AT&T Achieves Outstanding Efficiency for AI Training with AMD
AT&T trains telecom AI on AMD Instinct™ MI355X GPUs, running a full model on a single GPU to power inference for millions of customers.
AI Leaders Are Choosing AMD Instinct™ GPUs
Up to 6 GW AI infrastructure commitment
Up to 6 GW AI infrastructure commitment
Up to 2 GW AI infrastructure commitment
Microsoft brings the AMD Helios™ AI platform to Azure in a growing partnership with AMD.
Oracle set to deploy 50,000 AMD Instinct GPUs in 2026, then expanding in 2027 and beyond.
AMD Instinct will power the next gen Sovereign AI deployment in North America with Discovery at ORNL
AMD Instinct GPUs power the largest and fastest US exascale supercomputer, El Capitan.1
Trusted by AI Leaders
Building For the Next Era of AI Right Now
As AI matures, organizations need a trusted infrastructure for critical AI workloads. Leaders choose Al infrastructure powered by AMD Instinct GPUs for training and inference performance, reliability, scalability, security and open-software ecosystem.
Person using a tablet in a modern data center aisle between server racks
Hyperscale
Meta, OpenAI, and Anthropic have announced multi-gigawatt deployments of AI infrastructure powered by multiple generations of AMD Instinct GPU, starting in 2026..
Person using a tablet in a modern data center aisle between server racks
Cloud
Oracle Cloud Infrastructure and Microsoft Azure are deploying AMD Instinct at scale, with neocloud providers including TensorWave, Vultr, and Crusoe also adopting AMD Instinct powered AI infrastructure.
Person using a tablet in a modern data center aisle between server racks
Sovereign AI
Lux and Discovery at Oak Ridge National Laboratory, alongside France’s Alice Recoque program led by GENCI leverage AMD Instinct for sovereign AI.
Person using a tablet in a modern data center aisle between server racks
Person using a tablet in a modern data center aisle between server racks
Person using a tablet in a modern data center aisle between server racks
Person using a tablet in a modern data center aisle between server racks
Enterprise AI
AMD Instinct solutions integrate into existing infrastructure, helping enterprise customers quickly scale generative and agentic AI to run more models and serve more users without an infrastructure overhaul.
Up To
30%
Better performance-per-dollar than competitive solutions²
Leadership Performance
AI Infrastructure at Any Scale
Infrastructure must scale to meet any need, from AI startup labs to enterprise data centers, cloud services, hyperscale AI factories, and national labs.
AMD Instinct GPUs are the centerpiece of a broad portfolio that scales AI capacity as you grow - across GPUs, CPUs, networking, software, racks, and clusters—supporting inference, training, reasoning, and HPC at the scale their workloads demand.
At any scale, performance efficiency matters. That is why AMD Instinct GPUs are designed to optimize performance per dollar. The AMD Helios Rackscale Solution provides up to 30% better performance-per-dollar than competitive solutions.2
Open Infrastructure
Choice Without Complexity
AMD makes open innovations easier to adopt through deep integrations, validation, optimization, and ecosystem collaboration. With AMD ROCm™ software, teams can use an open, upstream-first software stack to build, deploy, and scale AI and HPC workloads with confidence on AMD Instinct™ GPUs.
Open Software
Build on AMD ROCm™
Use an optimized open software stack to deploy high-performance AI on AMD Instinct GPUs.
Familiar Frameworks
Use The Tools You Know
Develop with the most popular tools like PyTorch, TensorFlow, JAX, ONNX Runtime, vLLM, and SGLang.
Day-0 Readiness
Build and Deploy Faster
Validated builds, containers, and repeatable workflows reduce onboarding friction.
Operations
Manage AI at Scale
AMD ROCm software stack provides Kubernetes-ready deployment, observability, and lifecycle management across large AI clusters.
Hardware
Choose Open Standards
AMD advances open, standards-based AI infrastructure through OCP, UALink™, UEC and ESUN.
Ecosystem
Accelerating Innovation Together
AMD brings together software partners, ISVs, OEMs, ODMs, and neocloud providers to deliver scalable performance and greater customer value.
Success Stories
See How Customers Are Succeeding with AMD Instinct GPUs
A Broad Portfolio
The AMD Instinct Family
Resources
Learn more about AMD Instinct GPUs, AMD ROCm, and AMD rackscale solutions.
Discover AMD Instinct
The biggest names in AI have chosen AMD Instinct as the foundation for their next phase of AI development. See what makes AMD Instinct such a compelling choice for AI infrastructure at any scale.
Footnotes
- https://top500.org/lists/top500/list/2026/06/
- MI400-025: Based on AMD Performance Labs estimates as of July 2026, tokens-per-dollar performance was calculated using the Kimi K2 Thinking workload (32K input / 8K output) on an AMD Helios rackscale solution compared to an NVIDIA Vera Rubin NVL72 rack. Results reflect estimated aggregate throughput across low, medium, and high-interactivity operating points and hourly pricing projection of system GPUs based on market conditions. System configurations may vary by manufacturer and may produce different results.
Footnotes
- https://top500.org/lists/top500/list/2026/06/
- MI400-025: Based on AMD Performance Labs estimates as of July 2026, tokens-per-dollar performance was calculated using the Kimi K2 Thinking workload (32K input / 8K output) on an AMD Helios rackscale solution compared to an NVIDIA Vera Rubin NVL72 rack. Results reflect estimated aggregate throughput across low, medium, and high-interactivity operating points and hourly pricing projection of system GPUs based on market conditions. System configurations may vary by manufacturer and may produce different results.