Nvidia began its life as a maker of graphics cards, and it did so very successfully. But gaming is small potatoes compared to the vast sums of money being thrown around by mega corporations as they bet big on AI. Improved AI needs ever more powerful hardware, and Nvidia is taking advantage of that.
Tae Kim, a staffer from media outlet Barron's (via Tom's Hardware), cites consulting firm Raymond James in estimating Nvidia is making as much as 1,000% profit on each H100 compute GPU it sells. The estimate puts the base cost of a H100 GPU at $3,320, while the shipped product can sell for upwards of $30,000.
Of course, these estimates are just that: estimates. No one outside of Nvidia will know the true costs involved, not to mention the specifics of different SKUs, software, licencing and support, or even the prices on offer for those buying them. If I want to buy one H100 it's difficult, but if I want 10,000? My quote would be very much different.
The current wave of AI hype is being driven by the computational requirements of Large Language Models (LLMs). One physicist believes they are "glorified tape recorders", but whether it's automation, customer service or delivering better data insights, AI is here to stay, and Nvidia is in a position to dominate this lucrative market for years to come. Companies like Intel and AMD are scrambling to catch up.
Full credit to Nvidia, which deserves its success. Nvidia has been sowing the seeds of its tech for many years, and if predictions are true, the only way is up. It's a little analogous to the crypto currency GPU mining boom, but for companies with billions of dollars to spend instead of thousands or millions.
Nvidia's next-generation GPU is codenamed Blackwell. But while the consumer and enterprise versions might share DNA, they will be quite different. Each family could end up with a different name, as Hopper did compared to the consumer oriented Ada Lovelace. It's sure that whatever Nvidia has cooking in its labs, big corporations will be lining up with compute GPU orders.
Nvidia's success can be interpreted as being good for gamers in general. Nvidia has more resources than ever to put into its gaming GPUs. It can hire more highly paid hardware and software engineers (with a bit of help from AI). It isn't showing any sign of neglecting the gaming GPU market, but with profits on offer that are an order of magnitude above that of gaming graphics cards, it'll be understandable if the professional market is where its attention is increasingly drawn.
Nvidia's stock price dropped under the $5 mark in mid-2015 before reaching $480 in July of 2023. Not a bad increase. Not bad at all.