BREAKINGVIEWS-Markets have poor scorecard for spotting AI losers

By Edward Chancellor

LONDON, April 24 (Reuters Breakingviews) - When a groundbreaking new technology arrives, it’s often easier to spot the losers than the winners, says the writer and investor Alisdair Nairn. Over recent months, stocks in sectors from software to information services deemed to be facing a threat from machine intelligence have taken a hit. Yet the history of past technology booms shows the stock market has a spotty record of identifying losers. The best rule to follow at such times is biblical. As the Gospel of Matthew (20:16) states: “So the last shall be first, and the first last: for many be called, but few chosen.”

When the first British steam railway, between Stockton and Darlington, opened in 1825, it was cheaper and faster at moving goods than canals. Yet investors in British canal stocks remained complacent to this threat for the best part of a decade. Fifty years later, the dominant U.S. telegraph company, Western Union, showed a similar disregard to the threat to its business posed by Alexander Bell’s new telephones. Bell received his patents in 1875 and offered to sell them to Western Union for a pittance the following year. The company turned him down. Western Union outperformed the U.S. stock market for a decade to the mid-1890s, despite the rapid advance of telephony.

The market’s ability to identify both winners and losers during the bubble in technology, media and telecom (TMT) stocks in the late 1990s was abysmal. Companies in sectors that were not directly exposed to the internet were dismissed as “old economy” and their valuations marked down. At the same time, the valuations of so-called “new economy” stocks soared to the moon. On 29 February 2000, Jim Cramer declared that the internet was about to change economics of every industry: “It eliminates any bricks-and-mortar company that doesn’t embrace the Net…,” the hedge fund manager and CNBC pundit wrote on TheStreet.com. “It just destroys retail as we know it… How can Bank of America compete with Nokia as a way to bank?”

Cramer recommended that investors avoid banks, brokers, commodity makers, newspapers, machinery stocks – leaving them with just tech stocks. This pronouncement came just a few days before Nasdaq peaked. After that most of the sectors that Cramer had written off, notably miners and banks, went on a bull run. From 2000 to 2003, the Dow Jones Industrial Average, which was heavily exposed to “old economy” stocks beat the tech-heavy S&P 500 benchmark, while the Nasdaq Index dropped nearly 80% from peak to trough.

When the dust settled, formerly high-flying telecom operators that had invested hundreds of billions of dollars building out the information superhighway were spurned as providers of “dumb pipes”, obliged by the rules of net neutrality to charge a flat rate to the users of their networks. Likewise, European mobile phone operators failed to earn their cost of capital for building out their third-generation networks. “Content is king” was the clarion call of the boom period. Yet the newspaper industry was gradually destroyed by the internet, as advertising revenues migrated to the new tech giants, including Google and Facebook. The market was slow to observe the threat. The stock price of the New York Times reached an all-time high in the summer of 2002, after which it fell 85% over the following years.

During the so-called Everything Bubble of 2020 and 2021, the market once again confused the winners and losers. At the time, anything related to electric vehicles and clean energy soared, while an anti-bubble formed in fossil fuel companies and component suppliers to the makers of supposedly soon-to-be-redundant internal combustion engines. At one point, electric carmaker Tesla’s TSLA.O market capitalisation exceeded the entire US energy sector. Miners of platinum group metals, used in the catalytic converters for non-electric vehicles, sold off. After 2022, the green tech bubble burst and energy stocks rebounded. Valterra Platinum stock is up around 130% over the past year.

When it comes to identifying losers from the artificial intelligence revolution, the market appears to be similarly error-prone. In late February a paper by Citrini Research went viral. The research firm suggested that businesses from food delivery firm DoorDash DASH.O to payment giant Mastercard MA.N would soon be undermined by new AI-powered competition. Since the start of this year, stocks of software, data and analytics companies and various online platforms have sold off.

Jonathan Tepper, chief investment officer of Bahamas-based Prevatt Capital, is sceptical. In a recent letter to investors, Tepper suggests that while some information and analytics firms may be disrupted, most have proprietary data and operate in highly regulated sectors. Online platforms for selling travel, cars, and real estate enjoy strong network effects, Tepper says, pointing out that Google has sought to displace them for years without success. Prevatt’s portfolio is picking up supposed AI “losers”, including London Stock Exchange Group LSEG.L and Veeva Systems VEEV.N, which provides cloud software, data and business consulting to the heavily regulated life-sciences industry. Tepper says ChatGPT owner OpenAI has given up seeking to book travel directly for customers, reducing the threat to another of his stocks, online travel portal Booking Holdings.

The key issue investors seem to be overlooking is the inherent unreliability of the large language models. These models generate responses through probabilistic calculation rather than deterministic reasoning. As a result they are prone to hallucinations. Alasdair Nairn points to a performance study of 166 LLMs from last October which found that the best-performing model had a hallucination rate of 0.6% while the twentieth ranked model had an error rate of 1.9%. Compare that to Six Sigma, the approach to corporate management which aims to reduce defects to under 4 per million incidences.

Given the current state of technology, mission-critical activities cannot be run by AI. Amazon’s AMZN.O cloud service suffered outages late last year after employees used an AI tool for coding. This should have been a wake-up call. Picking AI winners during the ongoing investment frenzy is a mug’s game. Looking for value among the supposed losers offers more assured gains. As the veteran investor Marc Faber commented during the TMT boom: “The greater the mania in one sector of a market, or in one stock market, the more likely that neglected asset classes elsewhere offer huge appreciation potential. This is one of the cardinal rules of investing, and will always work for the patient long-term investor."

Follow @Breakingviews on X.