SpaceX buying is out of this stratosphere

44m

S&P 500 futures and Nasdaq 100 futures are building on this week’s gains ahead of a mammoth day for corporate earnings.

Overnight, SK Hynix helped reassure investors on the longevity of the AI boom by reiterating that memory chip shortages won’t be alleviated until at least the start of the next decade. Sales and operating profits came in well above estimates, sending ADRs of the South Korean company and shares of peers higher in premarket trading.

This good news for memory stocks belies a passing of the torch by Robinhood’s retail traders:

SpaceX has recently supplanted Micron as the most-active stock among Robinhood traders. In the process, a mammoth gap has opened between net purchases of Elon Musk’s rocket AI company and everything else:

Net purchases of SpaceX by Robinhood retail traders over the past month exceed the accumulation of the next two most-bought names — Micron and Nvidia — combined!

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.

Geminis are born in October, actually

Two days, three hyperscaler earnings.

Google and Microsoft report after the close today, with Amazon’s quarterly results slated for release in the postmarket on Thursday.

Among retail traders, there’s been a clear favorite among these names over the past month.

Google has enjoyed significant net buying amid the unveiling of Gemini 4 Argon, while there’s been little accumulation of Microsoft (which recently announced buyout offers for about 60,000 employees!) and slight net selling of Amazon.

“How fast is your cloud business growing and how much do you plan on spending for that cloud business to grow more?” will be the common questions facing the trio.

Here’s a smattering of that and everything else analysts are focusing on as they appraise the results and futures of these tech heavyweights:

Amazon

What do agents do to e-commerce, and how will you respond?

BofA’s Justin Post:

“...addressing Cloud capacity ROI and Agentic risks to retail ad revenues could be key for stock performance post print.”

“We expect positive commentary on AI demand (expanding beyond AI Labs to core enterprise), custom silicon adoption, and Bedrock adoption.”

Deutsche Bank’s Benjamin Black:

“Prime Day’s shift into June from July creates a comparison headwind for both e-commerce and advertising. The subsequent improvement in checks since earnings is encouraging, although we remain mindful of the event’s demand pull-forward impact.”

“Looking further ahead, Amazon’s decision to block Meta’s Muse highlights an emerging debate over ownership of the customer relationship and advertising economics. Agent-led shopping could reduce consumers’ exposure to Sponsored Products, while restricting third-party access could redirect demand toward participating competitors. We expect Amazon to eventually develop its own agentic shopping offering, with its proprietary purchase data and commerceinfrastructure positioning it well.”

Google

Do TPUs/AI growth more than offset deterioration in the core Search business?

Bloomberg Intelligence’s Mandeep Singh:

“Alphabet’s external TPU sales stand to support modest upside to 3Q Cloud revenue estimates and more meaningful 2027 revisions, though Search growth below the midteens could offset part of that benefit Meta Muse’s Sept. 8 launch is unlikely to have materially affected 3Q Search, but broader agent adoption can divert high-value travel and e-commerce queries.”

BofA’s Justin Post:

“From a results standpoint, search growth (given soft 3P estimates), relative Cloud acceleration (MSFT reporting same day), depreciation expense growth, and margins are key focus areas. Seven other focus areas for the earnings call commentary includes: 1) Search volumes given renewed agentic risk, 2) a potential Google consumer agentic launch, 3) comments on Gemini 4 capabilities, 4) capex and capacity ROIs, 5) 4Q expenses given $3bn of est. 4Q SpaceX capacity costs, 6) any impact from token price declines, 7) any data on TPU sales, inventory, and outlook given TPU sales are key upside case for 2027. We do not expect 2027 capex guidance on this call, though we believe Street could be ~$300bn+.”

Microsoft

Quite simply, are your AI apps good enough to support continued enterprise growth?

Goldman Sachs’ Gabriela Borges:

“Our two biggest takeaways from our fireside chat with CFO Amy Hood on 9/9 and IR meetings 9/15-16 were: a) Microsoft appears to have control over more variables in their capex equation today (which in turn drives more flexibility), in part because short-dated capex and CPUs are a bigger piece of spend than 6 months ago; and we think Microsoft may choose to provide more datapoints on ROIC soon; and b) the Copilot apps strategy is working.”

“We believe Microsoft has more capex variables in its control today vs. a year ago; this in turn suggests that upward revisions to FY27 Street capex may be limited. This is a good thing, in our view, as upward revisions to capex without commensurate revisions to Azure revenue have weighed upon the stock in prior quarters.”

Deutsche Bank’s Brad Zelnick:

“Fundamentally, we see evidence of improving AI adoption among Microsoft's 1P Apps user base and higher monetization rates across its stack more broadly providing underlying support for shares to continue moving higher into calendar year end, while tactically for F1Q we think investors will be looking for a couple points of topline outperformance, guidance supportive of further acceleration through the year in buyside numbers, and reiteration of its commitment to remain FCF positive in FY27.”

Yorkville Ives’ Dan Ives:

“Copilot is the second debate, and we think the monetization story is about to get clearer. Microsoft moved to per-seat plus usage pricing in September, added the $99 E7 SKU in October, and from December sells Copilot without a Microsoft 365 base subscription. Thirty million seats is the last disclosed figure; we want the update and color on usage revenue, since management has said Microsoft 365 commercial should accelerate through the year as consumption layers on top of seats.”

“In a week when Meta's $95 billion raise showed what staying at the frontier costs, Microsoft's ability to fund the same build from operations with double-digit operating income growth is what makes it our top pick, and it is why we view a higher capex number here as demand underwriting the build rather than a margin problem.”

Spoon-Fed

Some days at work, you get so busy you forget to eat. Or you get so busy you forget to be fed. Fed. Hmm…

Oh yes, there’s a Fed decision today! It really does feel like something of an afterthought, given:

All the AI news swirling;
Expectations for the central bank to be on hold, largely because many important Fed officials have done what their new leader does not like and provided forward guidance on their desired outcome ahead of the event (and in some cases before key data, it must be said);
Fed Chairman Warsh’s desire to not really say too much at press conferences; and
There being no updated “dot plot” or projections today

(I confess that I scheduled a haircut for 2 p.m. ET today because I forgot — only canceling it after being reminded by one of the world’s best Fed reporters, who surely thinks even less of me now. Onwards.)

I’m far from alone, it seems. 22V Research’s Dennis DeBusschere surveyed clients, and came back with this:

Quick summary, investors:

Are watching the FOMC tomorrow with less intensity than the prior two meetings
Anticipate a mixed/negligible reaction (62%)
- Risk-on (22%)
- Risk-off (16%)

When it comes to the Fed, I’m more focused on the potential for accelerating three/six month core inflation trends coinciding with another round of surprisingly large hyperscaler capex budget announcements as a potential scenario that could kick off another “inflationary boom, the central bank is behind the curve” type of market reaction. That probably isn’t a story for today.

But of course, when powerful policymakers speak, there’s always the potential for market fireworks.

Via SpotGamma:

“The SPX implied volatility term structure shows a pronounced event-driven volatility premium concentrated in this week’s expirations, with October 30 standing out near 16% IV, above the roughly 14% levels seen across subsequent weeks. This elevated pricing reflects the unusually crowded catalyst calendar, including Wednesday’s FOMC decision, major technology earnings, and key economic releases. Notably, the premium extends through Friday rather than being isolated to the Fed announcement, suggesting traders are pricing the cumulative risk of both FOMC and mega-cap earnings.”

Taking a bit of a longer view, JPMorgan strategists attempted to game out how both the Fed’s impact on short-term yields and investors’ assessment of the longer-term outlook for nominal growth can influence price action over time:

“A shallow hiking cycle (i.e., a reversal of last year’s insurance cuts) should be manageable for equities. If inflation reaccelerates and the forward curve starts to price in a broader hiking cycle (i.e., 4-5 hikes), equities will likely face more significant downside, although this is not our base case. The bigger question, however, is what happens to long-term rates and how much further the back-end could sell-off. Based on ~80 years of historical analysis, there is an inverted "U" relationship between the 10yr bond yield and S&P 500 multiples. The relationship is positive up until an inflection point, which is a function of the earnings growth backdrop. During periods of above-trend EPS growth (~15% y/y), 10yr yields can continue to rise to ~5% before the equity multiple starts to de-rate. Currently, forward consensus EPS growth is calling for 20%+ and the S&P 500 is trading at ~18x on 2027 EPS. If these strong growth projections are realized, history suggests that the multiple not only has room to re-rate higher but that equities should be able to withstand a 10yr yield level closer to ~5.5-6.0%.”

Impatiently Waiting

The CEOs of Nvidia and Broadcom have (at least) two things in common:

Along with providing the brains that power the AI buildout, they’ve also shown themselves to be very impatient when it comes to getting financial institutions to fund the boom.

Most recently, the Wall Street Journal reported that Hock Tan “was running out of patience” and personally intervened to get Wall Street to commit to a debt deal in which the custom chip specialist was offering some financial guarantees.

Now cast your mind back a few months to when Nvidia announced that a group of deep-pocketed lenders had agreed to finance $500 billion in AI compute. Per an August 14 Bloomberg report:

With slow progress on the complex deals, Nvidia’s chief executive officer, Jensen Huang, decided to change tack: He went public this week with the effort, saying the group is aiming to collectively finance AI computing deals totaling $500 billion — a round figure with no obvious provenance.

Consider the implications if the Bloomberg report is accurate — which given how little we’ve heard about this purportedly coordinated, mobilized pool of capital since, it certainly seems to be.

That would mean the leaders of some of the biggest financial institutions on the planet were somewhat blindsided by the head of the world’s most valuable publicly traded company, and then all went on TV together and smiled and thanked him for it!

This reminds me of a quote from author and statistician Nassim Nicholas Taleb:

“You are free in inverse proportion to the number of people to whom you can’t say ‘f**k you.’ But you are honorable in proportion to the number of people to whom you can say ‘f**k you’ with impunity but don’t.”

To modestly adapt Taleb, I suppose how powerful you become and remain is a function of how willing you are to swallow your pride in order to maintain profitable relationships. Less honor, more prudence.

The upshot: both Broadcom and Nvidia are pulling out all the stops to try to expedite the delivery of chips to customers, presumably in order to speed up their customers’ ability to use those products to make money as quickly as possible (so they can order more). But also, these measures are geared toward reducing the financial vulnerabilities of customers with less balance sheet strength than the publicly-traded hyperscalers. If the unit economics work, then the faster those clients are able to turn “spending commitments” into “profitable top-line growth” would be a significant positive for the ecosystem and eliminate some prospective air pockets.

For Nvidia, this strategy also plays into a larger theme of the company attempting to set industry standards: both by making AI compute a fungible asset class, and outlining quick-to-assemble data center designs that integrate its chips with offerings from preferred AI supply chain infrastructure providers. In particular, the latter is something you pursue when you’re trying to grease the wheels of AI commerce.

“Move fast and break things” was a fitting motto for the software industry. That axiom doesn’t lend itself too well to hardware, though, so “move fast to make things” is the mantra these chip designers are pushing, and it’s a vision that seems to be getting more difficult for capital markets to accommodate.

Oracle — the least creditworthy and most indebted of the major publicly traded hyperscalers — was the canary in the coal mine for investors’ distrust of externally funded AI capex campaigns.

In response, the company has adopted an approach that’s arguably the exact opposite of what Mag 7 hyperscalers and the biggest chip companies are doing. Oracle can’t really engage in “credit wrapping” because investors don’t love its wrapping paper. Instead, the company has used tactics like “bring your own hardware” to put the onus for covering some of the biggest expenses on its customers, rather than itself.

The cloud giant hosts its investor day on Wednesday, which may be a more useful barometer of investor appetite for the AI trade than its larger peers who report after the close and on Thursday.

“With FY27/FY28 expected to be peak CapEx years, we believe directional commentary on CapEx in FY28 and beyond and clarity around FCF inflection could ease investor concerns,” wrote Mizuho analysts led by Siti Panigrahi. “More visibility into OCI operating margins, asset useful lives and returns could help investors assess the economics of Oracle’s AI buildout.”

(Note: Oracle is currently in its fiscal 2027, which runs through the end of May)

The debate around Oracle not only encompasses concerns about AI financing needs, but also the nuts-and-bolts matters of getting capacity up and running amid supply constraints and, in some jurisdictions, rising political opposition.

“While facility-specific datapoints may continue to drive near term volatility, Oracle reiterated confidence in its broader capacity ramp given the diversity of supply coming online,” wrote Goldman Sachs analyst Gabriela Borges. “While Oracle's diversified pipeline reduces over-dependence on single sites, capacity delivery remains one of the most important risks to supporting its substantial backlog.”

Whispers from Wall Street

I’ve often written about the idea that semiconductors are the guiding light for markets, and everything else is just a residual running downstream of that price action. Well, this chart and commentary from Goldman Sachs really does underline the idea:

“Tech has become an increasingly important driver of risk appetite. The correlation between our cross-asset RAI and Nasdaq 100 performance relative to the S&P 500 equal weight has risen over recent years, reaching almost 1 in recent weeks. Differences in sensitivity to Semiconductor stocks help explain cross-asset performance over the past 3 weeks, more so than sensitivity to oil prices.”