Industry Icon Jim Keller Proposes a $1 Trillion Alternative to Altman’s $7 Trillion AI Fab Network Plan

Sam Altman made headlines earlier this week when his ambitious call to raise a staggering $7 billion came to light. The tech maestro of Silicon Valley primarily aims to raise enough funds to establish a series of fabs that would produce abundant chips for AI applications around the world.

NVIDIA CEO Jensen Huang and Chip Architect Jim Keller Believe that Innovation in AI Solutions is the “Efficient” Way Forward

Raising trillions of dollars is no small deal, which is why this news is bound to raise many eyebrows. It’s a matter that was brought to NVIDIA’s CEO, Jensen Huang, for his thoughts. He seemed to believe that what should be done is the opposite of what Sam Altman is trying to achieve. Jensen believed that the focus should be on ‘improving‘ the GPU architecture, rather than attempting to build a new semiconductor supply chain.

You can’t assume just that you would just buy more computers, you also have to assume that the computers are going to become faster, and therefore, the total amount you need will not be as much.

Otherwise, the mathematics, if you just assume that computers never get any faster you might come to the conclusion we need 14 different planets and three different galaxies and four more suns to fuel all this. But obviously, computer architecture continues to advance.”

Jensen Huang on Sam Altman’s $7T AI supply chain proposal

Industry icon and chip designer Jim Keller seems to have the same vision as NVIDIA’s Huang, where he asserts that the way forward is mainly through architectural innovation and not focusing on quantity. Interestingly enough, in reply to Sam Altman’s tweet—”f* it why not 8,” commenting on raising the amount to $8T, Jim replied: “I can do it for less than $1 trillion,“

This statement in itself doesn’t refer to fulfilling the roadmap Altman has proposed but rather to fulfilling the vision that he believes is necessary for the next generation of AI applications. Tenstorrent, where Jim currently works, has a very sophisticated roadmap of its own, where they believe that fulfilling the industry’s demands is largely possible through improving the technology at hand.

While it is true that ChatGPT’s AI models require a lot of computer power than people may realise, the problem is that money doesn’t always guarantee what we want. For instance, China, despite pouring in almost $150B, in 2017 into achieving self-reliance till 2025, still heavily imports semiconductors. This shows that while funding is important, it only solves one part of the puzzle.

This is all we know for now, but rest assured that we will keep you updated as new information becomes available.