AMD Data Center Revenue Hits $8.5B, Doubles Yearly Growth

AMD closed the third quarter of 2026 with revenues of $13.4 billion, a 45% increase compared to the same period in 2025, and a figure that shifts the company's perspective: the Data Center segment generated $8.5 billion, 97% higher than a year ago, according to the earnings report released on Tuesday (3) after market close. In practice, nearly two out of every three dollars that AMD sells today come from servers and accelerators for cloud and AI. Shares rose between 8% and 9% in after-hours trading.

The adjusted earnings per share stood at $2.09, compared to $1.46 on a GAAP basis. For the fourth quarter, the company projects revenue in the vicinity of $17 billion, with a margin of 3% up or down, according to guidance presented by Lisa Su and CFO Jean Hu during the conference call. If the midpoint is confirmed, AMD will have grown about 27% quarter-over-quarter, a rare pace for a company of this size.

MI450 Contributed to Revenue

The quarter marked the first time AMD recognized revenue from the Helios racks and the Instinct MI450 family, the company's bet to compete with Nvidia in the large-scale training and inference market. According to the announcement, Anthropic, Meta, Microsoft, OpenAI, and Oracle Cloud Infrastructure are already running models and workloads on the first Helios systems. The list is more significant than the volume: these are five of the largest buyers of AI capacity in the world testing an alternative to the industry's de facto standard.

Lisa Su, CEO of AMD, stated in the announcement that the company delivered a record quarter, with revenue exceeding the high end of its own forecast, and that growth is expected to accelerate in Q4 as production of the Helios and the EPYC "Venice" processors scales up. During the conference call, management described the transition from the third to the fourth quarter as a "very significant" jump in the volume of MI450.

The rest of the portfolio is moving at a different pace. The Client and Gaming division, which includes processors for PCs and console chips, shrank on a year-over-year basis, while the Embedded unit grew 29%, reaching $1.1 billion. The focus on the data center is the strength of the results and also the point that analysts will press: AMD now depends largely on the capital budgets of a handful of hyperscalers and AI labs.

Market Scrutiny Ahead

The results arrive at a time of skepticism. On the same Tuesday, the Nasdaq fluctuated as investors were divided between high valuations and doubts about the return on AI investments, and Palantir's shares fell even after surpassing estimates the previous day. The skeptical argument against AMD is not that demand is false, but that it is concentrated among a few clients whose capex decisions can shift from quarter to quarter.

There is also a distinction that the "AI bubble" debate often overlooks. Microsoft, Meta, and Oracle finance their chip purchases with cash flow from already profitable businesses. Labs like OpenAI and Anthropic rely on capital raising, and Anthropic filed its IPO request with Nasdaq on October 26. A data center revenue powered by the first group is more resilient than one supported by the second, and AMD's announcement does not differentiate between the two sources.

Nvidia is set to release its Q3 fiscal results on November 18, according to the earnings calendar.

Implications for Capacity Buyers

For CIOs and consultancies designing AI architectures, the relevant detail is the entry of a second source of complete racks in the same clouds. In the United States, this directly affects the capacity that Microsoft Azure and Oracle Cloud can allocate to corporate clients, as both appear on the list of Helios users.

Outside the American axis, the channel is the regional allocation. Banks and industries in Europe and Japan negotiate dedicated capacity with the same American clouds, and Oracle operates regions in São Paulo and Vinhedo that serve Brazilian clients. A second source of accelerators expands the total available volume, but AMD's announcement does not specify in which regions the racks will be installed. For clients outside the United States, the question to ask the provider is in which region and in which quarter this capacity will be available.

The takeaway for corporate buyers is timing: the guidance of $17 billion assumes that AMD will produce and deliver the Helios at the promised scale. Those signing capacity contracts for 2027 now have two potential suppliers behind the same cloud, and the hourly pricing for accelerators no longer has a single reference point.