentioned this keyword.

By the numbers, SenseTime’s 2025 net loss was RMB 1.782 billion, narrowing by 58.6% year over year. Adjusted net loss was RMB 1.956 billion, narrowing by 54.3% year over year. To highlight performance improvement, SenseTime emphasized that in the second half of 2025, EBITDA was RMB 376 million, turning positive for the first time since listing.

In fact, in the first half of 2025, SenseTime had already separated the loss-making smart driving and world model businesses within its “Jueying” intelligent automotive business, thereby narrowing overall net loss. In addition, AI GPU chips were no longer within the scope of consolidated financial statements since the beginning of 2025. Likewise, terminal chip companies have not been consolidated into the financial statements since the second half. And according to SenseTime’s plan, the X innovation business will ultimately be carved out from the group’s consolidated financial statements.

In SenseTime’s 2025 financial report, there is an “other gains, net” line item of more than RMB 1.9 billion, mainly consisting of gains of RMB 1.313 billion from selling subsidiaries and associates, and fair value gains of RMB 646 million from financial assets measured at fair value with changes recognized in profit or loss.

For comparison, this figure was RMB 539 million in 2024 and RMB 641 million in the first half of 2025.

Additionally, to achieve loss reduction, SenseTime put comprehensive controls on its “three expense items” expenditures. Among them, selling expenses had already decreased by 20% year over year in 2024, and continued to decline by 13.1% to RMB 569 million. Administrative expenses also fell by 16.2% year over year to RMB 1.226 billion.

More critically, research and development expenses decreased by 8.6% year over year to RMB 3.775 billion. Looking closer, this decrease was mainly due to lower employee benefit expenses, partially offset by higher server operating and cloud service expenses. In both 2024 and the first half of 2025, SenseTime’s R&D expenses increased by 19.2% and 12%, respectively. The main drivers were depreciation and amortization related to investments in its generative AI business, as well as costs for server operations and cloud services.

The disclosed number of employees in the financial report also supports this. By the end of 2025, SenseTime had 2,472 employees, down by 1,284 from 3,756 at the end of 2024. If you extend the timeline further back, at the end of 2022 SenseTime had more than 5,000 employees, showing a year-by-year declining trend.

Overall, after excluding the RMB 1.276 billion contributed by asset sales in the second half and the impact of reducing R&D expenses in the second half, SenseTime’s core business remains in a state of significant losses.

When will the profitability inflection point arrive?

Looking across a longer timeline, for more than a decade since its establishment, pressure to become profitable has always been the sharp sword hanging over SenseTime.

Based on financial reports over the years, from 2019 to 2024, SenseTime’s annual losses were RMB 4.968 billion, RMB 12.158 billion, RMB 17.177 billion, RMB 6.093 billion, RMB 6.495 billion, and RMB 4.307 billion. If you add last year’s loss of RMB 1.782 billion, SenseTime’s cumulative losses over the past seven years have already reached an enormous RMB 53 billion.

On the difficult road toward the profitability inflection point, SenseTime continues to rely on external “capital injections.” The 2025 annual report shows that as of December 31, 2025, the net proceeds from the December 2024 placement of HK$2.787 billion and the net proceeds from the July 2025 placement of HK$2.498 billion have already all been used. The net proceeds from the December 2025 placement of HK$3.146 billion are expected to be fully used by the end of 2026.

Most of the funds are used to support the development of SenseTime’s core AI business, including AI infrastructure construction, generative AI research and development, commercialization of large model products, and integration and application in innovative AI fields.

In contrast to the need for external funding to cover ongoing losses is SenseTime’s simultaneous “cash drain” in the capital market. As of March 26, the SenseTime share price was HK$1.85, with a total market capitalization of less than HK$75 billion.

This is a huge gap compared with its high point at the time of its listing at the end of 2021—back then, SenseTime’s offering price was HK$3.85. After the listing, the share price quickly climbed, with gains reaching as high as 22% at one point, and market capitalization once exceeded HK$150 billion.

It is rather dramatic that just before SenseTime’s listing, in late 2021, Yan Junjie, then SenseTime’s vice president, chose to leave and, in early the following year, founded an AI company called Minimax.

Earlier this year, Yan Junjie’s Minimax listed on the Hong Kong Stock Exchange. On its first day, the share price doubled, and its market capitalization exceeded HK$100 billion. Over the following more than two months, it continued to be chased by capital, and now its market capitalization has reached HK$300 billion. In contrast with the surge in its market value, this new AI large-model darling reported full-year 2025 revenue of US$79.038 million, up 158.9% year over year; and its full-year loss was as high as US$1.872 billion, up 302.3% year over year.

Behind the valuation gap is the capital market’s completely different pricing logic for new versus old AI companies. Compared with SenseTime’s “loss reduction” story, investors are more willing to pay higher premiums for “growth stories” like Minimax and Zhipu, even though their loss scales are larger, thanks to their combined advantages in AI application business models, high-growth expectations, and the scarcity of their underlying assets.

By Cai Shumin

Text editor Ma Yunfei