The Thinking Machine

If AI is powering cloud growth today, then Nvidia chips are powering AI. The Thinking Machine is a page-turner of a book focused on it founder, Jensen Huang, and the story of how he built the company. The story is one of hard work, brilliant scientists and a degree of luck and timing. Nvidia started out as a firm manufacturing graphics acceleration cards, for serious PC gamers. The business became a leader in this field. The needs of gamers led to the development of parallel processing chips or GPU’s. Once these because available scientists found they could use the cards to solve complex simulation problems. Around 2003 Nvidia developed CUDA, its software to enable these users to run multiple Nvidia cards together in a parallel processing machine. Software increasingly because central to its business as newer domain specific applications were developed.

The next inflection point for Nvidia occurred after the demonstration of AlexNet at University of Toronto. This system, based on Nvidia cards, demonstrated the best image recognition capability of an AI/ML system. A new market for Nvidia chips was discovered. The rest of this story is about Nvidia’s efforts to improve performance and software for AI applications. To their credit, so far, they have stayed on top of the market.

What does the future hold? One vulnerability seems to be that Nvidia does not make its own chips. They use fabs at TSMC, in Taiwan.  Others can do the same thing, or even a political crisis could interrupt the flow of chips. Another issue is the enormous power requirements for AI training and inference. Is there enough electrical power to enable future AI application growth, with the same architecture we use now.? Also, the business has been run by one man, Jensen Huang. Will an effective handoff be made to address future tech and business challenges?

All in all, a highly recommended book!