In January 1848, gold was discovered in Sutter’s Mill, California. Shortly afterwards, 300,000 people arrived in the state during the famous ‘gold rush.’ A German migrant, called Levi Strauss, set up a store in San Francisco to sell them the clothes, picks and shovels they needed. Most of the miners never made any money. When he died, Strass was a millionaire.
Surely, Jen-Hsun "Jensen" Huang, NVIDIA’s CEO, must see Strauss as a kindred spirt? Huang’s ‘GPU’ chips are the essential tools that are driving the AI revolution. Whether Silicon Valley’s entrepreneurs find AI gold or not, they still need NVIDIA’s chips to dig for it.
An insight that drove NVIDIA’s market-cap to an eye-watering $3.52 trillion. As The Guardian says, ‘The simple reason for the explosion in valuation is that Nvidia is the company to bet on if you want to invest in AI.’ Even more remarkable – NVIDIA’s stellar performance owes as much to serendipity as strategy. The Guardian again, ‘Nvidia is an overnight success, decades in the making, thanks to the fortunate coincidence that another sector of computing also needs very, very simple calculations to be done very, very fast: gaming. The G in GPU initially stood for “graphics.”’
In short, chips developed to render Super Mario in glorious technicolour, are, it turns out, perfectly adapted to instantly perform the mind-boggling calculations needed to build LLMs (Large Language Models). Nothing number crunches like a GPU.
Now, The Telegraph reports, as ‘Large language models such as ChatGPT and Google Bard require thousands of GPUs, both for their initial “training” and for the subsequent interactions known as inferences. Almost every company and government is falling over themselves to invest in AI. Amazon, Microsoft and Google, which operate the giant cloud computing data centres on which AI models generally run, are placing orders worth billions of dollars with Nvidia.’
But Huang has been far more than just lucky. Like Strauss before him, he has perfectly understood the direction of the market. As Forbes points out, ‘There is no question that Huang is an exceptional chief executive and that Nvidia is an exceptional company, successfully moving rapidly to capitalize on the AI opportunity that suddenly presented itself over a decade ago.’
Recently, Mark Zuckerberg announced plans for Meta, to create the ‘world’s fastest AI supercomputer.’ To do that he has ordered 350,000 of NVIDIA’s fastest chips. Each one costs $30,000. Huang may not have dug for it, but he has struck gold.