Which semiconductor maker has gone from being a “chip wreck” around a year ago, dragging down the whole sector, to a sharemarket darling with skyrocketing market capitalisation in 2023?
That’s Nvidia of course. Last week, the Californian company sparked investor frenzy as it reported way better results thananyone expected.
The rally added almost half a billion dollars to shares, which includes makers of the components Nvidia uses when it assembles graphics cards and other electronic devices around its chips.
As a result, Nvidia now dwarfs Intel, once known as Chipzilla due to its market dominance. Nvidia is headed towards becoming a trillion-dollar company like Apple, Microsoft, Meta, Alphabet and Amazon. Intel’s market capitalisation has dropped from US$282 billion in 2020 to US$120b (NZ$198.1b) now.
The giddy stratosphere rise in Nvidia’s share price is due to its early work in artificial intelligence paying off handsomely. We might not quite know where AI will take us, but an investor curse on any technology company’s house if they don’t announce generative models of some kind.
From a geek journo’s perspective, Nvidia’s been great to follow over the years. The company was set up by three engineers at a Denny’s in California in the 90s, the lore goes, because they thought graphical computing was the future.
Wafting away the swirling mists of time reveals an interesting ride, at times bumpy, over the years for Nvidia.
Their first large-scale commercial product was the Riva TNT graphics card, which piqued market interest as it could accelerate both 2D and 3D imagery. It wasn’t that long ago that 3D graphics was an aftermarket add on that cost heaps, and it set the stage for the GeForce range of cards that have been very successful for Nvidia. (Please tell me I’m not the only one left who remembers what “graphics cards” are?)
Nvidia focused on making graphics processors with lots of small “cores” that could handle large amounts of instructions in parallel. That’s much more difficult to make a good fist of than it appears, and general purpose processors went in a different direction trying to do more per ever-increasing clock cycle.
Long story short, we nowadays have both kinds of processors, but the highly parallel chips are very well suited for AI, and super speed guessing of passwords - cracking - and of course, hyper realistic gaming.
Nvidia’s cards were also very popular with crypto currency miners, who built “rigs” with lots of them installed. A few years ago, I was admiring a neatly designed crypto miner in a plexiglass case, with eight Nvidia GeForce cards installed and very beefy power supplies to feed them.
High end graphics cards were almost impossible to source at the time, as crypto miners bought as many as they could.
Soon after, the giant crypto currency pyramid scam imploded, and with it, Nvidia’s share price as graphics card demand fell. Gamers rejoiced however.
In March this year, Nvidia officially denounced crypto as adding nothing useful, but does that mean the environment can take a breather as there won’t be millions of graphics cards consuming vast amounts of electricity?
No, unfortunately. AI requires massive amounts of computing power. ChatGPT apparently was trained on a computer with more than 10,000 Nvidia cards. There’s going to be more of the same as every person and their virtual dog jumps aboard the AI train.
The level of innovation at Nvidia, as it goes after existing and evolving sectors, is nothing short of impressive. Nvidia’s Geforce NOW cloud gaming is developing and there’s already a provider in the region, Pentanet in Perth that’s about 8 millisecond pings away on an Orcon fibre connection.
Speaking of games, Nvidia has released its Avatar Cloud Engine for Games. This could be the killer application for AI, as it promises to add pseudo-intelligence for non-playable characters in games among other things, and I’m sure nothing could possibly go wrong with that tech.
Nvidia’s trying its hands on mobile chips and is getting ready to butt heads with Apple over which tech giant will control cars’ infotainment systems. Amazingly enough, infotainment systems are a crucial area of vehicles that car makers never realised they needed to own, and now it’s too late.
There are the Grace and Grace Hopper “superchips” as well, Nvidia’s foray into the data centre, piggybacking on AI. They have general-purpose processors in them, designed by British semiconductor architecture firm Arm, which Nvidia wasn’t allowed to buy for $66b as the deal gave regulatory watchdogs and national security agencies around the world acid reflux.
If it doesn’t become sentient and tries to kill us all as some depressed researchers and futurists predict, what will chip makers pivot to when AI fizzles out?
It could be quantum computing and communications, provided the necessary milestones are reached and the devices can run at room temperature.
Actually, it’ll probably be quantum AI which Google’s already working on. If you’re a tech journalist, start the countdown for the next AI email inbox flood now. Yes, Nvidia has a finger in the quantum pie as well.