An employee works on a printed circuit board at a factory in Dongguan, Guangdong province, China.Tingshu Wang/Reuters
Everything related to artificial intelligence has turned to gold recently. Nothing, though, has shone quite so brightly as tiny slivers of heavily engineered silicon. Which is rather odd when you think about it.
The business of making microchips (or semiconductors, as they are more formally known) has historically been a boom-and-bust affair. For decades, the semiconductor sector has chewed up enormous amounts of capital with only intermittent success.
Investors, though, are suddenly seeing a lot to love in the chips biz. The PHLX Semiconductor Sector Index – better known by its ticker symbol, SOX – surged to a 47 per cent gain in April.
The spectacular run-up in chip stocks explains why Wall Street just enjoyed its best month since 2020 despite the continuing troubles in the Strait of Hormuz. What isn’t so clear is why the market has suddenly fallen so hard for chip makers.
Any explanation has to begin with investors’ growing belief that AI will transform the economy. A lot is riding on that belief: AI-linked companies, from chip makers to power providers to the “hyperscalers” that run big cloud-based data farms, now make up about 45 per cent of the U.S. stock market, according to The Wall Street Journal.
The stock market’s bet on AI is only going to get bigger if AI pioneers OpenAI and Anthropic follow through on their plans and go public later this year. Each of those chatbot giants is likely to be valued in the hundreds of billions of dollars by investors.
Oddly, though, it’s not at all clear whether a) AI will actually have a big impact on the economy and b) who will benefit if it does. The AI pioneers and the hyperscalers such as Alphabet, Inc., Amazon.com Inc. and Microsoft Corp. have already spent massively on their AI ventures. No one is sure, though, who is winning – or, for that matter, if anyone is.
Which brings us back to chips. One of the few certainties in the current tumult is that semiconductors are vital. Any company that wants to develop AI applications has to buy huge amounts of chips to do the “inference” and other work necessary for AI models to operate.
CIBC’s chief market technician Sid Mokhtari reveals his top 10 stock picks for May
This means that chip makers offer a dependable way to bet on the continued growth of AI. It doesn’t explain, though, what has abruptly turbocharged these stocks over the past month.
One possibility is that the boom results from technological shifts that are allowing one type of chip known as central processing units (CPUs), to replace another type of chip, known as graphics processing units (GPUs), in certain applications.
“The latest rally in the SOX seems to have been fuelled by signs of rising demand for CPUs as lower-cost alternatives to GPUs for AI inference,” writes John Higgins, chief economic adviser at forecaster Capital Markets. He argues this is why a long-time laggard, CPU-maker Intel Corp., has suddenly sprung to life and outperformed long-time chip leader, Nvidia Corp., which specializes in GPUs.
The run-up in chip stocks could also reflect worries about a continued blockade of the Strait of Hormuz. While microchips aren’t manufactured in the Persian Gulf region, they depend heavily on certain products, such as helium and naphtha, that come out of the Gulf region and play key roles in the intricate process of chip making. If the Strait were to remain blocked for months and the stockpiles of these vital ingredients were to run down, chip shortages could result and chip prices could soar.
So should investors bet on the chip boom continuing? The leaders in the field – companies such as Nvidia, Taiwan Semiconductor Manufacturing Co. and Broadcom Inc. – are anything but cheap. Their ebullient share prices reflect a belief that spending on chips will continue to grow as AI spreads throughout the economy.
That could happen. It’s worth noting, though, that the amounts of money being wagered on AI are now gigantic even by the standards of past investing booms. According to earnings reports this week, four hyperscalers – Alphabet, Amazon, Microsoft and Meta Platforms Inc. – are on track to spend a staggering US$725-billion this year on AI-related capital expenditures.
That is amazing. Equally amazing, none of these big spenders has figured out a way to make money from their AI ventures. It’s quite possible they never will.
Critics such as tech observer Ed Zitron argue that the only way to make AI profitable is for providers to charge users for the actual cost of the service – and that would be so prohibitively expensive that it would chase away the vast majority of potential customers.
Others are more optimistic and argue that costs will decline as scale grows and technology improves. Perhaps so. But it’s difficult to avoid the impression that the hyperscalers are pouring huge amounts of money into AI not because they have a clear business plan in mind, but because they feel they can’t afford to be left behind.
Even a modest slowdown in AI investment would cause pain in the semiconductor sector. Investors may want to keep that in mind before jumping into the chip frenzy.