AI semiconductors are the single most important earnings complex in the market. A handful of names — $NVDA, $AMD, $AVGO, $TSM, $ASML, $MU, $MRVL — set the tone for the entire S&P 500, because their guidance is the cleanest real-time read on hyperscaler capex. This playbook is how to prepare for the cluster: which print to anchor on, what numbers actually move the group, and how to size positions when the whole complex re-rates in a single after-hours session.
Key numbers to watch
Consensus, prior-quarter or whisper context, and why each line matters.
| Metric | Consensus | Prior / whisper | Why it matters |
|---|---|---|---|
| NVDA Data Center revenue | ~$36B | Prior $35.6B (+66% YoY) | Single most important line — sets the floor for everyone else. |
| NVDA next-Q revenue guide | ~$42B | Street range $40–44B | Guide drives the gap; in-line at the top often isn't enough. |
| AMD Data Center revenue | ~$5.2B | Prior $4.6B | MI300/MI325 ramp — the only credible second source to NVDA. |
| AVGO AI revenue (ASIC + networking) | ~$5.5B | Prior $4.4B | Hyperscaler custom-silicon proxy — Google TPU, Meta MTIA. |
| TSM CoWoS capacity commentary | 2x YoY by year-end | Prior 'on track to double' | Packaging is the bottleneck; capacity = ceiling for the whole group. |
| ASML China revenue mix | <25% | Prior 27% | Export-control sensitivity — moves the equipment names. |
| MU HBM revenue | ~$2.0B | Prior $1.5B | Sold out through next year — pricing power read-through. |
| NVDA options-implied move | ±7% | Avg realized 4Q: 8.2% | Group beta to NVDA print is 0.7–1.0x for AMD/AVGO/MRVL. |
Consensus figures are Street estimates as of publish date and shift as analyst revisions land. Live consensus and implied move are on the stock page.
Why the order of reporting matters
The AI semi cluster reports in a sequence that lets you build conviction (or fade it) as the season unfolds. The typical order — $TSM first, then $ASML, then the US designers ($NVDA, $AMD, $AVGO, $MRVL), with $MU bridging memory — means every print updates the model for the next one.
The practical implication: $TSM's CoWoS commentary and $ASML's bookings are the leading indicators for the entire group. If TSM walks back capacity expansion or ASML's China mix surprises, you're getting a free preview of NVDA two weeks early. Don't trade NVDA into the print without watching these two first.
The four numbers that actually move the group
Across every print in this complex, four lines drive the after-hours tape more than anything else:
- Data Center revenue — the AI proxy. Beats here lift the whole group; misses crush it regardless of total revenue.
- Next-quarter revenue guide — an in-line guide after a beat usually sells off.
- Gross margin guide — expanding GM = pricing still holds; compressing GM = the supply bottleneck is clearing.
- Hyperscaler capex commentary — any management mention of customer capex plans is worth more than the headline EPS beat.
Hyperscaler capex: the upstream signal
The four US hyperscalers ($MSFT, $GOOGL, $META, $AMZN) collectively set the demand floor for every AI chip and every networking switch. Their quarterly capex prints come before most semi names report, giving you a one- to two-week window to position into the semi cluster armed with fresh capex guidance.
The read-through is mechanical:
| Hyperscaler signal | Semi read-through |
|---|---|
| Capex guide raised | Bullish NVDA, AVGO, MRVL, TSM |
| Capex held, mix shifts to custom ASIC | Bullish AVGO, neutral NVDA |
| Capex held, mix shifts to GPUs | Bullish NVDA, neutral AVGO |
| Capex guide cut | Bearish entire group, especially $MU |
One hyperscaler cutting is noise. Two cutting in the same season is a regime change — that's when you flatten the book.
Top 3 beneficiaries — and top 3 most at risk
Once you have the hyperscaler tape and TSM/ASML in hand, the cluster splits cleanly. The top tier compounds the AI demand signal; the bottom tier is exposed if any of the four key numbers wobble.
Two scenarios for the next cluster
Scenario A — capex confirms, beats continue. Hyperscalers raise capex guides; TSM lifts CoWoS capacity again; NVDA beats and guides above consensus. The group rallies 10–20% over two weeks and the AI trade re-asserts leadership.
Top 3 to lean into:
| Ticker | Setup |
|---|---|
| $NVDA | Highest beta to the signal; cleanest expression |
| $AVGO | Re-rates on custom-ASIC pull-through |
| $MU | HBM pricing power; smallest cap = biggest move |
Scenario B — capex held, mix shifts, NVDA guides in-line. Group sells off 5–10% on the gap, then bifurcates. Custom-ASIC and networking outperform GPUs; memory weakens.
Top 3 to avoid (or hedge):
How to size into the cluster
Three rules that have held up across the last six AI earnings seasons:
- Size based on dispersion, not direction. Implied moves regularly run 7–10%; use defined-risk option structures sized so a two-sigma loss is survivable.
- Anchor positions on $TSM and $ASML, not $NVDA. The upstream names price the bottleneck, not the narrative.
- Have one hedge on at all times during the cluster — a short basket of high-beta AI second-derivatives, or a SPY put 5% OTM dated past the last big print.
The AI semiconductor complex is the highest-information, highest-volatility cluster of the year. Treat it like one trade, not eight separate ones.
Use this on Earnings Compass
Frequently asked questions
- Which AI semiconductor stock reports first each quarter?
- $TSM typically reports first, followed by $ASML, then $NVDA, $AMD, $AVGO and $MRVL, with $MU bridging memory. TSM's CoWoS packaging commentary and ASML's bookings are the cleanest leading indicators for the rest of the group.
- What is the most important number in an NVDA earnings print?
- Data Center revenue and the next-quarter revenue guide. A beat on headline EPS with an in-line guide often sells off because the stock is priced for continued acceleration. Hyperscaler capex commentary on the call is the third line that moves the group.
- How do hyperscaler earnings affect AI semiconductor stocks?
- $MSFT, $GOOGL, $META and $AMZN set the demand floor for every AI chip. A raised capex guide is mechanically bullish for $NVDA, $AVGO, $MRVL and $TSM. Two hyperscalers cutting capex in the same season is a regime change and the cue to flatten exposure.
- Why is CoWoS packaging the bottleneck for AI chips?
- $TSM is the sole-source supplier of CoWoS advanced packaging used in every leading-edge AI accelerator. Its capacity expansion pace caps how many $NVDA and $AMD GPUs and $AVGO custom ASICs can ship — making TSM the ceiling for the whole group's revenue.