Ahead of Jackson Hole — and, perhaps just as importantly when it comes to the near-term direction for equity prices, Nvidia earnings — markets lacked conviction.
To let some folks tell it, the fate of the world is inextricably bound up with Nvidia’s prospects. It’s hard to disagree, and I’d note that some observers take the whole “fate of the world” characterization quite literally.
The A.I. hype might’ve abated over the past month or two, but industry experts and some computer scientists still insist we’re witnessing the dawn of a new technological epoch. A Bible-themed instruction manual for removing peanut butter sandwiches from VCRs is just the beginning, they’ll tell you.
In the narrower context of equity markets, the risk around Nvidia’s report (and guidance) is clear enough: The stock is a linchpin of this year’s rally both on its own and through a kind of external halo effect on other big-tech names.
The bar for Nvidia is now very high indeed, and any disappointment would surely ripple across markets.
There’s a lot more at stake than this year’s A.I. hype cycle. Some are betting on A.I. to drive a new productivity boom in the US, and that’s a factor in the r-star debate which’ll take center stage in Wyoming this week. Beyond economic concerns, many worry the future of our species is on the line — that if we fail to regulate A.I. the way we failed to get ahead of the social media revolution, it could be an existential mistake.
The notion that the weighty questions implicit in all of this will be answered by a chipmaker’s quarterly results is obviously laughable, but it’s not far-fetched at all to suggest that a big miss (or another big beat) could mean the difference in new record highs for US shares in the very near future and a continuation of this month’s equity swoon with a kicker from September’s sour seasonal.
Options imply a 10% move around Nvidia’s earnings. We saw on Monday how meaningful that kind of move from Nvidia can be for the broader market.
Calls on the shares are more expensive than puts, and even if the upsurge in A.I.-related spending proves to be a “one-off,” one-off in this context would still mean several quarters. That’s not a prediction, it’s just to say that Nvidia’s success last quarter likely sowed the seeds for more demand given the hype the company managed to generate. (Suddenly, everyone not throwing money and resources at generative A.I. felt compelled to catch up.)
Meanwhile, in rates, US two-year yields are back near cycle highs. Traders spent most of this month obsessing over the long-end, but it’s worth noting that, as BMO’s Ben Jeffery and Ian Lyngen observed, two-year yields were “back to a level only seen during three other sessions this cycle.” Two of those three sessions were on the eve of March’s mini-banking crisis, echoes of which were heard in S&P’s sector downgrade on Tuesday.
Remember: Shocks arrive at the front-end of the curve under the normal mode of functioning. The post-GFC era (the QE mode) was anomalous in that regard — with the front-end anchored at the lower-bound, shocks could only arrive at the back-end.
“Bull steepening or bear flattening is a renewal of a more Fed-driven reaction function and emblematic of what we anticipate will be the theme as supply considerations and r-star debates give way to a renewed focus on the data and the FOMC’s reaction function,” Jeffery and Lyngen went on, adding that the last time two-year yields were at current levels, 10-year yields were 30bps lower.
They sounded a somewhat ominous tone, although ominous is a relative term given the pair’s assiduously diplomatic, infallibly measured cadence. “We’ll classify it as an environment where there is an increasing asymmetry for something to go wrong either in terms of domestic risk asset valuations, the pace of hiring within next week’s NFP release or an erosion of the situation in China,” they wrote.
I don’t understand why Nvidia is the indicator of anything. There’s no question companies are investing into AI independent of how Nvidia performs
Well, I mean they have a de facto monopoly on the chip side. I suggest listening to three or four podcasts on why Nvidia is so critical here. All A.I. investment roads lead back to their chips pretty much.
For now. But many or most real world users, the ones who will pay for the servers and nerds to run them, will not require Large Language Models (LLMs). Nvidia has a window of two quarters to reap some big bucks, IF “they” can fill the orders from FOMO CTOs.
“They” because they rely on getting a share of the capacity over at TSMC in Taiwan which produces all of their high-end chips. Is TSMC going to cancel orders from AAPL, AMD and QCOM to fill more orders from NVDA? Really?
The longer orders go unfilled, the larger the opportunity for companies to examine if they really have any need for an LLM to support spending money on AI to reduce staff and save money. They might decide that tools that allow a more focused at a company’s own databases may be just fine. Tools that use “Edge Chips” instead.
There is no other reason to spend money on AI. It’s no different than the cost-benefit analysis when a company looks at introducing robots to the factory or warehouse floor.
But for the moment, that’s not an issue in the market. “It’s never wise to stand in front of a stampede.”
D. Thanks for the keen insight. Fast growth can be a killer burden. Control of the chain is critical. Prices do tend to rise when demand outstrips supplier capacity.
NVDA’s hype cycle is in full swing with Saudia Arabia and the UK both publicly committing to buying H100s. As someone in the software industry I can tell you that the expense of training your own model should be based on a unique data set and purpose, for most intents and use cases Inference from an existing model is pretty darned good… (and as they leapfrog how much do you want to invest in something obsolete 3 months later?). Buying for Inference is also silly because chatgpt demand fell way off.
There are breakthroughs: listen to InvestLikeTheBest podcast with Des Traynor (a founder of Intercom) and hear how customer support will never be the same.
Will NVDA go up, I don’t know but I wish I bought a year ago when this should have been obvious 😉
Thanks.
I wonder how many call center functions require a LLM. As opposed to meteorology, pharma research and such where looking for corellations in a massive database makes sense.
When it comes to customer service, why are the shares of NICE, AI and PATH trading so heavily? You’d think that they should have a head start on how to deploy and use machine learning (AI) in a customer service environment.
But what do I know? Only that I have no interest in trying to short the “it” stock of the moment.
Notably, NVDA was founded in 1993 at the same time that Microsoft, Apple, Google, Amazon, and Intel were taking the markets by storm. What impresses me is their ability to continually innovate their chip products. Their capacity to do so really distinguishes them against their peers.
Totally agree that innovating in hardware is both harder and has lower margins so hats off to Nvidia… I wonder when they’ll “pull an Apple” and expand into software/services?
Gen AI is cool, but its just a really good probability machine. It helps me with my productivity, but its not going to eliminate as many jobs as the talking heads claim.
Its going to allow the robots to generate even more content, making the search for actual knowledge/truth even more difficult.
Kind of like how Uber used to be more inexpensive than a taxi (netflix cheaper than cable) — and now look where we are.
I hope I’m wrong, would love some UBI, but it ain’t going to happen.