A designer of computer chips behind the Ryzen processors in many PCs, the Radeon graphics cards gamers use, and the semi-custom chips powering the PlayStation and Xbox game consoles. The company also makes EPYC server chips and Instinct AI accelerators that run giant cloud data centers. Founded in 1969 by Jerry Sanders and seven former Fairchild Semiconductor colleagues, it got its start making chips designed by other firms, and its flamboyant founder was nicknamed the "rock star of the semiconductor industry."
Data Center revenue more than doubled to $6.7B, driving AMD's Q2 revenue up 50% to $11.5B.
Data Center more than doubled, turning the into a $2.1B profit engine. Revenue rose 50% to $11.5B and expanded 14 points to 53.8% as the prior year's $800M export-control charges cleared and the product mix shifted toward high-end GPUs and CPUs. The business is scaling fast, but new multi-billion-dollar investment commitments and data center leases raise the stakes on sustained AI demand.
Key takeaways
Data Center rose to $6.7B from $3.2B a year earlier on demand for AMD Instinct MI350 GPUs and processors, swinging the to a $2.1B from a loss in the prior-year quarter.
expanded to 53.8% from 39.8% a year ago, primarily because the approximately $800M in and related charges from U.S. export controls on MI308 products did not repeat.
Client rose 23% to $3.1B on a 34% increase in unit shipments, while Gaming revenue fell 31% to $779M on lower semi-custom sales.
Section summaries
Management's Discussion and Analysis
Revenue surged 50% to $11.5B driven by Data Center GPU/CPU demand, with gross margin expanding to 54% from 40%.
⌄
Data Center more than doubled to $6.7B, fueled by processors and AMD Instinct MI350 Series GPUs, swinging the to a $2.1B .
reached $1,990M, up from a $134M loss a year earlier, as higher outweighed a 33% rise in R&D expense and a 41% rise in marketing, general and administrative expense tied to AI headcount and go-to-market activities.
rose 163% to $2,297M and rose 156% to $1.38, while for the first half reached $5.3B.
After the quarter, the company entered investment commitments of up to $5.0B and long-term data center leases with $9.5B in aggregate future payments.
What changed
The MI308 export-control charges that pulled Q2 FY2025 to 39.8% did not recur, allowing gross margin to recover to 53.8%.
Data Center growth accelerated to more than 100% , up from 57% in Q1 FY2026, as the MI350 GPU ramp gained pace.
The manufacturing-business sale, previously expected near end-2025 and then end-2026, was not flagged with a new closing date this quarter.
Embedded was not reported separately for a second consecutive quarter, leaving the trajectory of the prior 3% annual decline unresolved.
What to watch
Data Center in Q3 FY2026 to see if the triple-digit growth pace holds as the MI350 ramp continues and MI325 export license conditions take effect.
Closing of the $3.0B manufacturing-business sale and any update on the timeline or divestiture risk.
in Q3 FY2026 to confirm the 53.8% level is sustainable as product mix evolves and export-control charges stay out of results.
The $5.0B in new investment commitments and $9.5B in data center lease obligations, and whether they signal a step-change in capital intensity.
Client rose 23% to $3.1B on a 34% jump in unit shipments, while Gaming revenue fell 31% to $779M due to lower semi-custom sales.
improved 14 percentage points to 54%, primarily due to the absence of prior-year charges tied to export controls on GPUs and a favorable product mix.
Operating expenses grew significantly, with R&D up 33% to $2.5B for AI-focused headcount and MG&A up 41% to $1.4B for go-to-market activities.
was $5.3B for the first half, and total liquidity stood at $13.1B in cash and short-term investments against $3.3B in debt.
Post-quarter, the company entered investment commitments of up to $5.0B and long-term data center leases with $9.5B in aggregate future payments.
Quantitative and Qualitative Disclosures About Market Risk
Reference is made to “Part II, Item 7A, Quantitative and Qualitative Disclosures About Market Risk,” in our Annual Report on Form 10-K for the fiscal year ended December 27, 2025. Long-term Investment Risk. We hold long-term investments in marketable equity securities of publicl…
⌄
Reference is made to “Part II, Item 7A, Quantitative and Qualitative Disclosures About Market Risk,” in our Annual Report on Form 10-K for the fiscal year ended December 27, 2025.
Long-term Investment Risk. We hold long-term investments in marketable equity securities of publicly traded companies and non-marketable equity and other securities of privately held companies. As of June 27, 2026 and December 27, 2025, the carrying value of marketable equity securities was $1.2 billion and $0.2 billion, respectively, and non-marketable equity and other securities was $1.7 billion and $1.1 billion, respectively. These investments are subject to market risks that could substantially impact their fair value. The Company regularly reviews non-marketable securities for impairment.
There have not been any material changes in interest rate risk, default risk or foreign exchange risk since December 27, 2025.
For a discussion of our legal proceedings, refer to Note 10—Commitments and Contingencies of the Notes to Condensed Consolidated Financial Statements (Part I, Item 1 of this Form 10-Q).
⌄
For a discussion of our legal proceedings, refer to Note 10—Commitments and Contingencies of the Notes to Condensed Consolidated Financial Statements (Part I, Item 1 of this Form 10-Q).
Export controls on AI chips to China, competitive pressure from Nvidia and Intel, and supply chain dependencies are the most material risks.
⌄
U.S. export restrictions on AI accelerators (MI308, MI325) to China have caused $440M in charges, require licenses with uncertain conditions, and may include a 15% -sharing demand from the U.S. government.
Intense competition from Nvidia and Intel, including Nvidia's September 2025 partnership with Intel, threatens market share, pricing, and margins across data center, client, and gaming segments.
Heavy reliance on TSMC for leading-edge wafer fabrication and on joint ventures for assembly and testing creates significant supply concentration risk with no guaranteed capacity.
An industry-wide memory shortage and dependence on sole or limited suppliers for substrates and other components could increase costs and disrupt production of complex rack-scale systems like Helios.
The success of the Data Center hinges on sustained, but uncertain, customer demand for AI accelerators and their ability to secure sufficient data center capacity, energy, and financing.
New or heightened risks include the financial and operational impact of specific U.S. export license conditions for China-bound AI products and the competitive threat from the Nvidia-Intel partnership.