Late June.
That’s when the next selloff is coming. Or at least according to BofA’s Michael Hartnett who, in the latest installment of his popular weekly “Flow Show” series, suggested real yields aren’t high enough to undercut the burgeoning A.I. bubble.
“4% real yields popped the internet bubble, 3% popped subprime, crypto crashed on real-yield rip from -100bps to 150bps, but the market is telling you real rates may need to rise another 100-150bps from here to pop the ‘baby bubble’ in A.I.,” he said, in one of three hypothetical trader quotes describing the late-May “zeitgeist.”
The figure above is instructive to the extent it suggests meandering reals in 2023 left the door open for equities to adjust and recover thanks to a handful of mega-cap tech stocks sensitive to rates. I’ll leave aside the real-yield curve discussion, which is important for the dollar.
It’s a little early to be thinking about what might pop the A.I. “bubble.” It might not be a bubble, for one thing. Nvidia’s forward multiple was 61x on Wednesday afternoon. It’s 46x now. That’s still expensive, and the drop is only as legitimate as analyst forecasts are accurate, but it shows you how quickly things can change when “‘hope” becomes reality overnight.
If A.I. is a bubble, it’s inflating at a time when the Nasdaq 100 is still well off its all-time highs. Indeed, an equal-weighted index of the so-called “Magnificent 7” isn’t even back to its post-pandemic extremes, although I should note that Hartnett’s point with the chart shown below is obviously to suggest something ominous about 2023’s narrow rally.
My guess would be that another FCI tightening episode — tied to, say, a hawkish June FOMC meeting along with a run of cold (or hot!) US data — could undermine tech’s near-term momentum, particularly in the event any risk-off episode is accompanied by a stronger dollar and higher reals. But I doubt we’re anywhere near “peak A.I.”
The dot-com analogy may ultimately be a good one, though. I could certainly imagine a scenario where early A.I. optimism propels tech-heavy US benchmarks to new records against a perilous macro backdrop of still-high inflation which ultimately forces the Fed to adopt a bubble-buster mentality, culminating in a painful shakeout that incinerates sundry “pretenders,” clearing the way for the strongest names to rise from the ashes on the way to world domination.
If that’s the right analog, I’d suggest the frenzied crescendo is still a ways off. From a sequencing perspective, the way to get there would be a near-term pullback for risk in light of inflation realities and begrudging acquiescence to the Fed’s “higher for longer narrative,” then rekindled risk appetite around some manner of soft-ish landing where ~3.5% CPI is implicitly assumed to be the “new normal” and nominal growth stays perky alongside a resilient labor market. An extended Fed pause on the notion that eventually, “long and variable lags” will produce 2% inflation, could be fertile ground for speculation in risk, particularly if the first signs of labor market weakness are met with “insurance cuts.”
Assuming the hype around A.I. only grows, and no serious recession intervenes to curb “arms race” spending, it’s pretty easy to posit a new dot-com moment. But, again, that moment isn’t near at hand. If the A.I. bubble ends this summer, it’ll be the most disappointing bubble I can personally remember, especially given the world-changing potential of the technology.
Of course, that doesn’t mean Hartnett is wrong. Indeed, I think he’s probably right: The stars can align for another near-term swoon in risk. “June 14 FOMC Fed hike expectations are up from nothing to 12bps despite debt ceiling kabuki, the Nasdaq is taking 10-year Treasury yields back to 4%, the dollar is surging in May [and] financial conditions are tightening again,” he wrote. “Once spreads in creaking global high yield bonds are back above >550bps, we expect another bout of risk-off to return in late June.”
That’s entirely plausible. It’s just a matter of what comes next. Is it a proper recession that resets the economy and snuffs out inflation while the Fed stays resolute at terminal to ensure the job is well and truly done? Or is it merely a moderation in the macro that gives the Fed plausible deniability to indulge a year and a half’s worth of pent-up dovish inclinations? If it’s the latter, and the Fed prematurely hands investors the “dovish pause” markets have been waiting on since late last year (say, at the first sign of trouble in the labor market), policymakers would be sowing the seeds for the next mania.
For now, Hartnett wrote, “no conviction in the macro narrative of 2023 means big conviction in the new micro narrative of A.I.”
This is contingent on adults prevailing on the debt ceiling issue which doesn’t seem to be the case. To me it seems the only way a deal gets done is if McCarthy does a deal with the moderates knowing he will be removed from the speakership after the deal is done.
Based on the Congressional Clown Car’s success at forcing McCarthy to make concessions in order to become the speaker, his removal by the CCC for disappointing them is conventional wisdom. Yes, they can call a vote on his removal. But do they have the votes necessary to remove him? I don’t think they do and the failure to remove him would be a spectacular and well deserved ‘shut up and sit down’ moment.
IMO the so-called “dot.com” bubble came about because of a feeding frenzy by VCs who were suffering a a huge bought of FOMO. The trouble was that what these folks missing out on was a bunch of ideas that had no ability to make a profit. Too many dot.com companies had no viable business plans. As soon as many of these firms started trying to sell their wares at a level high enough price to make a profit, their customers disappeared and the firms went poof. With IA, there are many notable ideas out there, although to me the jury is still out on much of this tech. Many current searches produce ridiculous results not even close to what one is seeking. Like all new fields, most of the starters will fold. Some will figure it out. The biggest difference between the internet flops of yore and the coming AI business is that the latter will bring many more ethical, legal, and social issues to the table. And the secrets of the black boxes will create havoc.
Great points, Lucky One. I’ve yet to see much commentary about what firms will benefit from using AI. Everyone is obliged to say they are looking at it to please Wall Street, but we need to see some actual examples of end-users saving big money by using AI.
As our DL and many of you have noted, those savings will primarily come from reduced staffing needs. Most commentary glosses over that and only focusses on how you can generate cool graphics. But given the reluctance of many firms to spend money on tech beyond what is absolutely needed (an example is tepid spending on data security products) CFOs will need to see some strong examples of cost savings before spending large sums on AI and enduring the inevitable disruption as it is rolled out firm-wide.
Derek
Nice comments. One thing that occurred to me from your last (correct) sentence, is that money is not the only outcome of the corporate use of AI. The other is power. AI surely offers the possibility of creating unique and powerful capabilities that offer the prospect of massive moats for companies that deploy it properly. Not every AI will do. Only hippopotamus AI will do. But man, it will really do.
The other thing that occurred to me earlier today is that the impact of AI will expose the ultimate value of a college education (not a degree, per se). I read the following on a post today: “Every time I lost a job, it was an opportunity for a better job, even in my 50s. The trick is to always be learning new things.” Exactly! That is what education is for, to learn the process of how to teach oneself how to evolve. My daughter is now fifty. In the five years after high school she earned bachelors degrees in four fields, none of which she ever worked in. She evolved from being a medical assistant, to a CTO in a large bloodbank, to a software developer/project manager for a military supplier and several large tech firms, to being an ISO expert, quality manager and a senior director running a private cloud, data warehouse and applications developer. The well educated will easily survive AI. Hell, they’ll be creating it. The educational scoffers will end up at QuickTrip.
Thanks Lucky One.
I’ve been pondering all day about this. If I was at a fat check useless consulting firm I’d tell my clients to relax and see how some real like implementations go. Why not? Do any of you recall how it was common to blame earnings misses on the unexpected costs of switching to SAP ERP software?
And you know what? It’s probably right to just chill a few months to see what happens. It may turn out that the slow adapters maintain better client satisfaction until the AI algos catch up.
Why should I bother to be the “first mover” pioneer? How much competitive advantage falls to the fist users of a new technology?
What’s the hurry???
Okay, I’m going to go out on a limb and suggest that I am probably the only one here who was actually writing about AI in the early to mid 80s. For the likes of Computer Design magazine (long since defunct). Back then, it was purely a technology story, but the hype was the same: It’s about to explode—it’s just around the corner.
Since my interest at the time was software engineering, I did go so far as to learn Prolog, the only real AI language back then (Python, sometimes mentioned in AI circles, hadn’t been invented yet) and I got into it enough to have a sense of how language processing was going to evolve. It looked like at least 20 years of development to me before AI could really interact like a real person, and though I was interested in pursuing my ideas in that direction, the career change from journalism to software engineering required that I earn a living, and AI simply wasn’t near enough to the software front lines for that to be viable. So even for me, AI pretty much disappeared from sight until recently.
Forty years later, what has changed? Well, the way I see it, the main change is that AI is now being discussed on Bloomberg and presumably in the daily ink press; it’s still about to explode and just around the corner.
What’s evolved over the past forty years is a better AI interface: it can now converse in an excellent imitation of a speaking human as opposed to parsing rather carefully worded written input and responding in kind. And that’s a big deal, believe me! It takes a lot to do that. But no one reading these “pages” needs to worry about losing a job to AI. Not in your lifetime. Human intelligence is pretty amazing, and computers just aren’t there yet, at least not from anything I have seen. Once this current AI flareup dies down, I expect it will be another 10 years before AI does anything truly amazing. I could be wrong, but I’m not holding my breath.
At some point, we’ll need to see real life examples of massive staff “rightsizing” (in homage to Al Dunlop) to justify end users spending money on this. No doubt it will, in some cases, but when?