In this edition of Daily Science and Technology Innovation, we turn our attention to a relatively low-profile company on the Science and Technology Innovation Board in the field of computer vision. It doesn't manufacture mobile phones, cars, or sell cameras, but its algorithms are likely powering the photo-taking capabilities of your phone and the fatigue monitoring system in your car's cabin.

ArcSoft (688088) just released its semi-annual report for 2026: revenue reached 440 million yuan, up 7.19% year-on-year; net profit attributable to the parent company was 77.1384 million yuan, down 12.87% year-on-year; and net profit after non-recurring items dropped nearly 22%. With revenue rising and profits shrinking, where does this “scissors gap” come from? What is the company's competitive moat? In this article, we aim to clarify these questions.

01. Not Selling Hardware, but the 'Fundamental Skills' of Mobile Photography

What ArcSoft does can be summed up in one sentence: providing visual artificial intelligence algorithms for intelligent devices. Its business model is not about selling hardware or modules; instead, it “integrates” its self-developed visual algorithms into clients' end products through licensing, charging authorization fees on a project basis.

This model is characterized by extremely high gross margins. In the first half of 2026, the company's overall gross margin reached 89.03%, with the gross margin for mobile intelligent terminal visual solutions even higher at 91.33%. For every yuan of revenue, nearly nine mao is gross profit—almost unimaginable in the hardware supply chain but common in pure software licensing businesses.

Mobile intelligent terminals are ArcSoft's foundation. In the first half of the year, this segment generated 359 million yuan in revenue, up 5.78% year-on-year, accounting for over 80% of total revenue. Its client list includes global mainstream Android smartphone manufacturers such as Samsung, Xiaomi, OPPO, and vivo. In this field, ArcSoft is a leading global supplier of photography algorithms for Android phones, with a highly stable position.

However, the mobile phone market itself is struggling. According to third-party agency statistics, global smartphone shipments in the second quarter of 2026 fell 11% year-on-year. Rising memory chip prices further squeezed profit margins for mid-to-low-end models, leading clients to impose stricter requirements on algorithm memory usage, model size, and operational efficiency. ArcSoft's mobile business growth rate has slowed from double digits in previous years to single digits, reflecting industry-wide pressures rather than issues with the company's competitiveness.

02. 53% of Revenue Invested in R&D: Not 'Burning Money' but 'Placing Bets'

The core reason for the profit decline is stated bluntly in the semi-annual report: R&D expenses reached 235 million yuan, up 18.3% year-on-year, with the R&D expense ratio climbing to 53.41%, an increase of 5.02 percentage points year-on-year. In the second quarter alone, the R&D expense ratio surged to 57.7%.

What does this figure mean in the context of A-share tech companies? Most A-share software companies have R&D expense ratios roughly between 15% and 25%; exceeding 30% is typically considered “heavy R&D.” By investing more than half of its revenue in R&D, ArcSoft is making a clear strategic choice: trading current profits for future growth.

Where is the money being spent? Three areas. First, continuous iteration of “TurboFusion” technology for mobile devices, aiming to improve four scenarios most noticeable to consumers: portrait, telephoto, ultra-clear image quality, and video tone. Second, in-car AI vision, with DMS and OMS products moving from design wins to mass production, while the low-computing-power SouthLake platform is expected to begin large-scale shipments in the second half of the year. Third, emerging sectors, with multiple AI glasses projects entering mass production and shipments, and the core visual algorithm engine for robots completing adaptation to mainstream tech ecosystems such as NVIDIA, Intel, Qualcomm, and Rockchip, with deliveries already made to leading robot clients.

R&D personnel increased from 619 at the end of 2025 to 659, an addition of 40 people, with corresponding increases in salaries and social security expenses. This represents “upfront costs”—products are still in the validation and ramp-up phase, with revenue not yet fully realized, but labor and R&D investments have already been made. Huatai Securities, in a research report titled “Profit Pressure from Cyclical Industry Impacts,” gave a “Buy” rating while maintaining a long-term positive view on the steady growth of mobile intelligent terminal and in-car AI businesses.

03. In-Car Business: The Next Growth Engine Worth Watching

If ArcSoft is viewed solely as a “mobile algorithm company,” its potential would be underestimated. Revenue from in-car AI vision solutions reached 72 million yuan in the first half of the year, up 11.71% year-on-year. Although still small in scale, its growth rate has surpassed that of the mobile business.

More critically, its market position is improving. According to Gasgoo Auto data, among domestic passenger vehicle DMS (Driver Monitoring System) algorithm suppliers from January to November 2024, ArcSoft ranked third with approximately a 13% market share, behind only SenseTime Auto and Tesla, and significantly up from 8.3% in the same period in 2023 (10.7% from January to April 2025, still in the top three). Given that DMS products have low intrinsic value but high technical barriers, this is a market where market share is increasingly concentrated among top players. ArcSoft is gradually gaining share through its algorithm reuse capabilities, technical collaboration with sensor manufacturers like OmniVision and Sony, and deep adaptation with cockpit chip platforms like Qualcomm.

From client progress, DMS compliance products have secured design wins from multiple clients including VOYAH, Hyundai, and Chery, while OMS has added mainstream automakers like Geely, Xiaomi, and Great Wall. The intelligent parking assistance system has completed initial deliveries on the SouthLake platform, entering the commercialization phase. While domestic business growth is limited by vehicle price cuts and pressure from automakers, overseas cabin business progress is positive, with the company expecting European projects to scale up in the second half of the year.

04. Investment Logic and Risks: A Game of 'Timing'

The starting point of the investment logic is a clear timing gap. The core contradiction ArcSoft faces is that R&D investments and early-stage layout (layout) of emerging businesses have already occurred, but revenue realization still requires a one- to two-year ramp-up period. New directions like AI glasses, robot vision, and intelligent commercial photography (PSAI) are mostly in the stage of “transitioning from technical validation to mass production.” If these businesses scale up around 2027, the company's revenue structure and profit quality will undergo substantive changes; if scaling lags expectations, high R&D investments will continue to suppress profit performance.

The second logic lies in the scarcity of the business model. Pure algorithm licensing businesses, once integrated into a client's supply chain, have high switching costs and strong client stickiness. ArcSoft's top five clients account for roughly 60% of total sales (60.14% in 2024, 58.69% in 2025), representing both concentration risk and evidence of a competitive moat.

Risks cannot be ignored. The most direct is downstream market risk. The mobile business accounts for 80% of revenue, so fluctuations in global smartphone shipments directly impact ArcSoft's revenue. While the in-car business is growing rapidly, pricing pressure from domestic automakers is being passed upstream to algorithm suppliers. DMS products, already low in unit price, have limited room for price wars. Additionally, the expansion of new businesses is uncertain; the market timing for AI glasses and robot vision is not entirely controlled by ArcSoft, with client product definitions and market promotion efforts directly affecting algorithm procurement volumes. Furthermore, net cash flow from operating activities was -10 million yuan in the first half of the year, shifting from net inflow in the same period last year to net outflow. The divergence between profit and cash flow also warrants close monitoring.

Thus, the “rising revenue but declining profit” figure in the semi-annual report is less a sign of deteriorating performance and more a normal hiccup during a technological company's transition period. What truly merits tracking is not next quarter's profit figure but the timing of mass production orders for AI glasses, shipment rhythms of the SouthLake platform, and when robot vision clients transition from “technical cooperation” to “bulk procurement.”

- End -