I have always felt that the emphasis on having bleeding edge chips is something that is generally overstated. I don’t have deep knowledge of the wide-ranging applications so I can’t say authoritatively. But you don’t always need 0.00001 nm chips to be able to compute. There is an overemphasis on raw compute power as opposed to what the chips are supposed to be computing, whether it is necessary or even desirable, and whether software can be optimised to do more with less.
Right, the main advantage of smaller chips is that they consume less energy. This makes them really useful for stuff like mobile devices because you can extend battery life while improving performance. However, for vast majority of applications such as chips in cars and appliances, it really doesn’t matter because these chips aren’t doing computationally expensive tasks to begin with. There is far more demand for these types of chips then there is for bleeding edge ones for stuff like phones, tablets, and laptops.
its overstated because its kinda what is more interesting so people click it for a peek. Strictly in thr US for example, not many talk about Global Foundaries, or Texas Instruments because thats what they kinda do now, fab micro controllers and chips on older process nodes.
TI for example has a shit ton of fabs, but the chips they make go into your everyday kinda mundane stuff. No ones interested for your performamce of your washing machine for example.
Yes, this is why the “stop China’s military devwlopment” excuse is bullshit. Missiles and drones don’t use NVIDIA H100s, they use dedicated chips made on older nodes for durability.
Almost everything uses legacy nodes. Specially military applications, they prefer mature and proven fabrication processes.
