A sanctioned Chinese firm says cheaper models can still win
China’s AI race has no finish line. DeepSeek, Moonshot AI, Alibaba’s and even consumer electronics company Xiaomi In recent weeks, they have all released new models and competed for a place on the leaderboard.
From local AI startups to platform giants, companies in the industry face increasing pressure to innovate, expand their user bases and find ways to generate revenue. They also need to manage high research and development costs, as well as increasing spending on computing power and hardware.
SenseTime, one of China’s first artificial intelligence companies, has taken action to keep up with the era of productive artificial intelligence. The company, long known for facial and image recognition, is now developing multi-modal systems that can combine text, voice and visual data.
SenseTime, founded in Hong Kong in 2014, has faced US sanctions over allegations it spied on Muslim minorities in Xinjiang, but it has denied the allegations.
Its latest model, the SenseNova U1, combines language and image processing into a single system, eliminating the need to dial in different modes, increasing speed and efficiency.
SenseTime relies on cost efficiency as a competitive advantage. According to co-founder and chief scientist Lin Dahua, the company took cues from DeepSeek’s approach to delivering high-performance models under financial and technological constraints.
Co-founder and chief scientist Lin Dahua at SenseTime’s offices in the Hong Kong Science Park.
CNBC
Lin said OpenAI’s ChatGPT Images 2.0, an AI tool that generates images from text prompts, produces “excellent and beautiful” results, but SenseNova U1 costs ten times less.
“In many cases, you may not need the top model that can handle most tasks,” Lin told CNBC. “There’s still a gap between us and international leading models like OpenAI’s GPT Image 2 and (Gemini’s) Nano Banana, but our cost is much lower; it’s very efficient.”
With limited overlap between the US and Chinese AI markets, the real competition may be closer to home.
ByteDance’s AI video model, Sedance, initially raised competitive concerns, Lin said. SenseTime has since integrated some of its capabilities into short video tool Seko, allowing Sedance to combine background rendering with its own audio functions.
More than a model race
With business models becoming increasingly important, technology is only half the battle. OpenAI’s lack of reported revenue and user targets Wall StreetJournalThis situation points to risks for both Chinese and American players, Jefferies said in a note dated April 28.
Pure AI model companies face a challenging equation: low customer loyalty, limited differentiation, a crowded field and high training costs, Jefferies said.
The bank added that large internet platforms, by contrast, have stronger cash flows, access to user data and established customer bases to sell AI applications.
Including platform companies in China Alibaba’s, Tencent and ByteDance can use its core businesses to support AI development and enhance existing operations, said Vey-Sern Ling, senior equity advisor at UBP.
A Sensetime booth at the World Artificial Intelligence Conference 2021 in Shanghai, China, on July 7, 2021.
Cphoto | Future Publishing | Getty Images
“Frankly, they are in a better position than the independents who continue to make losses,” Ling added, noting that heavy AI spending is weighing on profits even at larger players such as Alibaba and Kuaishou.
Lin said SenseTime combines large artificial intelligence models, applications and infrastructure to differentiate itself, improving service quality while reducing costs per use. Many of its products are aimed at corporate customers, who generally demand higher quality services and are increasingly willing to pay more and less likely to switch providers.
SenseTime narrowed its net loss by 58.6% last year and reported positive second-half EBITDA for the first time since listing in 2021; It’s a trend that investors will continue to monitor closely. Lin said the company’s AI costs are “manageable” and are largely focused on making models more efficient.
Price to win or win at price
Pricing strategies vary by industry.
Some companies, including DeepSeek, have recently lowered prices and offered discounts to attract users. like others Zhipu He brought them up, signaling a push to commercialize advanced models.
Alibaba’s cloud units and baidu It has also increased prices due to increased demand for AI computing power. ByteDance is planning a subscription service for certain features of its popular AI chatbot Doubao.
“Price wars can serve a strategic function in short-term promotions, but long-term sustainability depends on differentiated value,” Lin said.
Given China’s huge market, analysts say some AI companies may be following a familiar strategy: They pour cash to gain market share before raising prices to make money later.
“They can’t continue to subsidize the use of AI because it’s too expensive,” UBP’s Ling said.
“They can either paint a picture of the huge future usage and demand and help investors understand that short-term losses are acceptable. Or they should start making money much earlier.”
Betting on the world beyond Washington
Faced with US export and investment restrictions, SenseTime has focused its international expansion on markets such as Southeast and North Asia, the Middle East and, most recently, Brazil.
The U.S.-Israeli war on Iran has caused short-term disruptions affecting flights and interactions, but Lin said the company’s long-term strategy in the region remains unchanged.
Cost efficiency and practical benefit are equally important in overseas markets.
“Most of the time, the reason for repeat purchase is not because the technology is particularly advanced, but because it offers the best service at a competitive price,” Lin said.
