AMD revenue rises 50% as AI data-center sales more than double

AMD reported its second-quarter results on August 4, and the numbers show how strongly AI infrastructure is reshaping the chip market. Revenue reached $11.5 billion, up 50 percent from a year earlier. The company also reported a 54 percent GAAP gross margin and $2.3 billion in net income. AMD chief executive Lisa Su said the quarter set records for revenue and profitability.

The central figure was the Data Center segment. It generated $6.7 billion, an increase of 107 percent year over year, and represented 58 percent of AMD’s total revenue. The growth came from demand for EPYC server processors and Instinct accelerators. AMD expects that business to accelerate again in the second half of 2026, forecasting roughly $13 billion in total revenue for the third quarter.

The results release was also a product and partnership update. AMD says its Helios rack-scale system is beginning to ramp and is being deployed by AI labs and cloud providers including Anthropic, Microsoft and OpenAI. The company highlighted the Instinct MI400 GPU family, new sixth-generation EPYC CPUs and ROCm.ai, an updated developer experience for building and optimising AI workloads on AMD hardware.

For the AI market, this matters because the competition is moving beyond model quality. Training and serving advanced systems requires a large supply of accelerators, CPUs, networking, memory and software. AMD is trying to offer a complete alternative platform, while its partnerships show that major AI companies want more than one source of compute. Anthropic has separately announced plans to deploy up to two gigawatts of AMD Instinct MI450 GPUs in Helios racks, according to the AMD release.

There is still a difference between an announcement and a working fleet. AMD’s performance, deployment and forward-looking claims come from the company, so buyers will need independent tests of cost, availability, energy use and software compatibility. The headline results nevertheless strengthen AMD’s position in a market long associated mainly with Nvidia.

For everyday AI users, nothing changes in a chatbot today. Over time, stronger competition for the hardware underneath those services could improve capacity, pricing and resilience. For developers and businesses, the practical signal is to evaluate the whole stack: model performance is only one part of the cost and reliability equation. The future AI platform may be determined as much by where workloads run as by which model writes the answer.