- Multi-year agreement between AMD and OpenAI to deploy up to 6 GW of GPUs
- Start with 1 GW based on AMD Instinct MI450 from the second half of 2026
- OpenAI Option to Purchase Up to 160 Million AMD Shares in Segments
- Expected financial impact: tens of billions in revenue and co-design at the rack level
Collaboration between AMD and OpenAI takes a leap forward with a strategic agreement to deploy a computing infrastructure dedicated to generative artificial intelligence on a scale unusual in the industry. The plan provides for up to 6 gigawatts of power in AMD graphics accelerators, with a first milestone of 1 gigawatt which will be launched in the second half of 2026.
The pact is multi-year and multi-generational, and integrates hardware and AMD software as a key supplier on OpenAI's roadmap. The relationship is strengthened following the series MI300X y MI350X, and materializes with the leap to the new generation Instinct MI450, designed for training loads and deployment of large models.
Keys to the agreement and roadmap
As communicated by both companies, the alliance guarantees a scalable supply of High-performance GPUs to support projects IA of great magnitude. The first phase will begin with the MI450 and an installation design that seeks to maximize performance per watt and efficiency of the data center.
The plan's horizon includes a progressive deployment up to 6 GW, subject to technical, commercial and financial milestones. This implies a progression by sections where each advancement triggers new phases of the agreement, aligned with technology maturity and operational expansion.
Co-design and optimization at the rack level
AMD and OpenAI will work side by side on the design and optimization of rack-level solutions, with a focus on scalability and orchestration of large AI clusters. This exchange of technical knowledge It aims to directly influence future generations of products, from the base software to the interconnection between nodes.
In the absence of complete technical data sheets of the MI450, companies point to expected improvements in Energy Efficiency , performance per watt and interconnection capacity, crucial aspects for distributed training and for the deployment of increasingly demanding models in terms of memory and bandwidth.
Financial conditions and capital structure
The agreement includes an option whereby OpenAI You will be able to purchase up to 160 million AMD common shares at a symbolic price of 0,01 dollars per share, structured in sections linked to technical and commercial milestones. The first consolidation will be activated with the implementation of the first gigawatt scheduled for 2026.
The remaining sections will be unlocked as the deployment is completed until 6 GW and objectives associated with it are met share price of AMD, with increasing thresholds reaching up to $600 for the final stretch. Company executives anticipate tens of billions in revenue linked to the agreement and a positive impact on the non-GAAP earnings per share.
Expected impact on AMD's business
The operation places AMD at the forefront of specialized computing for AI, reinforcing its positioning against a increasingly concentrated marketCompany executives have described the alliance as a movement of transformative character which could attract more customers to AMD-branded solutions, expanding the reach of the ecosystem.
Following the announcement, investor interest was noted with sharp increases in pre-opening, reflecting the weight that the GPU access and the ability to scale have today in the valuation of the sector. AMD sees in this agreement a path to sustained growth in the medium and long term thanks to contracts that ensure demand and encourage joint innovation.
What OpenAI is pursuing with this alliance
For OpenAI, the goal is to ensure computing power sufficient for its roadmap in advanced models and associated services. The collaboration with AMD provides a path to diversification of the supply of accelerators and one more piece in the puzzle of its global AI infrastructure.
The planned deployment is of enormous magnitude: speaking of 6 gigawatts equivalent to infrastructure comparable to large industrial facilities. As a guide, 1 gigawatt can support a data center with tens of thousands of GPUs operating in parallel, a necessary condition for mass training and for services with increasing demand.
MI450 Calendar and Technology
The first milestone set is the Boot of the initial gigawatt during the second half of 2026, with systems based on the series Instinct MI450. Although not all details have been revealed, a jump in performance per watt and in the efficiency of the interconnection fabric to reduce bottlenecks in distributed work.
Nature multigenerational The agreement will allow OpenAI to incorporate successive AMD enhancements as they are ready, with an upgrade path that seeks to keep the infrastructure in the performance frontier without breaking the operational continuity of the clusters.
With this alliance on the table, AMD strengthens its role as strategic partner benchmark for high-level AI workloads, while OpenAI gains room to plan for growth with a scalable supply and a technical collaboration that aims to squeeze every new iteration of hardware and software.
Passionate writer about the world of bytes and technology in general. I love sharing my knowledge through writing, and that's what I'll do on this blog, show you all the most interesting things about gadgets, software, hardware, tech trends, and more. My goal is to help you navigate the digital world in a simple and entertaining way.