Nvidia envy
The previously low-profile chipmaker has soared to top dog status because AI loves and needs it
Few people not into computers, video games or cryptocurrencies know much about Nvidia, the company that last week sparked a miniboom in stock markets, with its own market capitalisation soaring to within a whisker of $1-trillion.
Nvidia is now by far the most valuable semiconductor company in the world, well ahead even of Taiwan Semiconductor Manufacturing Co (TSMC), which makes most of Apple’s chips, and worth eight times as much as a now bruised and bloodied Intel a company once referred to as Chipzilla because of its dominance, but which has since had a dramatic fall from grace.
Only four US companies are worth more than Nvidia (and they’re also all in tech): Apple, Microsoft, Alphabet (Google) and Amazon.
How did Jensen Huang, Nvidia’s co-founder and CEO, take a scrappy start-up whose focus was building silicon chips to speed up computer graphics and turn it into the new Chipzilla?
Nvidia was founded in 1993 by Huang, Curtis Priem and Chris Malachowsky on the premise that for computer game graphics to advance significantly, PCs needed a discrete and dedicated graphics processing unit (GPU) to take the workload away from the central processing unit (CPU), typically made by Intel or AMD.
For most of its life, Nvidia was a minnow next to Intel and many other chip companies. GPUs were used mainly by gamers, and most computers, especially those sold into the corporate market, continued to offer only relatively weak graphics integrated into the CPU.
But the dynamics of the market started to change when cryptocurrency miners realised a few years ago that Nvidia GPUs are well suited to solving the complex mathematical problems needed to mint new digital tokens. Miners especially those working to extract new ether tokens on the ethereum blockchain snapped up Nvidia GPUs by the boatload to power vast crypto farms, fuelling a shortage and sending prices rocketing.
Those days are over, thanks in large part to changes last year in how ether tokens are mined moving from a so-called “proof of work” to a “proof of stake” consensus mechanism that obviated the need for all those energy-hungry GPUs. Prices returned to some semblance of normality, though they have remained higher than they were historically, mainly because of their growing complexity Nvidia’s flagship GeForce RTX 4090 GPU has 76.3-billion transistors in a chip just 6cm² in size.
As prices fell, Nvidia’s shares which had surged to a record high in 2021 lost more than half their value. The collapse in crypto prices resulted in demand slumping further. All round, it was looking to be a tough 2023 for Nvidia.
Then it happened.
In late November 2022, OpenAI a firm backed by billions of dollars in investment from Microsoft released ChatGPT, its generative artificial intelligence (AI) tool that has captured the world’s imagination. Microsoft saw the opportunity quickly: by integrating generative AI into its search engine, Bing, it could sock it to rival Google, undermining its business model built on serving ads on web search results and not on the output of generative AI. Microsoft is now integrating the technology into everything from Teams to Outlook to Word, potentially unleashing a step change improvement in global productivity this decade.
Google, which has been
Financially, Nvidia is the big winner so far from the generative AI craze
working on AI tools for years, had to play catch-up (at least on commercialising the technology), and now is working furiously to integrate generative AI tools into its search results. Other tech giants, including Facebook parent Meta Platforms, are also investing big money to ensure they’re not left behind in what could become the next big growth area in tech (sorry, Zuck, it won’t be the metaverse).
All these companies need silicon to power their AI tools truckloads of it, to fill giant cloud data centres. And one company happened to have just released the best AI chip money could buy: Nvidia. Made under contract by TSMC, its new H100 chip can cost as much as $46,000, and companies such as Microsoft and Google are scooping them up in their tens of thousands.
Nvidia can’t keep up with demand and has forecast blowout financial results, sending its share price into the stratosphere. Financially, Nvidia is the big winner so far from the generative AI craze.
“We went from a pretty tough year last year to an overnight turnaround,” Huang told the Financial Times last week. The success of ChatGPT was an “aha moment”, he said. “It created instant demand
This whole thing took off just as we’re going into production on Hopper.”
Hopper is the proprietary chip architecture on which the H100 chips are built.
For now, Nvidia has a commanding technological lead. But others, including big players like AMD and a host of start-ups, are racing to market.
One company is notably absent: last Wednesday, when Nvidia’s share price popped, pulling the stock market up with it, Intel shares bucked the trend, falling 5%. It is Chipzilla no more.