C3.ai (NYSE.AI) and Nvidia (NASDAQ: NVDA) are the two major stocks that people are using to bet on the AI revolution. As Paul Graham points out, this might not be the best set of possibilities: “AI is the first big new wave of technology we’ve had since start-ups started going public much later. The result is a unique problem: public market investors who want to invest in AI have few options. Most of the good investments are still private. “But we’ve got to deal with what we’ve got available on the public markets since we’ve only got those public market based investments available to us.
C3.ai did enormously well before its recent results. Up by multiples of the original price, 300% at least over the year to date. That all went more than a little pie shaped when the results weren;t as good as hoped. There are at least some people muttering that C3 is chosen simply because the dumb money is seeing that the ticker is “AI” and investing on that basis.
There is a point to the Nvidia valuation
Nvidia is rather different. They are making the chips that people are using in every greater quantities. That’s what drove the market cap up over $1 trillion as they forecast ever greater sales in this coming accounting period. Which is fine.
But to work out what’s really going to happen here we need to dive deeper. We are indeed in the middle of a technological revolution and while that does mean a lot of licking fingers to see which way the wind is blowing some basics can be thought through.
The real profit makers from AI will be you and me, consumers
The major beneficiaries of AI are going to be consumers. There’s simply too much competition in the market for that not to be true. There are no network effects here – it’s not like Microsoft, everyone uses Windows because everyone uses Windows. Nor like social media, everyone’s on Facebook because that’s where everyone is. Any and every half decent AI will be competing with every other half decent AI. Monopoly profits just aren’t going to be there.
This does mean that we’re in something akin to a gold rush. Sure, some people strike gold and make good money. But the people who consistently make money are those selling shovels to miners. That’s exactly the reason that people are piling into, or on, C3.ai and Nvidia. They’re providing the tools to, respectively, manage or run an AI. C3, well, maybe. We’re among those who thing the AI ticker has more effect than might be wise here.
Nvidia though, in the short term clearly it will do well. But long term not so much. Think through how computing actually works. We have processors, great. Intel, ARM, whatever. We have graphics processors, Nvidia – great. But what’s the difference? Graphics requires mathematical performance – floating point operations. So does mining Bitcoin – that’s why a few years back everyone switched from using general purpose processors to graphics cards to mine crypto. Because the graphics chips are optimised, designed even, for the floating point operations which speeds up any mathematical operation.
AI also requires that maths. Graphics chips simply will run an AI better – less energy, less time to gain the same result – than general purpose chips. So, Nvidia has a great position here, thus the $1 trillion valuation.
Ever heard of an ASIC, the thing Nvidia doesn’t make?
But now think. What happened in crypto? People stopped using Nvidia and other graphics chips. Actually, it became impossible to make money mining using them. Why? Because there’s the next stage of silicon, to go from general processor to application specific, to an AISC. Here the code, the routine, is embedded in the silicon which is always vastly faster than running code over a chip. Crypto mining became a game of who had the newest and best ASIC. The money for the shovels went to those designing (not fabbing, that’s a commodity business) those ASICS. Nvidia and other graphics chips makers lost out at this point. That’s why graphics cards got a lot cheaper this past 18 months, ASICs had finally taken over the crypto market.
This is going to happen again. Absolutely inevitably. AI will get to the stage where it is not run across general purpose chips, nor in the cloud, nor even on graphics of floating point chips. The code will be embedded in silicon, there will the ChatGPT ASICs (and, of course, other variants for other engines, Bard and so on). This will make AI much cheaper in time and energy – but clearly the money is going to move from those making general to those making specific chips.
ASIC makers for AI are only just starting
It’s entirely possible to think that people like Brainchip (ASX: BRN) are a bit early and might not make it – most of the people who try this won’t make it of course. But that is the angle from which the competition is going to come. It’s also where the real winners of AI are going to be. In those who embed the routines and computing necessities down into the silicon.
Now, where we find those who will succeed and how we sort through the pretenders, well, that’s another matter. So at this point the only real guide we’ve got is that those benefiting from the demand for general purpose silicone, like Nvidia, are going to find themselves cut off at the knees. As with crypto. The only question is when? Trying to go short NVDA at this point would be mad. But it is absolutely true that there will be a change of sentiment there, it’s when, not whether.
