Research Finder
Find by Keyword
AMD Q2 Record $11.5B Revenue Beat Overshadowed by Helios Margin Ramp?
EPYC carried the quarter, Helios ships later this quarter, and the TAM revision arrives well before the accelerator revenue does while peer prints from Intel, TSMC and the memory makers suggest supply, not appetite, is still setting the pace in AI compute.
08/06/2026
By the Numbers
- Revenue: $11.54 billion, up 50% year over year and 13% sequentially, ahead of consensus in the $11.25 billion to $11.35 billion range
- Non-GAAP diluted EPS: $1.66 versus consensus of roughly $1.61, up 246% as reported and up approximately 82% on the comparable basis management uses, which excludes prior-year MI308 export control charges
- GAAP diluted EPS: $1.38 versus $0.54 a year ago; GAAP net income of $2.30 billion included $483 million of gains on long-term investments recorded below the operating line
- Data Center: $6.72 billion, up 107% year over year and 16% sequentially, ahead of a roughly $6.5 billion segment consensus; segment operating income $2.10 billion (31% of segment revenue) against a $155 million loss a year ago
- Client and Gaming: $3.84 billion, up 6% year over year; Client $3.06 billion, up 23%; Gaming $779 million, down 31%; segment operating income $582 million (15%) versus $767 million (21%) a year ago
- Embedded: $977 million, up 19% year over year, segment operating income $386 million (40% margin)
- Gross margin: 56% non-GAAP (54% GAAP), up more than 200 basis points year over year on the adjusted comparable basis and up 1 point sequentially
- Non-GAAP operating margin: 27% versus 12% a year ago; non-GAAP operating expenses of $3.39 billion, up 40% year over year against 50% revenue growth
- Cash flow: operating cash flow of $2.37 billion and free cash flow of $1.56 billion (14% margin), against $2.57 billion and a 25% margin in Q1 2026; capital expenditures of $808 million versus $389 million in Q1 2026
- Balance sheet: cash, cash equivalents and short-term investments of $13.11 billion; inventories of $8.47 billion; total debt of $3.23 billion
- Capital allocation: no share repurchases in the quarter (the $221 million repurchased year to date occurred in Q1); $129 million of acquisitions net of cash acquired, following the MEXT purchase
- Q3 2026 guide: revenue of approximately $13 billion plus or minus $300 million, up roughly 41% year over year and 13% sequentially, above consensus near $12.6 billion; non-GAAP gross margin of approximately 56%; non-GAAP operating expenses of approximately $3.65 billion
The News
AMD reported second quarter 2026 revenue of $11.5 billion and non-GAAP diluted earnings per share of $1.66. The company’s Data Center revenue more than doubled year over year to $6.7 billion and accounted for 58% of the company quarterly total. The print landed ahead of consensus on revenue, earnings and the data center line, and the third quarter guide of approximately $13 billion sits above where the sell side had modeled it. Market attention seems to be focused on a gross margin outlook held flat sequentially as the Helios rack scale ramp begins. The full release is available from AMD Investor Relations.
Analyst Take
A record print, a raised guide, and a lower stock. That combination usually says more about positioning than about operations, and we suspect it does here. AMD delivered its strongest data center quarter to date and beat on every headline line, yet shares fell roughly 8% in extended trading. We spent time with AMD at its Advancing AI event a few weeks ago, where Cerebras and Cisco appeared on stage alongside Lisa Su and Jack Huynh, and the through line of that event was partnership breadth rather than silicon alone. This quarter is the first test of monetizing it. Our contrarian read is that the market appears to have repriced Helios as a margin event rather than a revenue event, and holding the third quarter non-GAAP gross margin guide at approximately 56% gave that interpretation room to breathe. Bulls will counter that flat is not down and that the opening quarter of a complex ramp is the most expensive one. That is a case on direction that will only be proven with timing on shipments and installations.
What The Numbers Mean
The transcript color matters more than the spreadsheets here, because the beat was carried by a business most models still treat as the secondary story. Management described server demand as largely unforecasted going into the year, which is why the first half ran tight, and pointed to work across wafers, back-end capacity, substrates and components to lift capacity through the year. Su noted that cloud and enterprise server sales each grew more than 70% year over year, with growth in both units and average selling prices and units contributing the larger share. That mix detail is the quiet quality signal, since ASP gains tied to higher core counts tend to persist while unit surges can reflect pull-forward.
Two disclosures struck us as underappreciated. First, management framed the chiplet approach as reducing the number of wafers required on a new node, which functions as a structural hedge against TSMC two nanometer tightness rather than a marketing point. Second, on memory, Su indicated visibility into HBM allocation for 2027 and, more interesting, a willingness to modify the memory footprint on workloads where the total cost of ownership benefit does not justify the content. The MEXT acquisition, which closed in the quarter and adds predictive memory optimization, sits directly against that pressure. That small deal has an outsized punch in this memory market.
Hu was careful on 2027 margin, describing puts and takes: server mix accretive, data center AI running slightly below the corporate average as it ramps, embedded providing a tailwind. Su separately clarified that Helios tracking ahead of forecast referred to demand volume rather than yields. On the client side, where management hedged on a softer second half, we see the Cisco collaboration pairing Ryzen AI Halo with Cisco networking, observability and security as the least covered part of the story. That relationship connects the PC franchise to a managed infrastructure motion rather than a unit cycle. Working capital is the ramp's shadow, and it lengthened this quarter.
Market Analysis
The same-quarter peer set suggests the constraint on AI compute remains physical rather than commercial. In this reporting cycle, TSMC reported revenue growth of 33.7% year over year with gross margin near seven-tenths of revenue, and guided the September quarter materially higher on a steep two nanometer ramp while lifting capital intensity. ASML came in above its own guidance and outlined meaningful capacity expansion for 2027, which is not the behavior of a WFE supplier anticipating a slow in demand for bu8iulding new chip generations.
Intel posted its fastest revenue growth in more than a decade, with Data Center and AI up 59% year over year and management describing demand as running ahead of what it can currently supply. We would resist reading the two x86 prints as a simple market share transfer. Both server franchises grew strongly in the same quarter, both cited supply as the limiter, and both flagged substrates and memory as gating items. That pattern looks more like a market expanding faster than either supplier planned for, which is a healthier condition for AMD than a pure displacement narrative would be.
The memory prints complete the picture. SK hynix and Samsung both delivered record semiconductor results on sharply higher DRAM pricing and HBM allocation, with Samsung's handset division suffering the cost on the other side of the house. That is the same input inflation AMD cited when guiding for a softer second half PC market, and it explains why the client commentary was hedged while the server commentary was not.
One structural question sits outside the prints. The Cerebras collaboration, pairing Helios with the Wafer-Scale Engine, is expected to reach Cerebras Cloud in the fourth quarter. That will be the early test of whether ultra-low-latency inference becomes a distinct product category rather than a configuration of an existing one. NVIDIA reports its July quarter on August 26, which should help indicate whether accelerator demand is broadening across suppliers or concentrating.
Looking Ahead
The key trend we'll be monitoring is whether the data center AI ramp proves dilutive to gross margin for one quarter, two, or through 2027. Management has guided the third quarter as the beginning of Helios shipments with a step up in the fourth quarter, which makes the December period the first in which rack scale volume and cost structure can be assessed together rather than inferred. Su held the server CPU line at more than 70% for 2027, calling it a strong statement given where the business already sits, while twice upgrading the data center language when analysts pressed on how much work the word more was doing in that guide. We read the asymmetry as deliberate. Restraint on the business that is shipping now, expansion on the one that is not. The offsetting variable is the embedded business, that company management now describes as an x86 growth driver in networking and control plane workloads, with design wins tracking above $18 billion for the year. That book suggests the segment recovery carries a multi-year tail rather than a single favorable comparison, and it is the least cyclical support under the 2027 margin case.
Stephen Sopko | Analyst-in-Residence – Semiconductors & Deep Tech
Stephen Sopko is an Analyst-in-Residence specializing in semiconductors and the deep technologies powering today’s innovation ecosystem. With decades of executive experience spanning Fortune 100, government, and startups, he provides actionable insights by connecting market trends and cutting-edge technologies to business outcomes.
Stephen’s expertise in analyzing the entire buyer’s journey, from technology acquisition to implementation, was refined during his tenure as co-founder and COO of Palisade Compliance, where he helped Fortune 500 clients optimize technology investments. His ability to identify opportunities at the intersection of semiconductors, emerging technologies, and enterprise needs makes him a sought-after advisor to stakeholders navigating complex decisions.