A ‘Made in USA’ win and the race for Nvidia graphics cards

Hello everyone! This is Lauly, sending my greetings from summery Taipei.

It’s been a very fruitful few weeks for me. Although I caught a bad cold during the annual Computex trade fair in mid-May, I was able to talk to numerous suppliers about their latest AI server solutions, from advanced liquid cooling systems and power supplies to cables, GPU boards and final assembly. Being able to touch all those components and parts, and hear directly from experts how these technologies come together to make the world’s most powerful AI supercomputers was a priceless experience.

That being said, the highlight of the past few weeks for me was attending Lady Gaga’s Mayhem concert in Singapore immediately after Computex wrapped up. Her performance was so powerful and her passion for music and art so infectious that I ended up clapping too hard and breaking my iPhone camera lens. It took me some hours to realize that was why all my photos of Gaga were so blurry.

After the show, my friend and I were stuck in front of the National Stadium MRT station for more than 30 minutes with tens of thousands of other fans leaving the venue. But the station staff clearly had a lot of experience handling these situations. One of them even led the crowd in singing Gaga’s songs, making the wait to get into the station much more bearable.

Back at work, I had a precious opportunity last week to join a small group of reporters on a tour of two Yageo manufacturing facilities in the southern Taiwanese city of Kaohsiung. Yageo is the world’s second-largest maker of passive components after Murata of Japan and counts Apple, Nvidia, HP, Dell and Tesla among its customers.

I was fascinated by how clean and automated Yageo’s plants are. I’ve heard a lot about passive components and was thrilled to see and touch the tiny components — the smallest one is smaller than a grain of ground pepper. They are so small that you can hold 3 to 4 million resistors in your hand at a time, and yet they are vital for everything from smartphones and notebooks to servers and cars.

Yageo is keen for people outside of Taiwan to know more about the company. It is trying to acquire a smaller peer, Japan’s Shibaura Electronics, a move that Yageo Chairman Pierre Chen believes would be a “win-win” for both sides.

Speaking of “wins,” U.S. President Donald Trump can look forward to some success in his push to onshore tech production: AI servers, but not iPhones, at least not anytime soon.

I recently met a longtime source, a former executive at an iPhone assembler, and we talked about why it is so challenging to make those devices in the U.S.

The source gave this example: Mounting the Wi-Fi module into an iPhone requires first inserting the tiny module — by hand — through the side of the phone’s metal frame, holding it in place on the printed circuit board, and then finishing the process by dispensing a small drop of glue onto it.

“I don’t see any automation tool to soon replace that process, let alone bringing the entire iPhone production flow to the U.S. … It’s more impossible than Tom Cruise’s Mission: Impossible,” the source said.

I agree with him.

AI over iPhones

While U.S. President Donald Trump has repeatedly called on Apple CEO Tim Cook to bring iPhone manufacturing home, it is the rapid formation of an AI server supply chain that may be his first “Made in the U.S.A” tech triumph, write Nikkei Asia’s Lauly Li and Cheng Ting-Fang.

Nvidia, the world’s AI chip leader, in April announced plans to invest $500 billion to make supercomputers in the U.S. Since then, at least eight of its suppliers have said they will build new capacity in the country.

Foxconn, the world’s biggest server builder and a key Nvidia AI supplier, is furiously expanding capacity in Houston, Texas, for the American company. The Taiwanese contract electronics giant plans to bring a significant portion of its production of graphic processing unit (GPU) modules and board to the U.S. for the first time, according to sources. TSMC’s $165 billion Arizona expansion will also play a crucial role in building a U.S. AI server supply chain.

Ren speaks

Huawei founder Ren Zhengfei broke his silence on the company’s artificial intelligence chip efforts in a rare interview with Chinese state media earlier this week.

Speaking to the People’s Daily, the reclusive executive said the U.S. has “exaggerated Huawei’s capabilities,” adding that the company “is not that strong yet,” writes the Financial Times’ Eleanor Olcott.

Ren’s comments mark a rare public intervention from the Huawei chief, who has largely withdrawn from the spotlight in recent years and had not previously addressed the company’s AI chip series, Ascend, viewed as China’s closest rival to Nvidia.

His remarks come amid growing concerns from Nvidia chief Jensen Huang, who recently warned that Huawei is emerging as a “formidable” competitor. Huang argued that Washington’s restrictions on Nvidia’s chip exports to China have inadvertently helped bolster the Chinese group’s position, threatening America’s lead in AI technology.

While Ren acknowledged that Huawei’s chips lag behind U.S. counterparts in performance, he suggested the gap could be bridged by clustering multiple chips in a server. He also emphasized China’s strengths in the AI race, citing its abundant energy supply, vast population and advanced telecoms infrastructure.

Cat-and-mouse industry

Ukraine’s successful strikes against Russian strategic bombers using inexpensive drones on June 1 highlighted how unmanned aerial vehicles (UAV) are reshaping modern warfare. Also known as drones, UAV are also emerging as a significant global business.

Japanese motorcycle maker Kawasaki Motors and carmaker Subaru are two examples of newcomers looking to foray into the rapidly growing market, Nikkei Asia’s Mitsuru Obe writes. Kawasaki Motors, which is working with French startup VoltAero, aims to produce 5,000 drone engines a year by 2030. “When it comes to small and efficient engines, no one can beat Japanese motorcycle makers,” Kawasaki said.

Australia’s DroneShield, on the other hand, is growing rapidly on the back of demand for counter-drone defenses, writes Nikkei Asia’s Shaun Turton. From some 90 employees two years ago to 275 today, the startup founded in 2014 is now an ASX-listed company with a market cap of 1.3 billion Australian dollars ($848 million). Its DroneShield systems are deployed in dozens of countries, including Ukraine.

A race against time

Leading gaming PC makers and Nvidia graphic cards suppliers MSI and Gigabyte are rushing to produce and ship as many products as possible to the U.S. market before the Trump administration’s grace period on tariffs expires on July 9, Nikkei Asia’s Lauly Li writes.

Joseph Hsu, chairman of MSI, said the Taiwanese company has been building up stock for the U.S. market ahead of the escalation of the tariff war in April. But because the newest products became available only around April, there was little they could do beforehand.

“Take Nvidia’s newly launched graphics cards for instance. However many we ship to the U.S., they are immediately sold out, so it is hard to build inventory there,” Hsu said. “We are racing against time.

Dandy Yeh, chairman of Gigabyte, which is also an AI server supplier for Nvidia, said his company received rush orders last month as clients in the U.S. wanted them to ship more products due to mounting tariff uncertainties. The tariff war and the White House’s unpredictable policy, he said, have brought uncertainty to the entire tech industry, and especially for its export-oriented companies.

Nintendo looks to break hit-and-miss cycle with Switch 2 launch

Welcome to the Tech Latest podcast. Hosted by our tech coverage veterans, Katey Creel and Akito Tanaka, every Tuesday we deliver the hottest trends and news from the sector.

In this episode, Akito speaks with Tokyo correspondent Tamayo Muto about Nintendo’s recent launch of the Switch 2, and how the company plans to navigate Trump’s tariffs and avoid repeating its history of poor sales following a previously successful console.