SenseTime Says Cost-Efficient AI Can Compete With Global Frontrunners

CNBC reported Tuesday that SenseTime, the U.S.-sanctioned Hong Kong AI company, is wagering that cost-efficient AI models can still claim meaningful market share globally, even against better-resourced Western rivals.

Co-founder and chief scientist Lin Dahua told CNBC the firm is drawing strategic lessons from DeepSeek’s ability to deliver strong performance under tight financial and hardware constraints. The approach underpins SenseTime’s latest release, SenseNova U1, a multimodal system that processes text, audio and visual data within a single architecture. Removing the translation step between different data types boosts both speed and efficiency, the company said.

Price Gap as Competitive Edge

Lin acknowledged SenseNova U1 falls short of OpenAI’s image-generation tools on output quality. But the cost differential is stark: SenseTime’s model runs at roughly one-tenth the price, he said. Lin argued that most enterprise tasks do not require frontier-level output, making affordability a decisive factor for clients choosing between providers.

SenseTime has also responded nimbly to domestic rivals. ByteDance’s AI video model initially posed a threat, Lin noted. SenseTime subsequently folded some of that competitor’s background-generation capabilities into its own short-video product, pairing them with in-house audio features.

A Crowded Market With a Platform Problem

The broader challenge facing standalone AI model companies is structural. Investment bank Jefferies flagged in a late-April note that pure-play AI firms face low customer loyalty, limited differentiation and punishing training costs. Even OpenAI has reportedly missed internal revenue and user targets, signaling industry-wide pressure.

Platform giants in China carry natural advantages. Alibaba, Tencent and ByteDance can cross-subsidize AI development using cash flows from their core businesses and tap vast proprietary data pools. Independent AI firms lack that buffer, leaving them loss-making for longer.

SenseTime’s Road to Profitability

Founded in Hong Kong in 2014 and historically known for facial recognition, SenseTime has reshaped itself for the generative AI era. The firm targets enterprise customers specifically, a segment that demands higher service quality, tolerates premium pricing and rarely switches vendors once embedded.

That focus is producing early financial results. SenseTime cut its net loss by more than 58% last year and reported positive EBITDA in the second half of 2025 for the first time since its 2021 listing. The company is also pressing ahead with international expansion, including unchanged plans in the Middle East, despite ongoing U.S. sanctions over allegations tied to surveillance in Xinjiang, which SenseTime has denied.

Investors will watch whether the cost-efficiency playbook can convert into sustained profitability before cash reserves thin.

Read Next: DeepSeek’s Efficiency Push Rewrites the AI Cost Playbook