Advanced Micro Devices (NASDAQ:AMD) executives used a short-notice investor call to outline a “significant expansion” of the company’s strategic partnership with Meta, centered on what AMD described as a new multi-year, multi-generation agreement for large-scale AI infrastructure deployments.
AMD outlines 6GW Instinct GPU deployment plan with Meta
CEO Dr. Lisa Su said the agreement positions AMD “at the core” of Meta’s next-generation AI infrastructure. Under the deal, Meta is expected to deploy 6 gigawatts of AMD Instinct GPUs across multiple product generations. Su said Meta has been a close partner over multiple generations, deploying “millions” of EPYC CPUs and “hundreds of megawatts” of MI300 and MI350 series GPUs across its global infrastructure.
Custom MI450-based accelerator and Helios rack-scale architecture
A key element of the expanded relationship is co-engineering a custom GPU accelerator based on AMD’s MI450 architecture, optimized for Meta’s workloads. Su characterized the project as AMD’s first custom GPU effort of this style using an MI450-based architecture.
Initial shipments supporting what AMD called the first gigawatt deployment are scheduled to begin in the second half of 2026. Su said the deployment will leverage AMD’s Helios rack-scale architecture, the custom MI450-based Instinct GPU, and AMD’s 6th Gen EPYC CPU, code-named Venice.
In response to analyst questions about the nature of the customization, Su said the effort started “with the workload first” rather than the chip, and that the optimization spans “chip level, board level, [and] system level.” CFO Jean Hu added that no additional tape-out is needed for the customer chip, and Su emphasized the approach is not a full ASIC. When asked about software transferability, Su said the software work is “extremely leverageable,” suggesting “+95%” of the effort should carry over because it is based on the MI450 architecture and remains within a programmable GPU framework.
EPYC CPU relationship expands alongside GPUs
Su also highlighted expanding CPU collaboration, saying AMD is seeing accelerating CPU compute demand as AI infrastructure scales across model development, inferencing, and data processing, including the rise of “agentic AI.” She said EPYC processors power the majority of Meta’s core services across its global data center footprint.
According to Su, Meta will be a lead customer for AMD’s 6th Gen EPYC Venice processor at launch later this year. She also referenced a new addition to the Zen 6 family, code-named Verano, which she said incorporates workload-specific optimizations aimed at performance per watt and total cost of ownership.
Financial framework: revenue timing, warrant structure, and long-term targets
Hu said the 6-gigawatt agreement is consistent with AMD’s long-term scaling plan for its data center AI business. She stated that the Meta deployment is expected to generate data center AI revenue of “significant double-digit billions of dollars per gigawatt,” with revenue beginning in the second half of 2026 and ramping alongside MI450 deployments with other customers.
As part of the agreement, Hu said AMD issued Meta a performance-based warrant for up to 160 million shares of AMD common stock. She described the structure as follows:
- The first tranche vests with the initial 1 gigawatt of shipments of AMD Instinct GPUs.
- Additional tranches vest as Meta’s purchases scale to 6GW.
- Vesting is also tied to AMD achieving certain stock price thresholds, with the final tranche vesting at $600 per share.
- Warrant exercise is tied to Meta achieving key technical and commercial milestones.
Hu said the structure is designed to align incentives and is “accretive to our non-GAAP earnings per share.” She also reiterated AMD’s long-term financial model, including “greater than 80% CAGR” for the data center AI business and a goal of generating “more than $20 in annual earnings per share” within the next three to five years.
On a later question, Su said AMD’s previously stated $20+ EPS target did not specifically include the Meta deal when it was outlined in November, noting the agreement was still “very, very much in the works” at that time. She added that AMD’s long-term model assumes a broad set of customers and ongoing strategic engagements.
Analyst Q&A: overlap with OpenAI, deal precedent, and commitment visibility
During Q&A, Su said there is no overlap between the Meta agreement and AMD’s previously announced OpenAI deal, describing the Meta arrangement as specifically optimized for Meta’s workloads.
Analysts also pressed on whether the warrant structure could set a precedent for future large customer deals. Su responded that “every deal of this scale is very unique,” and argued the partnership is “transformational,” citing deployment scale, ecosystem maturation, and software maturity. She said AMD is seeking “defining partnerships” and that not every MI450 customer will be of Meta’s “size and scale.”
On visibility and commitments, Su said the agreement spans five years for 6GW, with the first gigawatt committed and shipments beginning in the second half of 2026. She added that supply chain conditions remain tight and that AMD and Meta are planning deployments closely around data center buildouts. Su also said AMD is working beyond MI450, noting development of “MI500 and beyond.”
About Advanced Micro Devices (NASDAQ:AMD)
Advanced Micro Devices, Inc (NASDAQ: AMD) is a global semiconductor company that designs and sells microprocessors, graphics processors, chipsets and adaptive computing solutions for a broad set of markets. The company’s product portfolio includes consumer and commercial CPUs under the Ryzen and Threadripper brands, data center processors under the EPYC brand, and Radeon graphics processing units for gaming and professional visualization. AMD also offers semi-custom system-on-chip (SoC) products for gaming consoles and other specialized applications, and provides supporting software and platform technologies for OEMs, cloud service providers and end users.
Founded in 1969, AMD has evolved from a supplier of logic chips into a diversified, fabless semiconductor designer.
