Why are CPUs and GPUs so hard to make? Dr. Ian Cutress explains

Dr. Ian Cutress, also known as TechTechpotato on YouTube and the industry analyst behind Moore, is a frequent and welcome guest on PCworld videos and complete Nerd. And with him in the studio of San Francisco, it would be a crime not to choose his brain on the subject of the manufacturing of the processor. So that’s exactly what will do.

First of all: what does it take to make a silicon chip in a factory? Or to be more precise, a manufacturing plant or a “fab” to be short. I heard it once described as “a bag of VA in a machine on one side, the machine hits the bag and you get a processor on the other.” The reality is a little more complex. We are talking about thousands of individual stages with hundreds of different components and processes, some requiring precious fear and materials. And it’s just the physical things that enter a chip. In fact, putting the operational factory is the work of planning and development years, and billions – sometimes hundreds of billions – dollars from start to finish.

And of course, things become more and more complex as processors become more powerful, using smaller components and processes on the nanometric scale. Although a fab can continue to make flea for years or decades even after it is no longer at the forefront, the new things are so difficult that it is consolidated as giant competitors. Currently, it’s Taiwan Semiconductor Manufacturing Company (TSMC), Intel and Samsung. Between them, they make chips for almost everyone, from AMD and Nvidia to Apple, Qualcomm, Mediatek, and Cetera.

But there is a new “targeted effort” to create a new manufacturing plant for 2 nanometers from a company called Rapidus, which is spearhead for a new thrust for the production of fleas in Japan. And like the other major players (remaining), Rapidus would court the production of massive customers like those listed above, as well as partner companies in the country, notably Sony, Toyota and SoftBank, among others with the blessing and tax support of the Japanese government. He should be ready to start making bleeding chips absolutely by 2027.

So all these things are quite high-end corporate stuff, and not much to run on the consumer side. But what about fleas that enter graphics cards? And if Nvidia, and to a lesser extent AMD, can earn all the money in the world for the sale of fleas to the AI industry … Why would they even want to sell fleas to PC players?

As Ian explains, a large part of the same technology and the same development that arise at this MoneyPinner AI also enter standard graphics cards in consumers, to say nothing other applications such as game consoles. To say nothing about the extreme specialization – you cannot just get an AI chip with low reinforcement and throw it into a graphics card of $ 2,000 instead. It also helps to diversify and remain relevant for consumers … because investors are also consumers. Even a billionaire sometimes buys a laptop.

Thank you very much to Dr Cutress for having broken these complex subjects for us. To find out more about the latest news and trends in the technology industry, subscribe to PCworld on Youtube and watch our weekly the full Nerd podcast.