Updated 5 August 2026. AMD reported Q2 after the close on 4 August and beat on every headline line — record revenue $11.54bn (+50%), non-GAAP EPS $1.66, data center $6.72bn (+107%) — and guided Q3 above consensus at about $13bn. The stock closed regular trade at $518.58 (+7.00%) and then fell 8.82% to $472.85 in after-hours dealing at 7:59pm ET, per StockAnalysis.
Our call: the pre-print straddle bracketed $544 and $425; the realised move landed inside it. Against the $472.85 after-hours price, the mean street target of $579 is the base case, $730 (UBS) the bull, and the $429.56 low from 29 July the honest downside — a traded level, not an analyst target, because every recently updated sell-side target now sits above the stock.
The most dangerous assumption heading into AMD’s print was that a beat means the stock goes up. It did not survive contact with the tape. Advanced Micro Devices delivered the best quarter in its history on 4 August — revenue up 50% to a record $11.54bn, data center revenue more than doubling to $6.72bn, non-GAAP earnings of $1.66 against a $1.61 consensus — and then handed back the entire 7% gain it had built during the session, and more, within ninety minutes of the release. At $472.85 after hours, the stock sat 2.4% below where it started Tuesday.
Two things explain that, and only one of them is about AMD. The first is that the guide, while above consensus, was not the blowout the buy side had positioned for. The second is that a stock up roughly 140% on the year going into a print does not need bad news to fall — it needs the absence of a new reason to keep paying up. Both were in evidence. What was not in evidence, despite a good deal of same-night commentary to the contrary, was a gross margin miss.
Key Facts — AMD Q2 2026 results, reported 4 August
- Revenue $11.536bn, up 50% year-over-year — a company record, against a $11.28bn consensus — per AMD’s results release
- Non-GAAP EPS $1.66 versus $1.61 expected; GAAP diluted EPS $1.38
- Gross margin: 56% non-GAAP — in line with the company’s own ~56% guide — and 54% on a GAAP basis
- Data Center revenue $6.718bn, up 107%, now roughly 58% of total company revenue
- Client & Gaming $3.841bn, up 6% (Client $3.062bn; Gaming $779m); Embedded $977m, up 19%
- Q3 guidance: about $13bn ±$300m — roughly +41% year-over-year and +13% sequentially, above the ~$12.52bn consensus — with non-GAAP gross margin again guided to about 56%
- Capital expenditure more than doubled, from $389m to $808m, a figure Futurum Equities’ Shay Boloor said the market “fixated” on, per Benzinga
- Share price: closed $518.58, +7.00% (+$33.94); after-hours $472.85, −8.82% (−$45.73) at 7:59pm ET on 4 August, per StockAnalysis. Market cap $845.6bn; 52-week range $149.22–$584.73
The gross margin “miss” that was not a miss
Within an hour of the release, several outlets attributed the selloff to a gross margin miss — 54% against a 56% consensus. That comparison does not hold up. AMD’s own results release reports non-GAAP gross margin of 56% and GAAP gross margin of 54%. The 56% consensus was a non-GAAP number. Setting the GAAP figure against a non-GAAP estimate produces a two-point “shortfall” that exists only in the comparison, not in the business: on a like-for-like basis, margin came in exactly where the company said it would, and the Q3 guide holds that same ~56% line.
This matters because margin was supposed to be the trapdoor. Chief Financial Officer Jean Hu had been explicit that the Helios rack-scale ramp carries near-term margin pressure, and the bear case going in was that the ramp would force a guide-down. It did not. The company guided September-quarter non-GAAP gross margin to approximately 56% while simultaneously guiding revenue about $500m above consensus. Whatever caused the after-hours decline, an eroding margin structure is not supported by the disclosure.
The capex line is the more defensible concern. Capital expenditure more than doubling quarter to quarter, from $389m to $808m, is real cash going into infrastructure ahead of revenue — the correct behaviour for a company ramping rack-scale systems, but a genuine drag on free cash flow in the interim, and a number that gets scrutinised harder when the multiple is already stretched.
What the market actually wanted: Helios, quantified
“AMD earnings were good,” said Futurum Group chief executive Daniel Newman, but the market “wanted a blowout guide driven by Helios,” per Benzinga’s post-print roundup. That is the whole story of the reaction in one sentence. A $13bn guide against a $12.52bn consensus is a beat by any ordinary standard; it is not a step-change, and a step-change was what the positioning implied.
Helios is AMD’s rack-scale AI system — Instinct GPUs, EPYC “Venice” CPUs and Pensando networking integrated into a single rack under the ROCm software stack. Patrick Moorhead of Moor Insights & Strategy noted that shipments are now starting with volumes expected to ramp in the fourth quarter, and cautioned against drawing conclusions before the full call. That timing is the crux: Helios revenue is a Q4-and-2027 event, while the stock is being asked to justify its multiple today.
The anchor customer remains signed but distant. On 22 July, AMD and Anthropic announced a partnership to deploy up to 2 gigawatts of Instinct MI450-series GPUs, with AMD committing a strategic equity investment of up to $5bn tied to deployment milestones, per AMD’s announcement — the deal FinanceFeeds unpacked in its breakdown of how AMD bought its way into Anthropic. The first gigawatt begins deploying in the first half of 2027. It validates demand; it does not add 2026 revenue. That gap between a signed order book and recognised revenue is precisely the space the after-hours sellers stepped into.
Chief Executive Lisa Su framed the quarter simply: “We delivered an excellent quarter, with record revenue and profitability as Data Center revenue more than doubled year-over-year.” On the numbers she is plainly right. Data center at $6.72bn is now 58% of the company, and it grew 107% while the rest of the business — Client and Gaming together up 6%, Embedded up 19% — did comparatively little. AMD is now an AI infrastructure company with a large legacy attachment, and it is priced as one.
The options market got the size roughly right
Going into the print, the straddle priced a move of about 12.28%, bracketing $544 on the upside and $425 on the downside from the $484.64 close on 3 August, per TipRanks. The realised after-hours move was −8.82% — inside that band, and close to the roughly 8.0% average absolute post-earnings reaction this stock has delivered. Traders who paid for the tails overpaid again.
The directional record, meanwhile, looks set to extend. AMD has beaten EPS estimates in 10 of its past 12 quarters yet finished lower the day after each of its last four reports. Q2 beat on revenue, beat on earnings, and guided above consensus — and still went down. That is now five prints in a row where the guide, not the beat, has set the direction, and where the answer has been the same.
Where the levels sit now
| Case | 12-month level | vs $472.85 | What has to happen |
|---|---|---|---|
| Bear | $429.56 (29 July low) | −9% | The only support that has held this summer, and almost exactly where the pre-print straddle set its downside. Gets tested if the Helios ramp slips, capex keeps climbing without matching revenue, or the AI-infrastructure trade de-rates further. |
| Base | $579 (mean street target) | +22% | The $13bn Q3 guide converts, ~56% margin holds through the Helios ramp, and Q4 volumes arrive on schedule. Consensus positioning was 28 Buys against 7 Holds into the print. |
| Bull | $730 (UBS, 24 July) | +54% | Helios orders get quantified beyond Anthropic and 2027 AI GPU revenue tracks toward the higher end of street models — Bernstein has argued the ~$33.5bn consensus for 2027 is too low against its own $43bn estimate. |
One caveat on that table worth stating plainly: every recently updated sell-side target sits above the current price. Among targets revised in the two weeks before the print, the lowest is Susquehanna’s $500 (raised 30 July), which is still about 5.7% above the $472.85 after-hours level; the others run UBS $730, Jefferies $640, Mizuho $625, Bernstein and Wedbush $600, and Bank of America $560. That is why the bear row above is anchored to a traded price — the 29 July low of $429.56 — rather than to an analyst estimate. Presenting the lowest target as a “bear case” would imply downside where the sell side is in fact unanimously modelling upside. Note also that these targets predate the print; revisions in the coming sessions are the thing to watch.
The overhangs that did not go away
Two forces outside AMD’s control still frame the next two quarters. The first is China. AMD’s MI308 accelerator can only ship to approved Chinese entities under licences requiring 15% of those sales be remitted to the US Treasury, and Lisa Su has kept expectations deliberately flat — the company has not forecast China AI chip revenue much beyond roughly $100m per quarter amid licensing uncertainty. Any China number is upside rather than baseline, and Huawei continues to take share domestically.
The second is input costs. TSMC is pushing through price increases of up to 10%, a cost neither AMD nor Nvidia can route around, as FinanceFeeds detailed in its analysis of TSMC’s price rises. Holding ~56% gross margin through a Helios ramp while absorbing foundry inflation is the operational task of the next year, and it is why the margin line on the Q4 guide will matter more than this quarter’s beat.
The competitive clock also runs. Nvidia’s Vera Rubin platform entered full production in June with partner availability in the second half of 2026, meaning Helios ramps directly into Rubin’s launch window while hyperscalers keep building their own silicon. The sector backdrop remains unforgiving after the late-July washout that erased more than $1tn in chip market value — a slide FinanceFeeds traced through the AI chip selloff that hit SK Hynix 14.65% and AMD 8.3% in a day, and which continues to move storage names like Western Digital, which reports tonight.
Quick Take: AMD beat on revenue, beat on earnings, held non-GAAP margin at the guided 56%, and guided Q3 about $500m above consensus — and fell 8.8% anyway. The widely repeated “margin miss” compares a GAAP number to a non-GAAP estimate. The real issue is timing: Helios revenue is a Q4-and-2027 story, capex has doubled to $808m, and the stock had already run 140% into the print. Base case $579, bull $730, and the honest downside is the $429.56 July low — not an analyst target, because every recent one sits above the price.
FAQ: AMD Q2 2026 earnings
What did AMD report for Q2 2026?
Record revenue of $11.536bn, up 50% year-over-year, against a $11.28bn consensus; non-GAAP EPS of $1.66 versus $1.61 expected, with GAAP diluted EPS of $1.38. Data Center revenue was $6.718bn, up 107%, making up about 58% of the company.
Why did AMD stock fall after beating earnings?
The Q3 guide of about $13bn, while above the ~$12.52bn consensus, was not the step-change the buy side had positioned for — Futurum’s Daniel Newman said the market “wanted a blowout guide driven by Helios.” Capex more than doubling to $808m drew scrutiny, and the stock had already risen roughly 140% in 2026 before the release.
Did AMD miss on gross margin?
No. AMD reported non-GAAP gross margin of 56%, in line with its own guidance, and GAAP gross margin of 54%. Reports of a “miss” compared the 54% GAAP figure with a 56% non-GAAP consensus, which is not a like-for-like comparison. Q3 non-GAAP gross margin is guided to about 56% again.
What is AMD’s Q3 2026 guidance?
Revenue of approximately $13bn ±$300m — about 41% growth year-over-year and 13% sequentially — with non-GAAP gross margin of about 56%. Consensus had been near $12.52bn.
How far did AMD stock fall after hours?
The stock closed regular trading on 4 August at $518.58, up 7.00%, then fell 8.82% to $472.85 by 7:59pm ET — giving back the whole session gain and finishing about 2.4% below the previous close of $484.64. The move was inside the roughly 12.28% the options market had priced.
What are analysts’ AMD price targets?
The mean target was near $579 into the print, with 28 Buys and 7 Holds. Targets updated in the fortnight before results ranged from Susquehanna’s $500 up to UBS’s $730, with Jefferies $640, Mizuho $625, Bernstein and Wedbush $600 and Bank of America $560. All of these predate the results and sit above the after-hours price.
What price levels matter now?
Support: the $429.56 low from 29 July, the only floor that has held this summer. Resistance: the $557.89 July high, and above it the 52-week high of $584.73.
When does Helios revenue actually arrive?
Shipments are starting now, with volumes ramping in the fourth quarter and through 2027. The Anthropic agreement — up to 2 gigawatts of MI450-series GPUs with up to $5bn of AMD equity investment tied to milestones — begins deploying its first gigawatt in the first half of 2027.
This article is for informational purposes only and does not constitute financial advice. Results figures are from AMD’s Q2 2026 results release; price data as of the 4 August 2026 close and after-hours session, sourced from StockAnalysis; analyst targets and commentary as cited above and dated prior to the results. Trading and investing carry risk, including the loss of capital. Always do your own research.
