Q2 FY2026 Earnings Recap

Advanced Micro Devices, Inc. (AMD)

Reported Aug 4th 2026

News Summary

  • AMD’s Q2 2026 results were a record quarter with revenue of $11.536 billion, GAAP diluted EPS of $1.38, and non-GAAP diluted EPS of $1.66; revenue and adjusted EPS both beat consensus, while the release also showed strong gross-margin expansion and sharply higher operating income.

Financial Highlights

MetricQ3 '24Q4 '24Q1 '25Q2 '25Q3 '25Q4 '25Q1 '26Q2 '26
Revenue$6.8B$7.7B$7.4B$7.7B$9.2B$10.3B$10.3B$11.5B
Rev Growth (YoY)17.6%24.2%35.9%31.7%35.6%34.1%37.8%▲ 50.1%
Gross Margin50.1%50.7%50.2%39.8%51.7%54.3%52.8%▲ 53.8%
Operating Margin10.6%11.4%10.8%-1.7%13.7%17.1%14.4%▲ 17.3%
EV/Sales38.8x26.3x22.3x30.3x27.9x34.0x32.0x73.7x

Earnings Call Highlights

AMD's Q2 2026 earnings call was a high-conviction, bullish affair with management projecting extraordinary confidence in the AI infrastructure buildout, raising virtually every long-term target and drawing enthusiastic if probing analyst questions. Lisa Su's posture was assertive and expansive, repeatedly signaling that consensus estimates for both GPU and CPU revenue in 2027 are too low, and that the data center AI accelerator TAM has grown to $1.4 trillion by 2030 at a 45% CAGR. The dominant theme was the simultaneous acceleration of both the Instinct GPU ramp (Helios) and the EPYC server CPU business, with agentic AI cited as a new structural demand driver for CPUs. Analysts were clearly trying to size the 2027 Instinct revenue opportunity and were nudged upward by management, while gross margin questions were handled with measured optimism. The overall tone was one of a company that believes it has crossed a credibility threshold with hyperscalers and frontier AI labs and is now in a supply-constrained, demand-rich environment.

  • AMD announced a new strategic partnership with Anthropic for up to 2 gigawatts of MI450 Series GPUs in Helios, with the first gigawatt deploying in H1 2027, adding to existing gigawatt-scale commitments from OpenAI and Meta, and a new Helios-at-scale deal with Microsoft on Azure for frontier model inferencing.
  • Lisa Su explicitly told analysts that their 2027 Instinct revenue estimates are 'probably too low,' characterizing the data center AI segment as growing 'well over 100%' year-over-year in 2027, and nudging the implied Instinct figure above the $30 billion level one analyst floated.
  • AMD raised its data center AI accelerator TAM forecast to approximately $1.4 trillion by 2030 at a greater than 45% CAGR, up materially from prior estimates, and now expects the server CPU market to reach $220 billion by 2030 at greater than 50% CAGR, with agentic AI sandboxes cited as the fastest-growing and ultimately largest CPU sub-segment.
  • The Cerebras partnership was directly called out by a Citi analyst: AMD confirmed that the Helios plus Cerebras wafer-scale engine combination for fast inference will begin availability in Q4 in the Cerebras cloud and extend into 2027, framing it as part of AMD's open-ecosystem approach to workload-specific optimization.
  • ROCm reached what Su called 'an important inflection point,' with more than 3 million models running out of the box on AMD, open-source contributions up more than tenfold year-over-year, and the launch of ROCm.ai, an AI-assisted development platform supporting Claude, Codex, and Cursor for porting and optimizing code to Instinct GPUs.

Sources

Generated by Earnings Bot on 2026-08-13 07:46 UTC.