CNBC's The China Connection newsletter: What chip shortage? 10 questions with Temasek-backed AI startup unfazed by U.S.-China tensions

CNBC

28-05-2025

Hearing about artificial intelligence from Chinese businesses paints a very different picture of how the tech can be applied than is commonly understood.
After a stint as a data scientist at Meta, Caltech alumnus Jerry Ye returned to China in 2017 at the behest of an investor, with the goal of starting a global AI company.
The startup, Whale, sells AI software tools and related hardware to help retailers. It now claims clients such as Starbucks, Xiaomi and Unilever. Last week, Whale announced that it had raised $60 million in a Series C funding round that included Bosch Ventures, Singtel and MDI Ventures. The startup counts Hangzhou as its China headquarters and Singapore as its global base.
I caught up with Ye in Beijing this week as he passed through the city, where he has a small team. This interview has been condensed and edited for clarity.
What's the impact of the U.S.-China trade war?
We still think there's a global opportunity even if there are U.S.-China tensions. Business is business. Our customers are global: Procter & Gamble, Watsons, Starbucks.
Who will win the AI race?
The largest two markets are still the U.S. and China. The application layer will take off in both, in different ways: marketing and sales in China; very, very small vertical sectors in the U.S., such as in transportation or legal. Southeast Asia is more likely to adopt scenarios from the U.S. or China markets.
What's restricting AI development? Advanced semiconductors?
For China, we don't see any shortage in AI computing power. All the cloud [companies] have a lot of training power. A lot of AI companies don't actually train [models] a lot anymore.
Ultimately, on-device chips [often associated with edge computing, or processors built into devices like smartphones, home appliances, etc] are becoming powerful enough to run AI models and can ensure data privacy. In the long run, most of the competition will happen on the edge, instead of at data centers.
Where will AI make money for businesses first?
Marketing and sales, because they are very close to the money, and the return on investment is easier to calculate. If you adopt AI in marketing and sales over seven days, on the eighth day, you will see results.
What does that mean for the future of content?
The basic concept of marketing is communication, narrowing the gap between the customer understanding the product and the brand selling the product.
We talk a lot about personalization. But traditionally, we didn't have sufficient content, sufficient power to give each person personalized content. With AI's richer understanding and algorithms to bridge the content and the customer, the decision-making is faster and easier.
Generative AI definitely gives us a lot of opportunities to generate a huge amount of videos.
AI also enables us to send unique slices of live-shot photos and videos to social media. This has become the way ads are increasingly produced.
But if you use a lot of generative AI, they convey a lot of fake information. Most platforms are mostly anti-generative AI videos. My personal opinion is that real-life videos are still needed. It's very hard for generative AI to [convey] real emotions, real human touch.
How are companies using AI today?
Our company services Starbucks in China and Southeast Asia.
Each time you purchase a Starbucks coffee, huge amounts of AI are used to make the customer experience better. Right now, consumers can get their coffee order within 5 minutes, 98% of the time. If one location is too busy to handle online orders on top of in-store orders, AI analysis may send those online orders to a different location nearby.
How much money is Whale making? What are your costs?
Over the last three years, we have grown by about 100% [a year], and for the next two to three years, we will maintain 100% annual growth. We made revenue of 230 million yuan ($31.39 million) last year, and expect 400 million yuan in revenue this year. We have around 500 enterprise customers.
We spend about 1 million yuan a month in China on operating costs — outside China, it's more expensive, so the total is around 3 million yuan a month [36 million yuan a year].
Our biggest challenge is complying with local data privacy and security needs in every region.
Which AI large language model is best?
Different models have different capabilities. For Claude, they are really good at generating websites, PDFs. Gemini, video. It's like having PhDs in different domains. For most of our business customers, they don't want to be fixed into one model.
A lot of businesses want to keep their data private. That's why they use open-source models, deployed in their own data centers.
Is there potential for an AI device, such as glasses?
The biggest challenge for most hardware is the battery, not the AI itself.
Most of the hardware has been here for a long time.
Maybe a new cellphone. The most important thing is the application. The application will take over the large margin of the AI. Cloud has 5% to 10% of the margin, then hardware. There's a huge software layer [of which AI applications] can take 90% to 95% of the margins.
What do you think about former Apple designer Jony Ive's deal with OpenAI?
It will be hard. There's a camera on it. It's good to have large companies trying different AI. They have more potential to try different things.
Hong Kong will remain the most prominent destination for large Chinese companies to go public: Yiyi Capital
Theodore Shou, CEO of Yiyi Capital, discusses the recent strong performance of the IPO market in Hong Kong and says that the valuation premium between mainland China and Hong Kong stocks would narrow if more Chinese companies list on the Hong Kong Stock Exchange.
EV transition has sparked 'a tsunami of demand' for rare earths, Aclara Resources CEO says
Ramón Barúa Costa, CEO and director of Aclara Resources, discusses global demand for rare earths, China's dominance in critical minerals, and the strategic importance of rare earths in the energy transition.
HK debutant Mirxes says China-Singapore 'dual engine' will drive growth amid global changes
Lihan Zhou, CEO and Co-Founder of Mirxes, a Singapore-based biotech company, discusses the company's Hong Kong IPO and growth prospects.
The U.S. and China are talking. Chinese Vice Foreign Minister Ma Zhaoxu and U.S. Deputy Secretary of State Christopher Landau had a call last week, and both sides agreed to keep lines of communication open, according to readouts on Friday. While no specifics were announced, the call was the first since the U.S. and China agreed this month to pause most tariff increases.
China's industrial profits rose in April. Official figures out Tuesday showed profits climbed 3% year on year in April at major industrial companies, faster than the 2.6% recorded in March and growing despite U.S. tariffs. But margins at electric car companies are facing more pressure after BYD this month announced a slew of discounts.
Xiaomi targets Tesla and Apple at the same time. The Chinese smartphone company on Thursday revealed a new chip that it claims beats Apple's on certain metrics — and comes inside a phone far cheaper than the iPhone 16 Pro. Xiaomi also teased a new SUV called the YU7, which analysts predict will amp up the competition for Tesla in China.
Chinese and Hong Kong stocks traded mixed Wednesday as investors continued to assess the U.S.-European Union trade optimism.
Mainland China's CSI 300 added 0.11% while Hong Kong's Hang Seng Index — which includes major Chinese companies — slipped 0.4% as of 11 a.m. local time.
The benchmark 10-year Chinese government bond yield is at 1.672%.
The offshore Chinese yuan last traded at 7.1971 against the greenback.
—
May 31: Official Purchasing Managers Index for May
June 2: Mainland China markets closed for the Dragon Boat Festival holiday
June 3: Caixin China General Manufacturing PMI for May