Hi everyone, this is Lauly helming #techAsia this week. I’m writing this newsletter next to a rice paddy in the northern Taiwanese city of Taoyuan, where I have attended the opening of UPS’s largest logistics center in the Asia-Pacific.
The U.S. company didn’t pick this location for its 81,000 sq. meter, nearly $100 million facility because of its proximity to rice paddies, of course. Rather, it is only about 5 kilometers from the Taiwan Taoyuan International Airport.
Lauren Zhao, president of Asia-Pacific supply chain solutions for UPS, said Taiwan’s strategic position in the global high-tech supply chain was the driving force in the company’s decision to deploy a more comprehensive network here. It was interesting to learn that Applied Materials of the U.S., a critical supplier of chip production equipment to TSMC, is already the largest single customer for this logistics center — its goods occupy nearly half of the four-story building. The chip equipment supplier can access the logistics center to retrieve its components and parts 24 hours a day to serve its customer.
UPS said it also plans to open a center in the southern Taiwanese city of Kaohsiung later this year, which could help customers — again, mostly from the tech sector — dispatch goods quickly to the airport in Taoyuan for delivery the next day to the U.S.
This was not the only example this week of a company doubling down on investments in Taiwan. Air Liquide, Europe’s top industrial gas maker and a TSMC supplier, opened its first chip materials facility in Taiwan in the central city of Taichung to address the increasing demand for AI chip production. Taiwan’s Chief Telecom, the island’s top data center operator, also announced it will spend 3 billion New Taiwan dollars ($90 million) to build an additional AI data center at the Taichung Science Park, as it expects its current one — which just opened last summer — to run at full capacity next year or the year after.
I still remember the grim atmosphere in late 2022 and 2023, when many large tech companies asked Taiwanese suppliers to prepare alternative capacity outside the island due to fears of tensions across the Taiwan Strait. While many Taiwanese tech suppliers are acceding to such requests, as can be seen in the rise of Thailand as a printed circuit board hub, it is very encouraging to see companies continuing to invest and to show confidence in Taiwan, which should help secure the island’s strategic importance on the global stage.
Of course, not everything is rosy these days. I attended Asus’s launch event in Taipei for its new AI PC. What surprised me was not the product itself, with its Qualcomm CPU, but its price tag. When Asus first introduced its AI PC Zenbook 14 OLED in 2024, the starting price was NT$40,900, or $1,280. Guess how much the latest model is. NT$63,999, or $2,000.
As the cost of memory chips surges, brands are starting to pass some of the pain on to consumers. It’s no wonder many of my sources in the PC supply chain sound pessimistic about the outlook for the rest of this year, as they are afraid the rising price tags will scare away consumers.
Crunch time
The PC and server industries are facing a fresh blow. On top of the persistent memory chip shortage, a serious supply crunch is emerging in central processing units (CPUs) made by Intel and AMD, according to this exclusive story by Nikkei Asia’s Cheng Ting-Fang and Lauly Li.
So far this year, quote prices for CPUs — the heart and brain of electronics devices — have increased by an average of between 10% to 15%, and some even higher, said multiple sources with direct knowledge of the matter. Intel and AMD have recently told clients they will increase prices for all series of CPUs from March and April, respectively, they said.
Wait times for CPUs have also lengthened significantly, from one to two weeks to an average of eight to 12 weeks, while in some extreme cases extending to six months, the sources said.
The stronger-than-expected demand for general servers and storage servers amid increasing demand for AI inference was a key reason behind the crunch, as Intel and AMD have both prioritized their capacity to address this demand, further squeezing supplies for PC use.
An exception to the rule
SoftBank is testing a key self-imposed borrowing limit as it commits a further $30 billion to OpenAI, a move that risks unsettling investors already wary of its growing exposure to AI, writes the Financial Times’ David Keohane.
The Japanese group aims to keep its loan-to-value ratio — a measure of net debt against the value of its holdings — below 25% in normal conditions. But it now accepts the threshold could be exceeded in the coming months as it increases AI investment.
“I don’t deny the possibility in the future that we may go temporarily beyond 25%,” Chief Financial Officer Yoshimitsu Goto told the FT.
Although the group is adamant it will not change its policy, the move marks one of the clearest signs yet that SoftBank is ready to push up against its own financial guardrails to back OpenAI, despite mounting investor unease over the costs, competition and uncertain returns of the AI boom.
“There’s [an estimated] $50 billion of funding, between OpenAI, investments and refinancing, that they have got to put in place in the course of 2026,” said David Gibson, an analyst at MST Financial. “The loan to value will hit 25% or more. So to me that’s the story as I’m not sure the market is prepared for it.”
SoftBank’s LTV has already increased from 16.5 to 20.6% in the quarter to December. Shares have fallen more than 45% since last October as investors grow more cautious about its AI exposure, particularly as OpenAI faces intensifying competition from Google and Anthropic.
Goto said that, if needed, SoftBank would bring its loan-to-value ratio back below 25% as quickly as possible through a combination of asset sales, listings and asset-backed financing.
Keeping the pace
Taiwan’s leading chip substrate supplier Kinsus is betting big on expanding its capacity and upgrading its technology to keep pace with the needs of Nvidia and TSMC in the race to capture surging AI demand, the company’s chief executive told Nikkei Asia’s Lauly Li and Cheng Ting-Fang in an exclusive interview.
Kinsus, which is also a supplier to AMD and Micron, is embarking on its largest ever, multiyear expansion plan, looking to challenge rivals at home and abroad.
The company’s CEO, Scott Chen, said Kinsus is closely watching the new fab construction and technological progress of top chipmakers like TSMC to plan Kinsus’s own expansion pace. The substrate supplier plans to build a new factory every two to three years.
Chip substrates are the critical materials on which chips are mounted before placing them on the printed circuit boards. There are only a handful of suppliers in the world — mainly from Taiwan and Japan — that can produce reliable, high-end chip substrates at scale. Kinsus is among only three chip substrate suppliers verified by Nvidia and TSMC for use in the contract chipmaker’s advanced chip packaging technology CoWoS, which in turn is used to make Nvidia’s GPUs.
Designs for growth
Chinese e-commerce giant Alibaba Group earlier this week unveiled its latest self-developed central processing unit (CPU) design for artificial intelligence applications, as the company races to catch up with surging domestic demand for AI agents sparked by OpenClaw, write Nikkei Asia’s Cissy Zhou, Cheng Ting-Fang and Lauly Li.
Alibaba’s latest chip, XuanTie C950, is designed based on open-source RISC-V architecture and aimed at offering developers and startups cheaper and more flexible AI solutions, according to the company.
The company said at an event for its DAMO Academy ecosystem that the chip has been verified for 5-nanometer chip production technology. Alibaba did not disclose which contract chipmaker would produce the chip once it progressed to the production stage, but multiple sources told Nikkei Asia that it would be made by Taiwan Semiconductor Manufacturing Co., its longtime production partner.
Suggested reads
1. China reviews $2bn Manus sale to Meta as founders barred from leaving country (FT)
2. Samsung SDI’s Hungary woes cloud PM Orban’s reelection bid (Nikkei Asia)
3. Daughter of SoftBank’s Son to take over troubled Japanese unicorn Spiber (Nikkei Asia)
4. AI is moving from answering questions to taking action (FT)
5. SK Hynix files for US listing to fund AI chip-driven expansion (Nikkei Asia)
6. US must suspend Nvidia AI chip exports to China, senators say (FT)
7. Taiwan self-driving startup eyes Japan as launchpad for niche systems (Nikkei Asia)
8. Supermicro drops 33% after co-founder charged with smuggling Nvidia chips to China (FT)
9. Nintendo wins by not playing the power game (FT)
10. Bhutan’s hydropower offers big draw for Japanese AI investment, says PM (Nikkei Asia)
Podcast: Tech Latest
Nvidia aims to claw open the next frontier in AI
Welcome to the Tech Latest podcast. Hosted by our tech coverage veterans, Katey Creel and Shotaro Tani, every Tuesday we deliver the hottest trends and news from the sector.
In this episode, Katey speaks with Silicon Valley correspondent Yifan Yu about how the environmental impact of manufacturing AI chips and servers is coming under increasing scrutiny.
Find us on Apple Podcasts | Spotify | YouTube | YouTube Music | Amazon Music | Voicy
For more great stories like this delivered to your inbox every week, sign up to our #techAsia newsletter. Current subscribers, click here to update your newsletters preferences.
