Nvidia Way

Kim, a Barron's writer and former semiconductor analyst, wrote a history of Nvidia based on interviews with Jensen Huang, the co-founders and many current and former employees.

The corporate history covers the 1993 founding, the near-failure with the NV1 which used an unconventional rendering approach the market rejected, the RIVA 128 that saved the company, the GeForce line and the invention of the graphics processing unit category, the acquisition of 3dfx's assets, the Xbox contract dispute with Microsoft, the failed ATI merger and AMD's acquisition of it instead, the years of losses in mobile with Tegra, and the CUDA decision.

CUDA is the strategic centre. Nvidia spent roughly a decade from 2006 investing heavily in a software platform that let general-purpose computation run on graphics processors, with no clear market and against consistent investor objection to the expense. Kim argues that this was a bet on a capability rather than on an application, that the scientific computing users who adopted it early supplied the ecosystem, and that when deep learning researchers found that neural network training mapped onto the same hardware, Nvidia held a software position competitors could not replicate quickly.

The management material covers Huang's practices: a very flat structure with dozens of direct reports, the requirement that information be shared with everyone rather than filtered up a hierarchy, written status updates sent directly to him, no formal one-to-one meetings, criticism delivered publicly, and the stated principle of intellectual honesty about mistakes.

The book is written with company cooperation during a period of extraordinary success, which shapes it. It does not seriously examine the competitive risks, the customer concentration, the concentration of manufacturing at TSMC, or the questions about whether current demand levels persist. Later events in this sector move faster than publication.