What comes after the bubble could be electrifying

A version of this post first appeared on TKer.co

If AI is all it’s cracked up to be, stock market winners should they extend well beyond large-cap technology hyperscalers is currently building the AI infrastructure.

As BofA’s Savita Subramanian wrote June 2023: “The greatest profit can be made by old-economy and inefficient firms that can increase their earning power more permanently from efficiency and productivity gains.”

It is too early to say conclusively, but the market may be in the process of facing this phase of the AI narrative as small-cap stocks have recently outperformed large-cap stocks.

As Wells Fargo’s Ohsung Kwon argues, small-cap stocks (as tracked by the Russell 2000, or RTY) could see a bigger AI tailwind than large-cap stocks (as tracked by the S&P 500, or SPY).

“We see signs that small caps have been slower to adopt AI than large caps,” Kwon wrote on Monday. “We believe the next stage of AI adoption is in small caps – the long-term bull case for RTY. We estimate that every 1% labor cost savings This translates into ~2% EPS growth for SPX, but >6% for RTY.”

Small-cap stocks will benefit greatly from AI. (Source: Wells Fargo) · Yahoo Finance

This is what makes the promise of AI really exciting for investors: the beneficiaries are not limited to those who develop the technology. Companies in all industries have been exploring AI applications and many are already implementing them in their operations.

To be clear, we are still at first entries of the AI era, and it will be some time before we better understand the real impact on productivity across all sectors.

But if costs are materially reduced and productivity improved beyond technology firms, it would be consistent with past technological revolutions.

In a recent research noteBridgewater’s Greg Jensen drew parallels between AI today and the electrification of the 1920s and the Internet boom of the late 1990s.

“At least in the short term, AI seems likely to follow the classic J-curve productivity path of earlier general-purpose technologies like electricity or the Internet, which require a lot of upfront investment that doesn’t immediately improve productivity but is ultimately transformative,” he said. he wrote.

As you can see from the charts, it took years for the economy to realize the productivity booms promised by electricity and the Internet.

“We believe that AI investments will significantly support U.S. growth in the coming years, and that many of the second-order consequences of that investment are priceless,” he added.

Read Jensen’s full memo here.

Carlyle Group’s Jason Thomas looked closer in those historical experiences. And he wasn’t shy about tackling the risk head-on we may be in a bubble as we were in 1929 and 2000. In fact, the title of his note is: “Bubbles as a feature, not a bug.”

“Financial markets tend to be very good at detecting technological revolutions, but have rarely proven capable of anticipating their second- and third-order effects,” Thomas wrote. “Bubbles deflate not because the technology disappoints, but because much of its economy flows downstream to the companies that use it and the new industries it spawns.”

He discussed the decades-long process of harnessing and deploying electricity from the 1880s to the 1920s. “Earnings and valuations of electric holding companies and equipment manufacturers tripled; between 1925 and 1929, annualized returns for the entire industry exceeded 50%.”

These companies could not dodge 1929 accident. But what followed is remarkable.

“Electrified factories lowered manufacturers’ production costs, and electrified homes stimulated demand for their products, such as refrigerators, vacuum cleaners and radios,” Thomas observed. “This is where much more of the economic value of electrification ultimately accrued, with an annualized return of 11% for stocks of durable goods manufacturers over a decade stretching back to the depths of the Great Depression.”

Electrification helped reduce the cost of production of durable goods. Manufacturing stocks outperformed during this period. (Source: Carlyle) · Yahoo Finance

Thomas noted that after the dotcom bubble burst in 2000, companies that took advantage of the Internet performed similarly in the wake of the market downturn.

“These same phenomena were repeated in the Internet boom of the late 1990s,” he wrote. “Real capex grew at an annualized rate of 24% as investors focused on bandwidth, router architecture and processing power bottlenecks. But the economic value of this IT and telecommunications hardware manifested itself downstream. Networked computing allowed manufacturers and retailers to exploit new sales channels, more precisely calibrate production and inventories to sales and rationalize logistics networks and supply chains.

Companies that benefited from the Internet fared better in the wake of the dotcom bubble. (Source: Carlyle) · Yahoo Finance

“A technology doesn’t need to sell out because its bubble deflates,” Thomas continued. “Similarly, those inclined to focus on ‘irrationality’ or ‘market euphoria’ miss how bubbles appear to be a natural consequence of epochal changes. If the market did not misjudge the importance of hardware bottlenecks and first-mover advantages, who would be willing to fund the research and infrastructure that ultimately make revolution, technology and technological value possible?”

Read Thomas’ full note here.

As we did commented beforehistory is littered with examples of game-changing technologies leading to overinvestment followed by market corrections.

unfortunately, it is not easy to predict when these corrections will occur and what they will look like.

But the bottom line is that investors should be mindful of the risk that the hot stocks most directly exposed to the AI capex boom are at risk of being left behind.

Indeed, this maybe it’s already happening.

On Wednesday, Deutsche Bank’s Jim Reid shared this chart of where various AI-exposed stocks were trading relative to their all-time highs as the S&P 500 itself was climbing to new highs.