[DigitalToday reporter Jinju Hong] Chinese artificial intelligence company Sensetime is moving to secure competitiveness with a strategy that highlights cost efficiency over high performance, led by its latest multimodal model. Amid U.S. sanctions and fierce competition in China's AI market, it is seeking a way to survive by focusing on cheaper operating costs and an enterprise-customer business rather than competing to build costly, ultra-large models.

CNBC reported on Tuesday that Sensetime showcased technology in its recently unveiled multimodal model SenseNova U1 that integrates and processes text, voice and image information within a single system. The company said reducing the steps needed to connect different modes improved speed and efficiency.

Founded in Hong Kong in 2014, Sensetime initially grew on facial and image recognition technology but has recently reorganised its business structure in line with the generative AI trend. It is under U.S. sanctions over allegations it was involved in surveillance of Muslim minorities in the Xinjiang region. The company denies the allegations.

Its co-founder and chief scientist, Lin Dahua (린다화), said the company is drawing on strategies that raise performance efficiently within limited resources, like DeepSeek. He stressed that while OpenAI's image-generation model produces more sophisticated results, SenseNova U1 costs about one-tenth as much. He added that most real-world work can be handled sufficiently without a top-tier model.

Sensetime sees competition with AI companies in China as a more realistic challenge than competition with U.S. companies. In China's market, DeepSeek, Moonshot AI, Alibaba and Xiaomi have been releasing new models in succession, intensifying competition. AI companies are simultaneously bearing burdens ranging from research and development costs and computing resources to securing semiconductors.

Profitability concerns are also growing for pure AI startups. Jefferies said in a recent report that independent AI companies face structural limitations including low customer loyalty, excessive competition and high training costs. In contrast, platform companies such as Alibaba, Tencent and ByteDance can expand AI businesses relatively more steadily by using their existing user bases and cash flows.

As a differentiation strategy, Sensetime is strengthening an enterprise-customer-focused business by bundling AI models, application services and infrastructure. The company cut its net loss by 58.6 percent last year and posted positive EBITDA in the second half for the first time since its 2021 listing.

The AI industry is also split on pricing policies. While some companies such as DeepSeek are cutting prices to expand market share, Zhipu is instead raising prices for premium models and focusing on monetisation. Cloud units at Alibaba and Baidu also raised prices in line with growing demand for AI computing.

Lin said price wars may work in the short term but differentiated value matters over the long term. He stressed that repeat purchases happen not because the technology is the best, but because the best service is provided at a competitive price.