ted kang
← All briefingsSep 3, 2026 / Morning edition

AI earnings carry the tape.

Rates ease from their highs while oil keeps climbing. The market is rewarding company evidence more selectively.

Archived briefingU.S. premarket · Exact cutoff not recorded
10Y Treasury R~4.75%Below the prior day’s 4.8% areaBrent crude R~$97.30Fourth consecutive daily rise reportedInitial claims R206KLatest weekly reading in the briefingBroadcom AI growth R221%Reported year-over-year revenue growth
V Verified primary source R Reported in saved briefing X Social commentary A Analysis Method & sources ↗
01 / Rates A

Relief has a condition.

The 10-year yield had backed away from its recent high. Oil near $97 still made it difficult to assume that financial conditions would keep easing.

02 / Earnings A

Strong numbers are not enough.

Broadcom’s large AI growth rate and weaker premarket share reaction showed how demanding expectations had become. The starting valuation still matters.

03 / Labor A

Payrolls are the next test.

The briefing’s payroll consensus was 56K. The market needed a labor outcome compatible with softer inflation pressure and continued earnings growth.

Companies /Earnings & catalysts
V Company disclosures

The businesses
behind the tape.

Broadcom, Snowflake and HPE all show demand. Their different stock reactions make this an expectations and business-model problem, not a simple vote for or against AI.

AVGOResults

Broadcom

Q3 FY26 · Released September 2

Exceptional growth still meets a demanding price.

V Revenue was $29.6B; AI-semiconductor revenue was $16.7B, up 221%. Free cash flow was $13.7B. Q4 total revenue guidance was approximately $34.8B.

X The archived morning brief reported shares down roughly 3% despite the growth. “Disappointing” describes the reaction to expectations, not a year-over-year revenue decline.

My read A

I read the gap between strong growth and a weak share reaction as an expectations question. The next work is on sustainable earnings and customer economics, not extrapolating a single growth rate.

Broadcom · Q3 FY26 results · Sep 2, 2026 Broadcom’s demand meets physical capacity · @wallstengine
SNOWResults

Snowflake

Q2 FY27 · Released September 2

AI is reaching core consumption.

V Product revenue grew 37% to $1.49B. Full-year product-revenue guidance increased from $5.84B to $6.07B; net revenue retention was 126%.

X The saved morning briefing reported a gain of more than 20%; that reaction is retained as a reported indication.

My read A

This is the software evidence I care about: customers consuming more of the core platform. A sharp earnings gap also raises the valuation hurdle. I would separate the quality of the result from the price of the entry.

Snowflake · Q2 FY27 results · Sep 2, 2026
HPEResults

Hewlett Packard Enterprise

Q3 FY26 · Released September 2

The server cycle has more than one winner.

V Revenue was $12.2B, up 34%, with a 16.2% non-GAAP operating margin and $1.0B of free cash flow. HPE raised its outlook.

My read A

The useful comparison with Dell is growth per dollar of working capital and the contribution from higher-margin networking. I would test whether the order book can convert while component supply and integration costs remain manageable.

HPE · Q3 FY26 results · Sep 2, 2026
Research /Ideas from X
X Chatter → A Thesis

What I would
work on next.

Specific companies, a reason to care, and a clear test. These are conditional research ideas from this edition’s window.

01
Next earnings cycle

Broadcom: underwrite what can actually be deployed.

Broadcom (AVGO)

The X post ties the scale of AI demand to land, power and data-center constraints. I would build the earnings case from deliverable capacity and customer funding, rather than treating management’s long-range ambition as booked revenue.

What would confirm it
Quarterly shipments, margins and cash flow support the expansion, and customer capacity comes online on schedule.
What would weaken it
Power or construction delays push deployment out, or growth requires commitments that weaken cash generation.

Direct Broadcom discussion on X; long-range numbers in the post remain attributed claims, not a base-case forecast.

Broadcom’s demand meets physical capacity · @wallstengine
02
Several quarters

Networking is worth separating from the chip headline.

Broadcom (AVGO) · Hewlett Packard Enterprise (HPE)

The networking post points to a second stream of spending beside accelerators. I would compare Broadcom’s networking exposure with HPE’s data-center networking growth, asking where switching costs and margins make the spending durable.

What would confirm it
Networking growth keeps pace with deployments and is accompanied by stable margins and cash collection.
What would weaken it
Custom architectures change supplier economics, competition reduces pricing, or shipment growth comes with a weaker margin mix.

The X post discusses Broadcom. HPE is my comparison, supported by its separate earnings disclosure.

Broadcom’s networking opportunity · @wallstengine
Source notes & original X posts 2

September 3’s saved research briefing was reviewed alongside the September 2 earnings disclosures. Its Grok email contained only an excerpt; the full linked Grok chat required access. The Broadcom X posts below were recovered directly from the September 2 timeline.

Company facts were checked against the linked disclosures on September 11, 2026. X labels identify social commentary; reading the original post does not verify its claims. Analysis was reconstructed for this archive revision using information available within each edition’s window.

Sep 2, 2026 · @wallstengine · Original post read
Broadcom’s demand meets physical capacity

The post discusses management’s AI ambitions alongside constraints in land, power and data-center deployment. Long-range targets are claims, not contracted revenue.

Sep 2, 2026 · @wallstengine · Original post read
Broadcom’s networking opportunity

A post about the expected growth of AI networking alongside custom accelerators; useful for testing which parts of the buildout capture value.

01 /The day in data
R Original cutoff
R Dated observations

The relief session

September 2 · completed U.S. cash session

%
Original briefing source ↗
R Dated observations

Company reactions diverge

September 3 · approximate premarket moves

%
Original briefing source ↗
02 /What matters
A Research perspective

The useful distinction inside technology

The premarket split between software and semiconductors argued for looking at company-specific expectations, guidance and positioning. It did not establish a permanent leadership change. A one-day earnings reaction carries more information about the surprise relative to expectations than about the whole industry’s future.

The macro limit to the recovery

The decline in yields gave equities some room, but higher crude prices worked in the opposite direction. I would want the recovery to survive another difficult inflation input before treating the rate relief as durable. The following morning’s jobs report was the immediate checkpoint.

03 /What came next at the cutoff
Sep 4

Official payrolls

Compare the result with the 56K consensus cited at the cutoff.

Next session

Earnings follow-through

See whether the initial reactions hold beyond the opening move.

Sep 10–11

PPI and CPI

Separate energy pressure from broader underlying inflation.

Sources & reading notes

Visual editorial reconstruction of the saved daily briefing at its original cutoff; not a live market feed. Figures marked R were reported in that briefing and have not been independently re-verified for this archive. Approximate quotes remain approximate. Intraday indications, highs and completed-session prices are labeled separately. Weekend and holiday U.S. cash observations are carried forward from September 4. Analysis and upcoming events describe the perspective at that original cutoff.

R
Daily Research Market Brief · September 3, 2026

Saved original briefing. Market and company observations are reported evidence; no claim of fresh verification or historical point-in-time dataset certification is made.

Back to the briefing archive ↗