Kuaishou

Sanctioned SenseTime: Cheaper AI Models Can Win

SenseTime, a U.S.-sanctioned Hong Kong-listed AI firm, is pitching cost-efficiency as its competitive edge in the global generative-AI race. Cofounder and chief scientist Lin Dahua told CNBC the company's multimodal model SenseNova U1 integrates language and vision, improving speed and efficiency while costing significantly less than some frontier international image models. SenseTime has narrowed losses (net loss down 58.6% year-over-year) and posted positive EBITDA in the second half, and is pursuing international expansion across Southeast and North Asia, the Middle East and Brazil despite export restrictions. The company has integrated third-party capabilities (e.g., ByteDance’s Seedance) into products like its short-video tool Seko, and argues lower-cost, “good-enough” models can capture market share where top-tier models are not necessary.

Bytedance Raises Prices, Reshaping China AI Video Market

Following Sora’s shutdown in late March 2026, Bytedance’s Seedance 2.0 — surfaced via its Jimeng consumer app — rapidly raised prices and tightened usage limits, dramatically increasing per‑clip costs, queue times, and enterprise access thresholds. The moves exposed the underlying compute economics of AI video: generation is token‑intensive (roughly 300,000 tokens for a 15‑second clip) and previous subsidized pricing was unsustainable. Kuaishou’s Kling AI had already built substantial revenue, and a new challenger is preparing to contest the market, shifting competition from pure model performance to infrastructure and unit economics.

Alibaba Leads $290M Investment in World-Model AI

Alibaba Cloud led a 2 billion yuan (about $290 million) Series B investment in ShengShu, the startup behind the AI video-generation tool Vidu, to develop a "general world model" that uses multimodal data (vision, audio, touch) to better model real-world behavior. TAL Education and Baidu Ventures also participated. ShengShu—which raised 600 million yuan two months earlier from Qiming Venture Partners and others—says the funding will help bridge digital domains (games, AI video) and physical domains (autonomous driving, robots) by connecting perception and action. ShengShu’s Vidu Q3 Pro ranks among the top 10 video-generation models, according to Artificial Analysis. The story positions Alibaba’s investment as part of a broader push into world-model research and related startups such as Tripo AI and PixVerse, highlighting competition with other Chinese video-AI efforts and the importance of embodied AI for robotics.