The Morning 10

What

Alphabet printed $119.8 billion in revenue, up 24% against a $116.5 billion consensus — and none of it was the story. Quarterly capex hit a record $44.9 billion, the 2026 guide was raised to $195–205 billion from $180–190, 2027 was flagged to 'increase significantly' — and free cash flow went negative $5.9 billion, the first cash-burning quarter since the IPO. But this was not capex without payoff: Google Cloud grew 82% to $24.8 billion, cloud operating income more than tripled to $8.8 billion, and backlog reached $514 billion. The market sold the print 3.3% after hours to around 330.75.

If

The regular session holds the after-hours markdown above the 200-day near 323, the sell reads as multiple compression on a raised spend — digestible. An acceleration through that level would say the market is repricing the cash-burn itself, not just the guide.

Why

Alphabet is now the clearest test of the entire build-out: can revenue and operating profit compound faster than capex converts into depreciation, energy costs and standing infrastructure commitments? Last night it showed both sides of the ledger at once — the burn and the monetization Wall Street has been demanding from the hyperscalers. The naysayers will keep naysaying and the aficionados will keep cheering; the stock is sitting on a major level, and the tape gets to arbitrate.

Then

Watch the 200-day near 323 on Alphabet — and watch where the $44.9 billion lands. That guide is next year's revenue line for the semiconductor complex in points 4 and 7.

What

ServiceNow beat on every metric that matters — $0.90 against $0.86, revenue $3.99 billion against $3.93 — after going into the print down 6.47% on the day. The after-hours reaction faded from +6 during the call but settled at +4.76%: the first software beat that got bought after three consecutive beats were sold. The framing on the call was the real event: management positioned ServiceNow as the orchestrator of AI in the enterprise.

If

Today's regular session confirms the after-hours bid, the four-beat sold-streak in software is broken — and the repricing question shifts from 'is software over' to 'which software layer collects'.

Why

'Orchestrator of AI in the enterprise' is the same land-grab language that 'system of record' was in the SaaS era — the claim that built two decades of moats for the companies that made it stick. The layer that routes agents, holds the workflow and meters the usage is the layer that prices the platform premium. That is precisely the control-plane thesis our Agentic Winners cohort is built on, and last night the market paid the claim for the first time in four prints.

Then

Score whether the +4.76% survives the regular session — and watch whether the orchestrator framing spreads in tonight's SAP call. Two enterprise vendors making the same claim in one week is a category forming.

What

Tesla missed by a third — $0.33 against a $0.50 consensus — and settled down 4.1% after hours near 358.80. IBM's quarter was the subtler event: revenue of $17.16 billion missed, the 2026 growth outlook was trimmed to 4–5% from above 5% — and inside the print, Z mainframe revenue fell 42% while distributed infrastructure, Power servers and storage, grew 37%, its strongest result on record, as customers redirected capex toward supply-constrained AI hardware. On the same call IBM committed 8,000 forward-deployed engineers to installing AI inside client enterprises — the largest such commitment ever made. The market paid nothing for any of it: IBM settled at 204.95, directly on its 52-week low of 204.44.

If

IBM holds the 52-week low through today's session, the floor thesis survives the guide-down — a services-led AI pivot plus a record distributed-infrastructure quarter, priced at zero, is a cheap option if the deployment revenue converts in coming quarters.

Why

One income statement showed the whole rotation at once: the transaction machine shrinking 42%, the scale-out AI iron growing 37%, and the company betting its future on deploying intelligence rather than selling big iron. IBM says the mainframe drop is cycle timing, not migration — but the budget displacement is real, and it is the same shift Alphabet's server-heavy capex and Nokia's doubled AI-and-cloud sales describe from the receiving end.

Then

The 8,000 goes onto our Forward Deployment board today as the largest single entry of the series — and the mainframe-to-cluster shift is today's Pulse. Tesla's revenue detail gets read in daylight; the after-hours verdict was about the miss.

What

The two-part trigger that has governed this drawdown is complete. The technical half went into ink Tuesday at SOXX 552.69 — twenty points through the 530–532 band — and Wednesday's close at 555.52 held the break, up on a day the rest of the tape gave back. The fundamental half arrived last night, louder than required: the contract asked for a capex guide held at path, and got a guide raised fifteen billion at the midpoint, with 2027 flagged higher on top.

If

The complex opens heavy this morning on Alphabet's red print, that is the last trap to avoid mis-reading: the funder's stock being sold and the funder's spending being raised are opposite signals for the supplier complex — the second one is the one semiconductors trade on.

Why

A breakout that held its first pullback, funded by a spender who just raised the budget it feeds, is the aligned confirmation this entire sequence was designed to wait for. The discipline of the last three weeks — three sold reclaims, a defended floor, a tactical starter instead of a full position — existed for exactly this morning.

Then

Today is execution day, not analysis day. What that means for the book is point 11.

What

While semiconductors held their breakout, software fell a fourth consecutive session Wednesday: IGV down 3.05% to 89.02, the sharpest day of the streak. Then overnight the most-watched name in enterprise software beat everything and got bought, and Alphabet reported cloud growth of 82% — the demand line software bulls have been waiting to see land somewhere.

If

IGV bounces today on the ServiceNow print, the four-day slide reads as positioning ahead of a binary that resolved well — tactical, recoverable. If software stays sold through a bought beat and an 82% cloud number, the rotation out of the layer is structural, and one good print doesn't turn it.

Why

This is the cleanest test the software repricing question has been offered all month: the best possible news arrived overnight, from two directions at once. A layer that can't rally on its best news is a layer the market has re-rated, not one it is rotating around.

Then

Today's IGV close against 89.02 is the single most informative non-semiconductor number of the session.

What

Nokia opened the European print day with a beat: comparable operating profit of €434 million against a €382 million estimate, up 18%, on sales up 9% at constant currency — with Network Infrastructure up 12%, Optical Networks up 20%, IP Networks up 16%, and sales to AI-and-cloud customers up 105%. The full-year guidance range moved up to €2.1–2.6 billion, technically a reclassification, but the direction of the print was unambiguous. STMicroelectronics reports before the US open; SAP reports after the European close tonight.

If

STMicro and SAP extend the pattern — vendors into the AI build-out beating while the spenders get scrutinized — the spender-pays-vendor-collects frame stops being a US reading and becomes the global tape.

Why

Nokia doubling its AI-and-cloud customer revenue is the same signal as Alphabet's raised guide seen from the receiving end: the capex leaves one income statement and lands on another. AI clusters generate enormous east-west traffic — between accelerators, between racks, increasingly between data centers — so the network is becoming part of the compute system rather than the connection to it. Europe's networking, silicon and enterprise-software names are all downstream of budgets that were confirmed higher last night.

Then

Print-record cards for SAP and STMicro run today. SAP's call is the one to hear the orchestrator language test — point 2's category question, asked in Walldorf.

What

Intel reports after tonight's close — consensus near $0.21 on roughly $14.4 billion — and it carries the most binary reaction profile of any name this season: seven EPS beats in the last ten prints, but a 3-0-7 reaction record with zero flat outcomes, spanning a −31.7% single-day loss and a +26.6% single-day gain, on a 52-week range of 19.31 to 140.94. It went into the print down 2.68% at 102.62.

If

The foundry commentary lands well on a night the largest customer class just raised its infrastructure budget, the print catches the same tailwind the rest of the complex trades on. The record says the market will pick a violent direction either way.

Why

Intel is the referendum on whether the build-out's money reaches the West's own manufacturing floor. After a five-fold move off the low, the position sizing question isn't whether the story is real — it's how much of it tonight's number already carries.

Then

Score it tonight against the 3-0-7 record. It's the last US semiconductor print before the Mag-7 week begins Friday with Microsoft.

What

IEF closed Wednesday at 93.10, a hair above the 93 support that has held all month; TLT printed a fresh low at 83.44. Overnight the largest hyperscaler raised its capex path by fifteen billion at the midpoint and flagged 2027 higher still — spending financed at exactly the maturities that keep cheapening.

If

IEF loses 93 on a close this week, the long-end drift graduates to a break with the next shelf two points lower — and the equity market's re-rating of an accelerating capex cycle starts paying a rising discount rate on every incremental dollar of guide.

Why

A capex cycle this size is a duration bet — data centers monetize over decades, and last night the bet got fifteen billion larger while the long bond made a new low. Both trades cannot stay comfortable indefinitely. The equity tape is currently choosing not to look; 93 on IEF is where it would have to.

Then

Same line as yesterday, now with higher stakes: watch 93 into the ECB decision and US claims today. A confirmed capex boom and a broken bond floor on the same tape is only half a green light.

What

Behind the print drama the cross-asset board stayed orderly: gold rose another 1.15% Wednesday to 379.12, extending its quiet uptrend; VIXY was essentially flat at 20.72 — no vol bid into a four-print night, which is itself information; and Money Temperature reads 50, dead neutral. The three scoring lines all held: SOXX 555.52 above the band, SPY 747.41 above its 50-day, QQQ 705.35.

If

Vol stays offered through today's open and the give-backs keep getting absorbed at higher lows, the market is treating the verdict week as resolution, not risk — the regime the last three weeks were waiting to confirm.

Why

A tape that goes into its biggest binary night of the quarter without bidding volatility, and comes out of it with the epicenter green and the indices intact, is a tape that had already positioned for the answer it got. Gold rising alongside — hedging demand coexisting with risk appetite — is the same pairing as last week, and it keeps resolving upward.

Then

US claims and the ECB decision print into a calm board today. The bar for either to matter is high — unless point 8's bond floor is the thing that moves.

What

Gene Munster's framing — AI infrastructure spending is a multi-year compounding cycle, with the physical layer paid before software — was set against last night's numbers in this space yesterday. The grade came in: the largest spender raised its budget to $195–205 billion and flagged 2027 higher, funding his second wave outright. But the tape handed him an asterisk — the night's best-bought print was a software company, on an orchestration claim.

If

N/A

Why

Both things came true at once: the infrastructure wave got bigger, and software found the one layer the market will still pay for. His sequencing holds — infrastructure before software — but ServiceNow's bought beat suggests 'software eventually' may arrive selectively, through the control plane, rather than as a layer-wide rerating.

Then

Read his post-print take at Deepwater directly, and set it against the IGV close in point 5 — the one number today that says whether 'eventually' has started.

Gene Munster · Deepwater Asset Management ↗