As Chip Prices Rise, Investors Focus On How AI Stack Really Makes Money

Good news. The 'throw money at everything remotely related to AI' era appears to be winding down. The absurd data center build out costs, insupportable power and water demands, questions about actual increases in productivity as well as real sales and profits on top of all the related concerns are finally registering with investors. 

And what is following is not a repudiation of AI, but a much more interesting and nuanced assessment of how money can and will be made from this technology (Gosh, just a like for a real-live business!). Pardon the sarcasm, but what the rise in chip prices - and their manufacturers' stocks - is encouraging is a series of detailed analyses of where, in the chain of production, sales and marketing, profits will most likely be made. And there is one word for this evolutionary development: healthy. JL

James Mackintosh reports in the Wall Street Journal:

The explosive growth in Micron’s profit in the latest quarter is good news for shareholders, but it comes at the expense of the AI companies to which it sells fast-memory chips. Micron and SK Hynix are both up 290%, while Microsoft and Meta are down, Amazon’s flat and Alphabet’s only up 8%. Since April, when the memory stocks accelerated up, all the hyperscalers are down, along with Nvidia. This could be traders shifting money out of the last big wave of AI gains to chase chip stocks. (But) there are only three ways to deal with higher chip prices. Make lower profits (short-term), find efficiencies so you need less (in the long run), or wait for more supply (as fat profits encourage production). All three are likely in AI, and investors need to think carefully about which parts of the AI stack will make money, and how long it will last.

The explosive growth in Micron Technology’s profit in the latest quarter is extraordinarily good news for its shareholders, but it comes at the expense of the artificial-intelligence companies to which it sells fast-memory chips.

Micron, along with Korea’s Samsung Electronics 005930 -4.76%decrease; down pointing triangle and SK Hynix, are to AI what oil producers are to the airlines: makers of an essential input that this year suddenly became much more expensive. 

Because there is extremely limited capacity to make the high-bandwidth memory that AI needs—and it takes years to build new production facilities—soaring data-center demand has simply jacked up prices. Micron’s soaring profits are, for its customers, soaring costs.

We are witnessing an enormous transfer of cash from the providers of AI—and, perhaps one day, AI users—to the memory-chip makers. Profit shifts of this scale are rare events, and investors should be paying attention to where the money’s coming from, where it’s being spent and how long it will keep flowing.

In the quarter ending May 28, Micron increased prices for DRAM chips more than 60% on the previous three months, while increasing shipments by a low single-digit percentage, it said this week. Prices for NAND flash memory, also used in data centers, jumped more than 80%.

Usually, memory doesn’t matter that much. But for Micron alone, customers paid $18 billion more. And that was just in the quarter. Prices have quadrupled in a year, and it’s hurting outside AI, too.

Apple this week raised its prices for MacBooks more than 15%. Closer to home for me, the memory I bought on Amazon.com a year ago to build a super-quiet computer (I hate fan noise) has tripled in price, and now costs more than the CPU. For an industry in which prices usually drop every year, it’s a huge turnaround.

In consumer electronics, passing on higher prices helps limit demand for chips—just as higher oil prices reduce consumption. But the AI companies aren’t passing on higher prices, because they are able to throw money at supply problems.

The problem in AI is that the end users aren’t covering the costs of the service, with big losses recorded by the leading AI-model producers, OpenAI and Anthropic. Everything is still priced to bring in new customers, not yet to make money. So higher input costs create a nasty problem: Either losses will be even bigger, or higher prices will be needed, putting off potential customers.

Already big companies have begun rationing AI as employee usage of the “tokens” that measure chatbot and agent use skyrockets, without necessarily helping productivity. Pushing up the price of the tokens in this environment will only slow adoption.

Given that investors are, for now, focused on market share rather than the bottom line, I wouldn’t expect the higher costs to be passed on to AI users. 

That means they will be absorbed either by the data-center providers or the model producers. Total profits for the AI “stack” of businesses critical to supplying large language models remain about the same, but margins for buyers of chips are hurt and more of the profit is made by the chip suppliers.

This might already be showing up in the stock market, although the speculation in microchip shares is so extreme it’s hard to be sure.

This year, shares in Micron and SK Hynix are both up about 290% in dollar terms, and Samsung Electronics is up 166%, while the so-called hyperscaler cloud providers of Microsoft and Meta are down, Amazon’s flat and Alphabet’s only up 8%. Where it really shows up is in the performance since April, when the memory stocks accelerated upward. All the hyperscalers are down since then, along with Nvidia, which packages memory chips with its graphics processors for AI.

This could just be traders shifting money out of the last big wave of AI gains to chase soaring chip stocks. But it fits the idea of profits shifting from high up in the AI stack down to the memory-chip suppliers.

Even some bullish investors worry about the scale of speculation in chip stocks. Alec Cutler, a fund manager at Orbis, argues that the increased complexity of manufacturing and reduced number of competitors means there is no need for concern about new supply—my main worry, and the usual cause of collapsing chip prices and profits.

However, he’s sold his longstanding holding in Samsung Electronics in favor of its preferred shares, which normally track closely but have lagged behind the ordinary stock during the boom this year, and particularly since April.

Ultimately, there are only three ways to deal with higher prices from chip suppliers, as the airlines can attest when oil rises. Make lower profits (the short-term response), find efficiencies so you need less (in the long run), or wait for more supply (as fat profits encourage production). 

All three are likely in AI, and investors need to think carefully about which parts of the AI stack will make money, and how long it will last.