Q3 revenue hits $13.4 billion with adjusted EPS of $2.09, beating market expectations
Company projects over 80% revenue growth by 2027, citing expanding AI semiconductor supply and memory cost pressures as key variables
U.S. semiconductor giant AMD posted record-breaking third-quarter results, propelled by accelerating demand for artificial intelligence infrastructure. With data center revenue surging 97% year-over-year to drive overall growth, the company projected fourth-quarter revenue of approximately $17 billion. While forecasting that total revenue will expand by more than 80% by 2027, CEO Lisa Su cautioned that AI chip demand continues to exceed secured supply capacity.
According to Q3 2026 earnings released by AMD on the 3rd (local time), revenue rose 45% year-over-year to $13.377 billion, surpassing the consensus estimate of approximately $13.09 billion. Adjusted earnings per share (EPS) came in at $2.09, beating the projected $1.95.
The data center business spearheaded the overall performance. Revenue in the segment reached a record $8.546 billion, a 97% increase from the same period last year. Bolstered by combined demand for AI accelerators and server processors, the data center unit accounted for roughly 64% of total revenue.
AMD forecasted fourth-quarter revenue at approximately $17 billion with a margin of ±3%, topping the market consensus of around $16.23 billion. Based on the midpoint of the guidance, this represents an increase of about 27% from the third quarter.
However, the projected adjusted gross margin for the fourth quarter stands at approximately 55%, down one percentage point from 56% in the third quarter. How product mix and production costs will impact profitability amid rising revenues remains a metric to watch in upcoming quarters.
CEO Lisa Su expressed confidence that this growth momentum will extend into next year. She projected that total revenue will surge by over 80% by 2027, with server CPU revenue roughly doubling and data center AI revenue expanding well beyond double.
At the center of this growth is the next-generation AI system, "Helios." AMD initiated initial shipments of Helios in the third quarter, noting that major firms—including Anthropic, Meta, Microsoft, OpenAI, and Oracle—are currently running their AI models and workloads on the system. The company plans to ramp up volume production in the fourth quarter to broaden deliveries to corporate clients.
Moving beyond standalone AI accelerator sales, AMD is expanding its business scope into rack-level systems that integrate multiple semiconductors and networking equipment. This strategy responds to a broader market shift where AI chip competition centers not only on individual processor performance, but also on total system processing capacity, power efficiency, and supply stability.
On the supply front, CEO Su explained that AMD has secured a substantial portion of the volume required for its 2027 growth trajectory. Working closely with key customers to align product deployment schedules and volumes, AMD has lined up core components such as wafers, substrates, advanced packaging, and memory. However, with demand continuing to outstrip supply, the company is coordinating with supply chain partners to further expand production capacity.
Cost pressures driven by rising memory prices have also emerged as a variable. Su noted that upward price pressures are manifesting across various components, including memory, adding that AMD is establishing long-term contracts with select suppliers and customers while discussing cost-sharing arrangements with clients. The company aims to maintain competitiveness beyond product pricing, factoring in total AI computation costs and customer total cost of ownership (TCO).
Expanding the supply of AI systems also requires robust production capacities in memory and advanced packaging. High Bandwidth Memory (HBM)—which stacks multiple DRAM dies vertically to increase data transmission bandwidth—relies on a base die at the bottom to handle input/output and control functions. In next-generation products, the logic process technology of the base die is emerging as a critical factor influencing performance and power efficiency. As the technological importance of HBM grows, securing stable volumes has become a core prerequisite for scaling up AI accelerator production.
To address these challenges, AMD is broadening its collaboration with memory suppliers and chip foundries. In March, the company agreed to strengthen cooperation on next-generation AI memory with Samsung Electronics, designating Samsung as a priority supplier of HBM4 for AMD's upcoming MI455X AI accelerator. Both parties also agreed to discuss leveraging Samsung's foundry services for future AMD products.
During a visit to South Korea last October, CEO Su reaffirmed that AMD continues to actively explore collaboration opportunities with Samsung across memory and foundry sectors, underscoring AMD's dual-track approach of reviewing both memory procurement and semiconductor manufacturing partnerships.
AMD's future performance hinges on the successful mass production and monetization of its next-generation AI systems. As the data center business maintains its upward trajectory, the extent to which expanding Helios shipments translates into actual revenue gains—and how effectively the company absorbs cost pressures from rising component prices—will dictate its financial scorecard through 2027.
