Why Advanced Micro Devices Just Might Be The Next Big Thing In AI

Investors have been buzzing about artificial intelligence (AI) all year, and for the most part, that’s meant swarming around Microsoft, the software titan that’s bringing OpenAI’s ChatGPT into its suite of products, and Nvidia, the semiconductor behemoth that’s been designing the stuff that’ll power this new tech. But lately, another name’s been popping up that you might want to pay attention to: Advanced Micro Devices (AMD).

Why’s everyone talking about AMD now?

Immediately after AMD’s first-quarter earnings release, rumors began flying that the chipmaker’s been working with Microsoft on its secretive Athena project (speculated to be all about designing chips). And the timing’s interesting, to say the least. AMD’s CEO Lisa Su did her damnedest to distract investors from AMD’s fourth-straight quarter of plunging profit by liberally throwing around the words “artificial” and “intelligence” on the post-earnings conference call. It didn’t seem to work – AMD’s stock price sank 9% the next day – but then the rumor got going, and AMD’s stock reversed all those losses.

Now, AMD having AI ambitions shouldn’t come as a bolt out of the blue. The firm does compete with Nvidia in the artificial intelligence powering graphics processing unit (GPU) market, and any tech company worth its salt is trying to plunge its fingers in the AI pie somehow. What’s more, AMD announced it’s hosting an event on June 13th where it’ll “showcase its next-generation AI technology”. So whether a Microsoft partnership turns out to be fact or fiction, you can expect increased interest in AMD shares, especially in the run-up to June’s event.

What’s AMD all about then?

AMD has a market capitalization of $164.3 billion. Valuation appears expensive, with a forward P/E of 30.7x and EV/Sales of 6.6x. Sources: Finimize Markets and Google.

AMD was formed in 1969 by a group of employees who broke away from Fairchild Semiconductor, the OG chipmaker and Silicon Valley’s inaugural firm. Only a year earlier, a few other Fairchild engineers had done the same and incorporated Intel. In those first years, Intel and AMD didn’t compete: Intel was creating a computer-powering central processing unit (CPU), and AMD was focusing on memory chips. But by the mid-1970s, AMD had managed to effectively clone an Intel microprocessor, and so began a decades-long war between the firms.

Today, AMD is a diversified designer of semiconductors that power a whole host of gadgets, PCs, laptops, servers, and cars. This chart shows how AMD segmented the $23.6 billion revenue it generated in 2022.

AMD’s $23.6 billion revenue breakdown, by division, in 2022. Source: AMD company filings.

AMD’s AI-powering chips fall within the data center segment, but that division also houses regular CPUs too. In short, then, AMD’s AI exposure is just a fraction of that data center segment, probably lower than 10% of total firm sales.

Now, that’s not nothing, and having some exposure to AI is why the market is getting hot and bothered about AMD. But compare this to Nvidia, whose data center chips make up around 40% of its total sales. Most of those chips are consumed by supercomputers powering AI. (I wrote about Nvidia’s AI opportunity back in February).

Can AMD elbow in on Nvidia’s AI dominance?

No one really knows what snazzy Nvidia-killing products – if any – AMD will unveil. And, to be honest, for regular investors without a PhD in electrical engineering, it’s pointless trying to pit one semiconductor against another. The only way to know which chips will be winners is to see what customers wind up buying, and we won’t know that for some time to come. But I think there are other clues as to whether AMD could be successful in competing with Nvidia, and one of them is hidden in AMD’s success in closing the gap between it and another semiconductor leader, Intel.

In this game, if you’re not first, you’re last.

Semiconductors are a winner-takes-all industry. See, new consumer electronics like TVs or phones are constantly being released. And in the chipmaking world, you have to be ready with the fastest and most powerful semiconductors to make sure your products are designed into those shiny new gadgets. That’s not easy and it requires flawless execution. Any production delays, and your customers could miss important selling seasons, like Christmas, annoying them and sending them to your competitors. For decades, Intel was the king of execution, and its market share (blue line) punched above 80%, while poor old AMD’s (black line) dipped below 20%. But that’s been changing.

Central processing unit market share for Intel (blue) and AMD (black). Source: Statista.

See, AMD ditched its manufacturing efforts years ago to focus entirely on designing semiconductors: something a lot of other chipmakers were doing too. Most of today’s semiconductor companies don’t actually make semiconductors: they contract the manufacturing to firms like TSMC.

Working with the Taiwan-based powerhouse TSMC, AMD has been able to focus entirely on designing chips, leaving the challenge of actually making them to TSMC. And together they’re putting out ultra-tiny, seven-nanometer (a nanometer is a millionth of a millimeter) chips, while Intel, after a few stumbles, is still making ten-nanometer ones. That microscopic advantage is monstrous in the semiconductor world, and it’s meant that AMD has been able to steal market share from Intel.

Now here’s the thing: there’s a very big difference between the AMD-Intel battle and AMD’s possible tussle with Nvidia: manufacturing. See, Intel was able to build a lead in those early years by being able to make chips faster than AMD. And a manufacturing advantage can deliver a pretty sharp edge, with better sales meaning more money to build the most advanced factories and so on. And in reality, although AMD was smart to focus all its efforts on designing the best chips, it wasn’t until TSMC caught Intel in the manufacturing race that AMD was able to start retaking share. Against Nvidia, on the other hand, AMD won’t have that advantage: both firms contract with TSMC.

So the competition is likely to boil down to design. Now, no one’s saying creating super-complex chips that power the most complicated technology ever imagined is easy, and there are bound to be some competitive advantages that Nvidia – with its head start – has already racked up. But in theory, design gaps are easier to close than manufacturing ones, and so AMD might have better odds in its battle with Nvidia than it had against Intel. That’s exactly why the market pricked up its ears when the rumor sparked, and AMD announced its June AI event.

So, should AMD hold a place in your portfolio?

It’s probably not a good idea to rush into buying AMD on the hopes of AI success alone, but it’s an interesting time to be considering semiconductor companies anyway. And that’s because of a quirk in semiconductor investing. Chipmaking’s a cyclical – and therefore volatile – industry, and profits tend to swing in sync with economic cycles. Periods of booming sales are followed by sharp declines, which makes timing your investments important. Of course, with a long enough time horizon, you can ride out the ups and downs – something I’d personally favor with any stock. But when investing in the semi sector, it’s best to buy when profit has been on a downtrend, with the expectation that it’ll eventually move higher. And right now, AMD’s in the midst of one of those downturns.

AMD quarterly EPS in cents. Source: Company filings.

Of course, there’s a risk that things deteriorate from here – a nasty recession would put a chunky dent in semiconductor demand.

So it’s smart to look at valuations. Now, valuing semiconductor firms is extremely tricky because of those wild swings in sales, profit, and cash. For pure cyclical stocks – those that have little in the way of self-serving growth and that rely only on economic cycles – the best option is to compare the current price to historical average profit performance. The idea here is that a stock might look expensive using ratios like price-to-earnings (P/E) because current earnings are depressed, and so a comparison to a time when earnings were less depressed (assuming that’s what the future will bring) is a more accurate picture.

And through that lens, AMD’s expensive. The most profit the firm’s ever produced was in 2021 ($2.61 per share) which compared to the current $97 share price, puts AMD’s stock on 37x P/E, around a 50% premium to the S&P 500. But here’s the thing: AMD’s recent success against Intel, and its AI potential has got the market thinking it could be a supercharged growth firm in the future, and Wall Street analysts are pegging 42% profit growth next year, and 23% for 2025. That’d put the stock at 19.4x P/E using 2025’s profit estimates, a more appealing valuation. What’s more, if AMD is successful in its AI efforts, it’s highly likely that analysts today are actually underestimating future profit.

But that’s a big “if” and in my opinion, valuation’s the biggest risk. Plenty of other risks exist too though: Intel might get its act together and make back some ground on TSMC, and therefore, AMD. Plus, there’s that nasty recession risk in the near term.

Let’s face it, in the world of AI investing, nothing comes cheap. So if you’re convinced that the theme is going to be huge in the not-too-distant future, and you’re looking to diversify your exposure, then AMD might be an opportunity.

In the meantime, try to remember that when you’re dealing in uncertainties, which is what investing is all about, it’s important to draw historical parallels to guide your thinking. I do that all the time. This analysis has shown that AMD has closed the gap on a market leader before, so there’s probably a decent chance it could do it again.