AMD used its Advancing AI 2026 event to move artificial intelligence closer to developers and enterprise users, positioning Ryzen AI Halo as a compact platform for building and running powerful models locally.
TL;DR
- Ryzen AI Halo combines 128 GB of unified memory with support for models containing up to 200 billion parameters.
- It supports Windows and Linux, includes AMD’s ROCm software stack and offers up to 50 TOPS of NPU performance.
- AMD and Cisco are developing tools to secure, govern and monitor fleets of local AI systems.
- More Ryzen AI Halo systems are expected later in 2026.
AMD Makes Local AI More Accessible
At Advancing AI 2026, AMD presented Ryzen AI Halo as a one-stop platform for local AI development and inference, designed to reduce the configuration work required before developers begin experimenting.
The compact system is powered by the Ryzen AI Max+ 395 processor and includes 128 GB of LPDDR5x unified memory. AMD says it can run models containing as many as 200 billion parameters, giving developers room to test large language models, coding assistants, image generators and agent workflows without relying entirely on cloud infrastructure.
It also provides 60 FP16 TFLOPS of GPU performance and up to 50 TOPS through its neural processing unit. The system supports Windows and Linux, while AMD’s ROCm software stack supplies tools for developing and deploying AI workloads.
AMD is pitching local execution as a way to deliver greater workflow control and predictable costs. It claims sustained local inference can cost substantially less over three years than equivalent cloud API usage, although the comparison is based on AMD’s own workload, electricity and pricing assumptions.
Ryzen AI Halo Reaches Developers
Ryzen AI Halo began shipping through Micro Center in July, with the retailer serving as AMD’s first global launch partner. AMD used a retail price of $3,999 for the 128 GB system in its internal cost and performance comparisons.
More Ryzen AI Halo platforms powered by Ryzen AI Max PRO 400 Series processors are expected from AMD and its hardware partners later in 2026.
“The future of AI won't be defined by any single model. It will be defined by what you create,” said Jack Huynh, senior vice president and general manager of AMD’s Computing and Graphics Group.
AMD is also preparing a version powered by the Ryzen AI Max+ PRO 495 processor, with support for up to 192 GB of memory.
Topics For More Insights
Cisco Adds Enterprise Control
Powerful hardware alone is not enough for large businesses, where fleets of AI agents require consistent security policies, monitoring and network controls.
AMD and Cisco are collaborating to combine Ryzen AI Halo with Cisco networking, observability and security products. The goal is to help companies deploy and govern AI agents operating locally, across hybrid environments or alongside cloud services.
Cisco described the shift as an opportunity to “run serious AI where their people work, not just in a distant cloud.”
The planned architecture combines Splunk monitoring, Cisco AI Defense, DefenseClaw policy enforcement and Cisco Cloud Control. IT teams could track agent activity, compute utilization and token costs, restrict risky actions and quarantine devices following a serious model or agent failure.
Why AMD’s Personal AI Push Matters
The announcement reflects a broader shift in AI computing. Rather than sending every prompt or workflow to a remote data center, developers and enterprises can process more information near the user, potentially improving data control, responsiveness and cost management.
AMD faces strong competition across the AI hardware market. However, Ryzen AI Halo shows that its strategy is not limited to massive server racks such as Helios, which AMD expects to begin shipping by the end of the third quarter of 2026.
AMD also wants capable AI development and inference systems on desks, giving more users control over where their models and data run.
