SenseTime Says Cost-Efficient AI Can Compete With Global Frontier Models

CNBC reported Tuesday that SenseTime co-founder and chief scientist Lin Dahua believes lower-cost AI models can still claim significant market share, even where quality gaps with frontier systems persist.

SenseTime’s Cost-First Strategy Takes Shape

The Hong Kong-listed company, long associated with facial and image recognition, has repositioned itself around generative AI. Its newest release, SenseNova U1, processes text, audio and visual inputs within a single unified architecture. That design eliminates the translation step between data types, cutting both latency and operating costs.

Lin told CNBC that SenseNova U1 runs at roughly one-tenth the cost of OpenAI’s image-generation tool, even while acknowledging a performance gap with frontier models. His argument is straightforward: most commercial use cases do not require peak-tier output, and price sensitivity matters more than marginal quality gains for many enterprise buyers.

SenseTime has drawn direct inspiration from DeepSeek’s approach of engineering high-performing systems under tight resource constraints. That philosophy now underpins SenseTime’s own development roadmap.

Background: A Company Navigating Sanctions and Fierce Domestic Rivalry

Founded in Hong Kong in 2014, SenseTime carries U.S. sanctions tied to allegations of surveillance technology use against Muslim minorities in Xinjiang. The company has denied those allegations. Despite the restrictions, it has continued expanding, with Middle East growth plans reportedly unchanged.

Domestically, competition has sharpened quickly. ByteDance, Alibaba, Moonshot AI and even Xiaomi have all launched or updated models in recent weeks. ByteDance’s Seedance video model was an early concern for SenseTime, Lin said, though the firm has since incorporated compatible capabilities into its own short-video product.

Platform giants carry structural advantages here. Firms like Alibaba, Tencent and ByteDance can offset AI development costs against their core consumer businesses, analysts at investment bank Jefferies noted in a late-April research note. Pure-play AI companies face the opposite pressure: high training costs, low customer loyalty and a crowded competitive field.

Narrowing Losses Signal a Turning Point

SenseTime’s financial picture is improving, albeit from a difficult baseline. The firm cut its net loss by nearly 59% last year and posted positive EBITDA in the second half of 2025 for the first time since its 2021 listing. That trajectory matters because investors are scrutinising AI monetisation across both Chinese and American players.

Enterprise clients form the core of SenseTime’s revenue strategy. These buyers tolerate higher price points, demand service consistency and are less likely to switch vendors on short notice. That stickiness gives the company a more durable revenue base than consumer-facing AI products typically provide.

Read Next: DeepSeek’s Cost Discipline Is Reshaping How AI Companies Think About R&D