AMD Q2 2026 Earnings: The Beat, the $13 Billion Guide, and a Sell-the-News First Reaction. Updated With Results

---

UPDATE (August 4, 2026, evening): the print beat, the guide beat, and the stock fell anyway. Q2 revenue was $11.54 billion against $11.28 billion expected, above even the top of the guided range, with adjusted EPS of $1.66 against $1.62. The Q3 guide came in at $13 billion plus or minus $300 million against a roughly $12.5 billion consensus. The data centre segment was over 58% of revenue, which works out to about $6.7 billion against the $6.5 billion previews clustered on (our arithmetic from the disclosed share, not a company line item), roughly double the year-ago $3.24 billion. The first extended-hours reaction was a drop of about 9-10%: the stock had gone into the print up more than 8% on the day near $524, so that drop roughly hands back Tuesday's rally rather than anything structural. All of tonight's quotes are extended-hours snapshots taken with the 5pm call still underway, not a close; Wednesday's session is the real verdict. One flag: early coverage cited a 54% gross margin against the roughly 56% guided. Which basis that is was not clear at writing, and that line, next to a valuation near 53x forward earnings that needed a blowout rather than a beat, is the likeliest explanation for the reaction. Trade-log scoring notes are below the table.

---

TL;DR

• AMD reports Q2 2026 results tonight, Tuesday August 4, after the close, with the call at 5:00pm ET. It is the last big AI chip read of the busiest earnings week of the quarter, and it lands the same evening as SpaceX's first public print.
• Consensus wants $1.61-1.62 of adjusted EPS on about $11.3 billion of revenue, up 47%, against a company guide of $11.2 billion plus or minus $300 million. Consensus sits inside the range, so the headline is close to pre-agreed: the guide and the data centre line are the print.
• Sell-side previews cluster around $6.5 billion of data centre revenue, which would be up roughly 100% from $3.24 billion a year ago. That single line is the AI capex argument, receipts edition.
• Options price a move of about 12.3%, bracketing roughly $425 to $544 from Monday's close near $484.64, in a stock up about 122% this year and roughly 17% below its June 30 high of $584.73.
• Palantir just cleared its whisper overnight; both Monday prints landed hot. AMD decides whether the week's AI narrative confirms or cracks.

What Time Is AMD's Earnings Report?

Tonight, Tuesday August 4, shortly after the 4:00pm ET close, with the conference call at 5:00pm ET. The quarter being reported ended in late June. SpaceX reports the same evening, which is why the week-ahead playbook said not to hold short-dated premium through both.

The Board

Consensus sits inside the guide. The quarter is pre-agreed; the data centre line and the Q3 guide are not.

The Headline Is Pre-Agreed. The Data Centre Line Is Not

Management guided $11.2 billion plus or minus $300 million with adjusted gross margin near 56%, and consensus has settled at $11.3 billion, comfortably inside the range. That makes the revenue beat close to worthless as a signal: this season has taught, repeatedly, that the guide beats the quarter.

The line with actual information in it is data centre revenue, where previews cluster near $6.5 billion against $3.24 billion a year ago. AMD is the seller in a week full of buyers: Microsoft showed $678 billion of contracted backlog, Amazon showed AWS accelerating, and every one of those capex promises has to convert into somebody's orders. If the hyperscalers are spending what they say, it shows up here. A data centre miss would be the most damaging single data point the AI trade has had this season, because it would mean the promised spending is not converting.

Underneath it, management has already claimed server CPU revenue growing more than 70% this quarter. The year-ago comparison is friendly: Q2 2025 carried the dent from curtailed MI308 shipments to China, which is part of why the adjusted EPS comp ($1.62 consensus against $0.48 a year ago) looks so violent.

The Anthropic Deal Is the Call Topic

On July 22 AMD and Anthropic announced a partnership to deploy up to 2 gigawatts of Instinct MI450-series GPUs, with the first gigawatt targeted for the first half of 2027 and AMD investing up to $5 billion in Anthropic. That deal reframed AMD's AI story from "second source to Nvidia" toward "named anchor customer at gigawatt scale", and tonight is the first chance to size the MI450 and Helios rack order book on the record. Any specificity on 2027 deployment timing moves the stock more than the Q2 numbers do.

The Setup Cuts Both Ways

The stock is up about 122% this year and closed Monday near $484.64, roughly 17% below its June 30 record of $584.73. A 12% implied move on a $780-odd billion company is the options market admitting it has no idea: the bull case is a data centre blowout plus an accelerating Q3 guide into a tape that just rewarded Palantir for exactly that shape; the bear case is that a stock that has more than doubled this year needs acceleration, not arrival, and an in-line guide gets sold the way Broadcom's +200% AI guide got sold.

The Options Angle

The implied move is about 12.3%, bracketing roughly $425 to $544. Two structural facts shape the plays: tonight is a two-event evening (SpaceX prints alongside), and July's calibration lesson still stands: realised moves have been beating implied all season, so "the straddle looks expensive" is not, on its own, a reason to pass.

• Skip short premium in any structure. Consensus inside the guide means the reaction is all about the unguided lines, and a 12% gap through a wing on a $484 stock is unrecoverable.
• The straddle is a genuine coin flip, not a donation. At ~12.3% you need a Palantir-sized reaction to profit. We pass, narrowly, because unlike the memory names there is no run of AMD realised moves beating implied to lean on, and we log the pass to be scored.
• The confirmation trade is the play. If data centre clears ~$6.5 billion and the Q3 guide implies acceleration, own it Wednesday with shares or calls and time on your side.

Trade log

#	Stance	Structure	Strikes and expiry	Cost or credit	Spot at writing	Implied move	Breakeven
1	Pass	Long straddle into the print	~$485 line, Aug 7 weekly	Live chain not sourced; implied ~12.3% of spot	$484.64, Aug 3 close	~12.3%	needs a move beyond ~12.3%
2	Pass	Short premium into the print (any structure)	Aug expiries	Not sourced	$484.64, Aug 3 close	~12.3%	scored on whole position; both tails live
3	Conditional	Post-print long (shares or 1-2 month calls) if data centre clears ~$6.5bn and the Q3 guide implies acceleration	Struck off the Aug 5 open	Struck off the Aug 5 open	To be struck Aug 5	n/a	Scored against the post-call entry if triggered

Scoring update, August 4 evening. Row 1, the straddle pass, is currently right: the first after-hours reaction of about 9-10% sits inside the ~12.3% implied, exactly the premium-seller's outcome we declined to pay for. Row 2 is inside the implied so far and gets scored on Friday's tape. Row 3's conditions both printed: data centre cleared the bar and the guide beat consensus, so per the plan the long triggers off the Wednesday, August 5 open, and it will be struck and scored against that opening price. The uncomfortable part of taking it is the reaction itself; a written plan does not get to renegotiate with the tape. Final scores for all three rows land in the next scorecard.

The One-Line Read

The revenue number was agreed months ago inside AMD's own guide; tonight is really a referendum on one line and one sentence: a data centre print near $6.5 billion says the AI capex promises are converting into orders, and the Q3 guide says whether they keep converting, in a stock priced at more than double where it started the year.