Are US tech shares a bubble in 2023?

If you can answer that question correctly, congratulations: You have the keys to a fortune and, more importantly, you’re apparently possessed of deep insight into the viability (or not) of generative A.I. as a world-changing technology.

There are a number of salient questions facing investors this year, and if you had to make a list that didn’t involve speculating on whether the US and China will ultimately end up at war, the top two spots would almost invariably be occupied by questions related to r-star and A.I.

Admittedly, “Are large language models a precursor to human obsolescence?” is a lot more exciting than “Is the neutral rate higher?” But don’t forget: The two queries do overlap.

If you ask Goldman, the handful of “early A.I. winners” in the US likely don’t constitute a bubble. In an expansive new piece that runs three-dozen pages, the bank’s Peter Oppenheimer ran through the numbers (he ran through a lot more than that, but the full note, by virtue of scope, isn’t amenable to any sort of summary treatment).

Today’s dominant companies “remain at much lower levels than the biggest companies in other periods,” Oppenheimer remarked, noting that the so-called “Magnificent Seven” sport an average P/E of 25x, with an EV/Sales ratio of 4x. At the height of the dot-com bubble, those figures for the biggest companies were 52x and 8x.

For the Nifty 50 bubble, the P/E was nearly 10 turns higher.

Oppenheimer also endeavored to pull out the implied future growth that’s currently priced into the S&P 500. That’s a critical consideration given that runaway growth expectations were at the heart of the dot-com bubble.

For that exercise, Goldman used a fairly straightforward methodology, shown below. It’s the difference between the ERP and the risk-free rate (on one side), and the dividend yield and 10-year BKEs (on the other). It assumes a 4% ERP.

As the figure shows, current long-term growth expectations, despite having increased, are “much less than the expected annual growth implied in the late 1990s,” as Oppenheimer put it.

He also noted that mega-tech in the US has 4% cash as a share of market cap, double the share seen during the tech bubble, and boasts ROE and margins that are twice as robust.

After presenting a few additional statistics and metrics, Oppenheimer made a simple but crucial point: Profitless tech was largely decimated last year as rates rose. What we’ve seen in 2023 are rising valuations for profitable companies.

The figure above shows a veritable collapse in the relative performance of unprofitable versus profitable tech.

“This is an important difference,” Goldman dryly remarked. “The biggest profitable tech companies have benefited both from less competition and from their strong balance sheets and cash flow generation, which has made them increasingly defensive on a relative basis under the weight of higher interest rates.”

Here’s the bottom line, from Oppenheimer again:

Given the valuations of the dominant incumbent companies are high but not excessive, we believe we are still generally in the first phase of a typical technology wave. If this is the case, it suggests there will be further emergence of new entrants in the space and still higher valuations in this part of the market. There is a risk that the current enthusiasm leads to a bubble, or to a point where incumbent valuations rise excessively relative to their future growth potential, but we do not think we are at this point yet.