4On March 17, the Wall Street Journal published an article:Trump enthusiastically said on social media this week that Nvidia would build an "AI supercomputer" in the United States. Why does Trump want “AI supercomputers” to be built in the United States so much? What exactly is it?
‘Supercomputers’ can perform calculations and simulations at speeds and scale far beyond those of everyday computers. They are designed for tasks that require processing large amounts of data simultaneously, such as weather forecasts or simulating what happens inside an atom. Traditionally, supercomputers use large amounts ofCPU. In a supercomputer, there may be tens of thousands of CPUs working in parallel through a high-speed network.
The AI supercomputer envisioned by NVIDIA is a little different. This is a machine that houses hundreds or thousands ofGPUA large computer that focuses on training artificial intelligence models—such as those used by OpenAI’s ChatGPT.GPUs are particularly useful for these training tasks.
In recent years,Nvidia's high-end GPUs play key role in advancing artificial intelligence, high-performance computers built with them are often called AI servers. These servers are centralized in large data centers.
Nvidia didn't precisely define when AI servers will transform into AI supercomputers, but it may want to emphasize that,New machines using a large number of its latest Blackwell chips can bring about qualitative changes with high-performance quantitative changes.
Nvidia said it has commissioned more than 1 million square feet of manufacturing space in Arizona to make and test its Blackwell chips, and AI servers in Texas. The company is working with Foxconn, the world's largest electronics contract manufacturer, to build a factory in Houston, and with another manufacturer, Wistron, to build a factory in Dallas.
High-tech products like the iPhone are difficult to produce in the United States, but AI servers are another matter. Smartphones are highly price sensitive, and consumers may refuse to buy them if the price increases even slightly. butBig U.S. companies buying servers are willing to pay more for hardware produced closer to home. They can afford the higher costs of producing in the United States.
In addition, making iPhones can require hundreds of thousands of manual laborers, so hourly wage levels matter. andAssembly of AI servers is highly automated, the key elements are engineering, design and software. The United States has comparative advantages in these areas.
The Stargate project touted by Trump has said it will spend $500 billion to build artificial intelligence. The initial phase of the data center is planned for Abilene, Texas. Apple has also said it plans to build a new factory in the state. Adriana Cruz, executive director of economic development for the Governor's Office, said,Texas “will become the epicenter of AI infrastructure.”
However, industry observers sayUntil these companies give concrete results, be wary of "dramatic" announcements.