The Week Ahead for Semis and Tech: Four Catalysts, and What Each One Can Actually Move
📅 The Week Ahead for Semis and Tech: Four Catalysts, and What Each One Can Actually Move
Welcome back, everyone! 📊 I'm your Quant Analyst, filtering out market noise using data, statistical modeling, and systematic insights. 👩💻✨
Next week gives semiconductors one company event and three macro prints. Most previews will list them and stop there, which is the part I find useless — a calendar is not analysis. What matters is which channel each event transmits through, because that determines whether it moves earnings expectations or the discount rate, and those two do very different damage.
My headline view: Broadcom’s print is the loudest event but the least uncertain one, because its AI revenue is already largely pre-sold by backlog — the genuine unknown is the half of the company nobody is talking about. Meanwhile the highest-variance event for semiconductor prices is Friday’s jobs report, for reasons that have nothing to do with chips.
▲ Four dated events, two transmission channels — the calendar is the easy half
📌 First, the confirmed calendar
Every date below is from the issuing organisation, not a third-party calendar. Tue Sep 1, 10:00 ET — ISM Manufacturing PMI (August data). Wed Sep 2, after the close — Broadcom Q3 FY2026 results, call at 2:00pm PT. Thu Sep 3, 10:00 ET — ISM Services PMI. Fri Sep 4, 8:30 ET — the August Employment Situation from the BLS.
Broadcom’s last reported quarter (Q2 FY2026, ended May 3, 2026) was revenue of $22,187 million, up 48%, with AI semiconductor revenue of $10.8 billion, up 143%, semiconductor solutions of $15,009 million and infrastructure software of $7,178 million. Q3 guidance is ~$29.4 billion total (+84%) and $16.0 billion of AI semiconductor revenue.
Sources: Broadcom Q2 FY2026 results release; ISM report calendar; BLS Employment Situation.
1. Wednesday: Broadcom, and Why the AI Line Is the Boring Part
Start with the segment dispersion inside Broadcom’s own last quarter, because one income statement holding management and balance sheet constant tells you more than four tickers do.
One Company, Three Very Different Businesses
*Reported figures from Broadcom’s Q2 FY2026 release, not estimates. AI semiconductors are a subset of semiconductor solutions, so the first two bars overlap by construction.
AI semiconductors grew 143%. Infrastructure software grew 9%. That is not one company having a good quarter — it is two companies stapled together, and only one of them is in the AI trade. It is the same gradient I found inside the broader semiconductor complex: growth amplitude tracks AI exposure almost mechanically.
Now the part that changes how I’d read Wednesday. Broadcom disclosed over $30 billion of AI semiconductor bookings in Q2 against $10.8 billion shipped — a book-to-bill near 2.8×. Those Q2 bookings alone are 1.9× the entire $16.0 billion AI revenue the company guided for Q3.
The Revenue Is Largely Already Sold
*From Broadcom’s Q2 FY2026 disclosures. Bookings and revenue are different measures over different periods — read the gap as backlog depth, not as a quarter-to-quarter comparison.
When a quarter’s revenue is that heavily pre-committed, the revenue line stops being a surprise variable and becomes an execution check. So the useful question is not “will AI revenue be big?” — it will — but what the rest of the guide quietly assumes.
Here is the arithmetic nobody seems to be doing. Q3 total revenue is guided to about $29.4 billion, of which $16.0 billion is AI. That leaves roughly $13.4 billion of non-AI revenue. In Q2 that same non-AI bucket was about $11.4 billion. So the guide embeds roughly +18% sequential growth in the non-AI half — the half whose software segment just grew 9% year over year.
That, to me, is the line to watch on Wednesday. It may well be met; software revenue is lumpy and renewal-driven. But it is the assumption carrying the most weight relative to how little attention it gets, and it is where a “beat on AI, miss on total” outcome would come from.
2. Tuesday and Thursday: The ISM Prints Are a Breadth Test
The two ISM surveys are the week’s cheapest information. They matter for semiconductors in a specific way that gets lost in the “is the economy okay” framing: they speak to the non-AI half of the chip cycle.
Analog, industrial and automotive silicon tracks manufacturing activity with a lag. If ISM manufacturing is expanding, the recovery I found in the broader complex has a floor under it; if it rolls over while AI keeps compounding, the market gets a cleaner story than it wants — the sector isn’t recovering, one end market is. Services matters less directly for chips, but it is the better read on whether the labour market Friday will confirm.
3. Friday: The Jobs Report Is the Highest-Variance Event for Chip Prices
This is the claim that sounds wrong and isn’t. Broadcom’s print carries more information about Broadcom; the jobs report carries more price risk for the sector, because it moves the denominator under every multiple in it.
I published the arithmetic on this two days ago: a +100bp move in the discount rate takes roughly 23% off the fair value of a 30× multiple and about 11% off a 12× one. Semiconductors are, as a group, at the expensive end of that range. A jobs number that repriced Fed expectations by even a quarter of that would do more to semiconductor prices than a normal earnings beat.
Context for Friday that most previews will skip: on August 28 the BLS published its preliminary benchmark revision, marking payrolls down by 79,000 for the year through March. That is a real downward revision, but a smaller one than the consensus had feared — which means the labour-market narrative going into Friday is genuinely unsettled rather than one-sided.
| Event | Channel it moves | Who it hits hardest |
|---|---|---|
| ISM Manufacturing Tue |
Earnings — non-AI end demand | Analog, industrial, automotive chips |
| Broadcom results Wed, after close |
Earnings — custom-silicon read-across | AI accelerator and networking names; the ASIC-vs-GPU debate |
| ISM Services Thu |
Discount rate — via inflation path | Broad market more than chips specifically |
| August jobs report Fri, 8:30 ET |
Discount rate — directly, via the front end | The highest multiples in the sector, mechanically |
4. What I Am Not Claiming
I am not forecasting these numbers, and I would be suspicious of anyone who does with confidence. What I have tried to do is something narrower and more defensible: establish what is already known (a heavily pre-sold AI revenue line, a guide that embeds 18% sequential non-AI growth, a benchmark revision that landed softer than feared) so that the new information on the day is easier to isolate.
The framework is falsifiable, and here is how it fails:
- If Broadcom moves the sector more than the jobs report does. That would mean the earnings channel is dominating the rate channel right now, and my ordering of the week is simply wrong.
- If the non-AI half comes in at +18% sequential without drama. Then the guide was conservative in exactly the place I flagged as stretched, and I should stop treating the software segment as the soft spot.
- If ISM manufacturing contracts and analog names rally anyway. That breaks the “non-AI chips track manufacturing activity” link the sector map above depends on.
- If high-multiple semis don’t underperform low-multiple ones on a hot jobs print. The duration arithmetic says they should. If they don’t, the discount-rate channel is not currently the dominant driver, and I need to reason about this sector differently.
Quick FAQ
Q. Should I trade around these events?
I can’t advise on your portfolio and I’m not licensed to. What I’d say generally is that event weeks reward having decided in advance what each outcome means, and punish deciding at 8:31 on Friday morning. The plan is the edge, not the prediction.
Q. Why does a networking-and-software company matter for the GPU debate?
Because Broadcom is the largest merchant supplier of the custom accelerators and networking silicon that hyperscalers use as an alternative to buying general-purpose GPUs. Its AI number is therefore the cleanest public read on how fast that substitution is actually happening — a question I looked at in whether anything can replace the GPU.
Q. Isn’t a 2.8× book-to-bill just unambiguously good?
Mostly, but not entirely. Deep backlog also means revenue is capacity-limited rather than demand-limited, so upside surprises get capped by supply, and the bookings themselves become the number that can disappoint. A backlog that stops growing is a warning even while shipments look excellent.
💡 Quant Strategy & Takeaways
Broadcom’s AI line is roughly pre-sold — Q2 bookings alone are 1.9× the whole Q3 AI guide — so the real question on Wednesday is the non-AI half the guide quietly assumes will grow 18% sequentially. And for sector prices, Friday outranks Wednesday, because rates move the denominator under every multiple.
Separate the events that move earnings from the ones that move the discount rate, and the week stops looking like noise. 🤖
Which one are you actually watching — the print, or the payrolls? I’d like to know which channel you think dominates right now. 📈✨
Disclaimer: This article is quantitative research published for informational and educational purposes only. It is not financial advice or a recommendation to buy or sell any security. Scheduled dates are as published by the issuing organisations at the time of writing and can change; company figures are as reported.
Comments
Post a Comment