Text/Yang Jianyong

In the first half of 2026, CloudWalk's revenue dwindled to RMB 100 million, marking a 40.67% year-on-year decrease and resulting in a loss of RMB 78.77 million. Although the loss has narrowed somewhat, the substantial revenue decline underscores the company's struggle with business contraction.

The Generative AI Era: SenseTime Achieves Profitability While CloudWalk Continues to Struggle

Once a prominent figure among the top four AI companies, CloudWalk garnered significant attention, with its revenue exceeding RMB 1 billion in 2021. However, in the era of large models, its revenue has fluctuated, and persistent losses have yet to show signs of improvement. Notably, the previous expectation of achieving profitability by 2025 has proven to be an unfulfilled promise.

In stark contrast, SenseTime has exhibited robust growth and achieved profitability. In the first half of 2026, it is anticipated that SenseTime will record a profit ranging from RMB 500 million to RMB 700 million, successfully turning a profit compared to a loss of RMB 1.489 billion during the same period last year.

It is noteworthy that SenseTime, too, faced the dual challenges of declining revenue and significant losses in the past. The turning point came with the advent of ChatGPT, which ushered in the era of generative AI. Facing the industrial development opportunities presented by the new wave of AI large models, SenseTime reorganized its business in 2023 to better adapt to the generative AI era.

Today, the generative AI business has emerged as the core growth engine for SenseTime's revenue. In 2025, SenseTime's overall revenue reached RMB 5 billion, representing a 33% year-on-year increase. Among this, revenue from the generative AI business surged to RMB 3.629 billion, a 63.7% year-on-year increase, with its revenue share further rising to 72.4%. This not only establishes it as the largest business segment but also underscores the effectiveness of the company's strategy to transition to generative AI.

With the realization of economies of scale in the generative AI business and changes in the fair value of AI ecosystem companies built by SenseTime, profitability is expected to further improve. The anticipated positive IFRS consolidated profit for the first half of 2026 dispels market concerns about long-term losses and lays a solid foundation for future comprehensive profitability goals.

Facing the wave of large models, CloudWalk has further iterated and upgraded its human-machine collaborative operating system CWOS to Growth OS, promoting an AI infrastructure + AI agent business architecture and advancing its strategic layout in areas such as controllable computing power bases, content safety large models, industry agents, and embodied intelligence. However, the business scale has yet to fully materialize. In particular, the significant decline in revenue and persistent losses starkly contrast with SenseTime's successful turnaround to profitability.

Computing Power Base: The Pillar Supporting SenseTime's Transition and Profitability in Generative AI

The key difference between SenseTime and CloudWalk lies in SenseTime's establishment of a vast AI infrastructure, forming a computing power infrastructure moat that underpins the development of its generative AI business.

As large model technology continues to evolve, the essence of industry competition has shifted to computing power competition within AI infrastructure. Since 2018, SenseTime has been constructing a new generation of AI infrastructure, SenseCore, forming the foundation for its trinity strategy of large devices, models, and applications.

As of March 2026, the total operational computing power of the SenseCore platform has reached 40,400 petaFLOPS, with a daily average of 2.42 trillion Token services. It is projected that the service scale for the entire year of 2026 will grow 25-fold.

Meanwhile, with the global surge in generative AI and the continuous iteration of AI large models, their capabilities have become increasingly formidable, and their applications have transitioned from initial knowledge question-and-answer modes to Agent AI. The application of Agents is entering a new phase. The SenseCore platform provides robust infrastructure support for the Agentic AI era.

Overall, the rise of generative AI technology and the significant changes brought about by large models have gradually unlocked market demand potential. SenseTime has capitalized on this market opportunity with the advantage of its AI infrastructure, SenseCore, driving comprehensive revenue growth. This also signifies its successful transition from traditional AI to generative AI, initiating a new growth cycle.

Finally

With continuous technological breakthroughs in large models, they are accelerating the empowerment of industrial upgrading, and various industries are rapidly integrating large model services. Both the consumer (C) and business (B) ends are exhibiting significant commercial potential.

Leveraging its large model infrastructure, SenseCore, and the DayDayNew multimodal large model, SenseTime deeply empowers Agent AI applications, driving growth in its generative AI business. This showcases its unique path to industrialization and unlocks new growth potential.

Yang Jianyong, a contributor to Forbes China, expresses views that are solely his own. He is dedicated to providing in-depth interpretations of cutting-edge technologies such as AI large models, artificial intelligence, the Internet of Things, cloud computing, and smart home appliances.