(Bloomberg) — With much of the world’s attention on the AI race between the US and China, this week’s Computex conference served as a stark reminder of the central role that Taiwan continues to play in the global technology industry.
Most Read from Bloomberg
The 2025 edition of Asia’s signature tech gathering didn’t break as much new ground as last year, when Nvidia Corp.’s Jensen Huang unveiled a multi-year roadmap for AI chip design. But it still drew the attention of an industry parsing the fallout from Washington’s campaign to curb Beijing’s tech ambitions in the aftermath of DeepSeek.
Here are the major takeaways from Computex 2025:
Consumer AI Needs More Time
There was scant mention of consumers this year. In 2024, Qualcomm Inc. devoted its presentation to how AI would make everything better and we’d never again need our laptop chargers on the move. This time around, it flipped to a script focused on enterprise applications for AI.
Foxconn, also known as Hon Hai Precision Industry Co., used its first keynote address at the show to go big on heavy-duty AI, from making cities smarter to bringing robots to manufacturing and making EV design more efficient. The company that assembles the bulk of the world’s iPhones had nothing to excite consumers.
Asustek Computer Inc. offered a sobering outlook on the prospects of so-called AI PCs.
Samson Hu, co-chief executive officer of the Taiwanese computer gear maker, said AI PCs will take a year or two before they go mainstream. That’s because the software is still immature, plus new tariffs from the US are likely to flatten near-term growth. Asus may have to hike prices in the US by as much as 10% to deal with those levies, Hu said.
DeepSeek Changed AI
The debut of DeepSeek in January changed the nature of AI, highlighting China’s advances in the field while dismantling basic assumptions about computing needs.
Huang was among the few executives who addressed that first openly, saying that China is “incredible” at software development and has the resources to make up for any technological gap.
DeepSeek also popularized reasoning models, which require more raw computing power. “And so now the reasoning model is not one shot, but it’s hundreds of shots,” Huang told reporters this week. “DeepSeek increased the amount of computing need by maybe 100 to 1,000 times.”
China Looms Large
The most-clicked headlines centered around Nvidia’s delicate position as the US pursues chip restrictions intended to curtail China’s tech ascent. That came to a head Wednesday, when Huang branded that effort a “failure” and called for Washington to back away.
Story Continues