Mission AImpossible

23m

S&P 500 futures and Nasdaq 100 futures are trading to the upside on Monday after the continued bleeding of AI names handed the latter tech-heavy index back-to-back weekly losses for the first time since July.

The big news of the morning is Meta lining up a $95 billion equity issuance to fund its AI ambitions, with the three biggest US AI accelerator companies — Nvidia, Broadcom , and AMD — said to be among the buyers. Shares of the social media giant are getting a lift, probably because they were already sold hard in the aftermath of earnings when this offering was teased. And hyperscalers’ credits are catching a bid thanks to Zuckerberg following through in the stock market, rather than with more debt.

While earnings season never ends, it does peak — and that’s what we’re looking at this week. Over one third of the S&P 500 by market cap releases results, with most of the high-profile names coming after the close on Wednesday and Thursday.

Solid results from companies tethered to the AI boom have not always been greeted with a warm reception this earnings season. Case in point: Bloom Energy, which tumbled double digits on Friday despite delivering across the board beats and boosting its sales guidance for 2026 for the third time to about $4.5 billion (consensus: $4.1 billion).

Retail said the drubbing was a buying opportunity, with Robinhood traders making their second-largest day of net purchases on record for Bloom, trailing only September 28.

The iShares MSCI Brazil fundas well as the two most heavily-traded Brazilian ADRs on Robinhood, Banco Bradesco and Petrobras, are getting shellacked in premarket trading after the shock election results saw President Lula emerge victorious over Flavio Bolsonaro (the son of a former president).

If there’s a silver lining for retail traders, it’s that the most aggressive phase of accumulation in these stocks happened in Q3 and Q1 of this year, respectively, with activity much more muted since then.

As a subsidiary of Robinhood, Sherwood Media is restricted from writing about any company in which Robinhood is or was a selling group member of the IPO during the regulatory "quiet period" for that company.

Mission AImpossible

Quarterly updates from storage stocks, AI infrastructure companies, three of the Magnificent 7’s four hyperscalers, and Intel: this will surely be a week for trying to answer impossible questions about the longevity of the AI boom.

The near-term intensity of this buildout has rarely been questioned (and persistently underestimated), but investors appear to have adopted an “if something can’t go on forever, it won’t” approach to the situation.

“The debates around AI: I don’t think we will definitively answer any of the larger debates (open/closed, revenue required for capex, ROI) within the next 2–3 months – it would be impossible given how large and long-term the development and proliferation of AI will be,” writes 22V Research data infrastructure and commodities analyst Dauvin Peterson. “Therefore, as we have seen before, we reach cyclical peaks and troughs on these views, and I’d say it’s leaning pretty far to one side right now.”

So this week is really about whether our feelings change in response to this deluge of results and outlooks — if what we hear is good enough to assuage concerns and prompt an AI relief rally, or if a broader move below July lows for more of the space prompts soul-searching about whether the commercial applications of this quest for Digital God will bear enough fruit.

TS Lombard strategist Dario Perkins set the stage for this busy week by paying homage to Citirini Research’s James van Geelen with a “scenario analysis” of the 2028 bursting of an AI bubble.

Per Perkins:

“Although there had been bouts of market volatility before, most notably in the autumn of 2026, nothing looked broken. In fact, the various near misses had only convinced the bulls that their optimism about AI fundamentals were justified. They had certainly made a lot of money over the years, and everyone knows that money is irrefutable proof of intellectual superiority (no “intelligence crisis” here). Crucially, the hyperscalers had continued to spend like drunken sailors, and that continued to flood the AI ecosystem with huge amounts of cash.”

This hits on one critical point about the US stock market as a whole (and the AI boom) that needs to be repeated ad nauseam: its strength is due in no small measure to being a momentum trade!

In 2026, every fresh rise to all-time highs for the S&P 500 has occurred amid a brisk rally or rejuvenation of the long-short momentum factor and a deterioration in value and low volatility factors. Right now, momentum has given back all of its YTD total returns since early January, and value and low vol are in ascendance.

The path of toppling dominos that Perkins lays out will look familiar to everyone — it’s basically a supercharged endgame version of what we seem to price in and worry about once a quarter!

He offers an eminently reasonable takeaway to cap an evocative report:

“George Soros’s famous advice was that when you see a bubble forming, you rush to buy. That works if you’re Soros and you sell on time. Our advice is more modest: hold a normal amount of risk but protect yourself from the downside. In practice, that means a diversified portfolio of global stocks, with less of the AI complex than the index hands you. This is not like Dotcom, which spread across the developed world, Europe included. This time the bet is concentrated in US tech and an Asian hardware supply chain, so most of the rest of the world should fare better - eventually. And add duration as the vulnerabilities appear, because bonds are what rallied.”

On the other side of the spectrum, Ben Bajarin, CEO and principal analyst at Creative Strategies, went micro to macro in a detailed note pushing back against recent claims of an emerging multi-trillion dollar gap between the outlook for AI revenues versus spending.

The unit economics for providing inference already work, per Bajarin, but the problem is that not enough of these racks are being explicitly used to make money.

“The hardware earns its keep when it is busy and paid for, and the shortfall comes from the share of the fleet that is not paid for, at least not in a way we can count,” he writes.

He argues that industry-wide financials should continue to improve going forward with revenues covering costs by the early 2030s. This breakeven date could be pulled forward to the start of the new decade if one assumes a reasonable attribution of how hyperscalers’ compute bolsters their core businesses, according to his analysis.

“The realistic outcome is a buildout that earns its cost of capital in the early 2030s: a sound infrastructure investment, but well short of what the most bullish narratives imply,” Bajarin concludes.

The tension between capital constraints emerging now and returns being a half decade away is a great summary for how the perception of, and drivers behind, the AI trade have changed materially in the second half of this year.

The highest-flying names in the theme lost their momentum as semiconductors’ earnings revisions stopped accelerating. Perhaps more importantly, megacap tech’s quarterly reports in July were what prompted every analyst to throw in the towel on the idea that the hyperscalers, in aggregate, were poised to reclaim their status as free cash flow generators next year.

Guppies and GPUs

Data center stocks CoreWeave, Cipher Digital, IREN, Nebius, Applied Digital, TeraWulf, and more got smacked (again) this morning after the world’s richest man pronounced a harsh judgement:

One person’s “helpful market color” is another’s “kicking an enemy while they’re down.”

As my colleague Walt Hickey wrote in the Snacks newsletter after Elon replied “yikes” in response to a chart of CoreWeave CDS, Musk “simply owns a competitor,” that is, SpaceX. He is not a neutral observer!

For fans of Dodgeball, this is a tweet with strong, “Here at Globo Gym we’re better than you, and we know it!” energy.

Shares of all of these aforementioned neoclouds recovered off the lows on confirmation of Meta’s issuance plans.

treAIsury market

Before the term “dumb money” was ever bestowed upon retail traders specifically, it was implicitly applied to stock traders in general. The bond market is the “smart money,” it was known.

In that case, it must be very annoying for the smart money that the inmates are now in charge of the asylum. Across the yield curve, the US bond market is in thrall to tech stocks.

From JPMorgan’s head of global rates strategy Jay Barry, unpacking recent trends in the US Treasury market:

“We do not want to understate the impact of shifting narratives on domestic risk assets as well. Figure 4 shows that over the last month, Treasury yields have become increasingly correlated with the performance of the NASDAQ, while displaying a negative correlation with Brent prices. It is difficult to disentangle these two factors, as earlier in the Middle East conflict, Treasury yields began to display a negative sensitivity to oil prices when Brent broke above $110/bbl. Looking ahead, despite a resilient economy, equity markets are likely in the driver's seat for rates valuations.”

And here’s BMO head of US rates strategy Ian Lyngen on how stock market volatility has fueled some meaningful swings around the pricing of Federal Reserve action:

“The latest bout of FCI [ed’s note: financial conditions index] tightening has reinforced market expectations for steady policy rates next week while tempering the prospects for a rate hike at the December 9 meeting, although the market continues to favor such an outcome this morning. For context, the market-implied odds of a December rate hike stood at 87% heading into Thursday’s session, dropped as low as 65% intraday, and have since retraced to 83%.”