Fri, Mar 21, 2025 page9

• Nvidia planning massive outlay on US electronics

  • Bloomberg
  • • 
    • 
    • 

  Nvidia Corp aims to spend several hundred billion dollars to procure US-made chips and electronics over the next four years, the Financial Times reported.

  Nvidia chief executive officer Jensen Huang (黃仁勳) told the Financial Times that the latest chips designed by his company, and Nvidia-powered servers for data centers, can now be produced at US-based factories operated by Taiwan Semiconductor Manufacturing Co (TSMC, 台積電) and Foxconn Technology Group (富士康科技集團), known in Taiwan as Hon Hai Precision Industry Co (鴻海精密).

  It marked a major step forward in supply chain resilience for the US chipmaker, Huang said.

  

  Nvidia Corp chief executive officer Jensen Huang speaks at a news conference at the Nvidia GPU Technology Conference in San Jose, California, on Wednesday.

  Photo: Bloomberg

  Nvidia this week hosted its GTC developer conference, where Huang, 62, said his company ultimately plans to shift manufacturing onshore. It is already using TSMC’s Arizona factory to help produce some of its highly prized graphics processing units, which have become the most essential component for the current wave of artificial intelligence (AI) investment.

  In its annual report last year, Nvidia said that it had US$20 billion in total purchase commitments as of Jan. 28 last year.

  US President Donald Trump’s proposed tariffs on imports would have a small impact in the short term, Huang said.

  They will not take a major toll on Nvidia because it would continue to shift manufacturing for the most important parts of its product line onshore. The company is already using TSMC’s facility in Arizona and is to increase that as its supplier builds more capacity, he said.

  TSMC chief executive officer C.C. Wei (魏哲家) on March 3 announced an additional US$100 billion in investment in the US, alongside Trump in the White House, and Foxconn is working with Apple Inc to add an AI server assembly site in Texas.

  Separately, Huang on Wednesday said that his company has not been approached about purchasing a stake in Intel Corp.

  At a news conference at the developer conference in San Jose, California, Huang was asked whether Nvidia was part of a consortium to buy Intel.

  “Nobody’s invited us to a consortium,” Huang said. “Nobody invited me. Maybe other people are involved, but I don’t know. There might be a party. I wasn’t invited.”

  Huang at a meeting with analysts and investors on Wednesday said new types of AI models that produce more complex answers would only increase the need for computing infrastructure, adding that concerns caused by Chinese start-up DeepSeek’s (深度求索) R1 AI model — that fewer chips and powerful servers would be needed for such software — were misplaced.

  “The understanding of R1 was completely wrong,” Huang said. “Computation demand is much higher.”

  Huang was also asked about the efforts of customers to develop their own components — chips that might replace Nvidia’s AI accelerators in data centers. Google has been working with Broadcom Inc to develop their own application specific integrated circuits (ASICs) for this area, but Huang countered that many ASICs are designed, but not always actually put into data centers.

  Those customers need better chips to generate more revenue from their infrastructure, not cheaper ones to save costs, he said.

  Additional reporting by Reuters
• Most Popular

  • 1

    Missile mass production to commence in Q2: official

  • 2

    Flight disruption was deliberate ploy, experts say

  • 3

    Air quality to deteriorate nationwide from tomorrow

  • 4

    Taiwan closely monitoring Venezuela developments

  • 5

    Donovan’s Deep Dives: Taiwan’s basic services could begin to crumble this year