Image: AMD
Ask a personal AI agent what chip it runs on and, it turns out, many will simply tell you. That small question has turned into a friendly chip rivalry on X, with AMD and Intel processors showing up behind some of the best known agents.
Ryan Shrout put four agent apps to the test and shared the results in a single screenshot. His post sorted them neatly into two camps.
What CPUs are powering your favorite agents?
— Ryan Shrout (@ryanshrout) October 27, 2026
- EPYC: Meta Muse and OpenAI dots
- Xeon: Grok Bot and Claude Duet
What the agents said
Each app got the same plain question, what CPU are you running on. The answers were surprisingly specific.
The Muse agent, shown in the screenshot under the name Rex, reported 2 virtual CPUs of an AMD EPYC 9D25 126 core processor. It added a fair caveat, noting that the actual model inference happens on Meta's side and that the EPYC is what it is working on, not necessarily what it is thinking on.
The app labeled Flux reported an AMD EPYC 9V74 80 core processor with 9 logical CPUs in its cloud workspace. A third, unnamed app described a KVM virtual machine with 4 virtual CPUs of an Intel Xeon at 2.1GHz, which it identified as an Emerald Rapids part. The fourth, labeled GB, reported 8 cores of an Intel Xeon at 2.4GHz, a Sapphire Rapids class chip, with about 16GB of memory.
Image: Ryan Shrout on X
Shrout's tally put Meta Muse and OpenAI dots on EPYC and Grok Bot and Claude Duet on Xeon. Grok confirmed its side in a reply under the post, again describing an 8 core Xeon machine running as a sandbox for tools and code.
AMD takes a victory lap
AMD clearly enjoyed the moment. The company had already shared a short clip of someone asking Muse the same question, with the agent answering that it runs on an AMD EPYC 9D25 with 2 cores allocated to its virtual machine.
ICYMI: @Muse knows what's up. EPYC to see @Meta's success.
— AMD (@AMD) October 27, 2026
What Muse is
Muse is Meta's personal AI agent, which the company launched on September 8. Meta says Muse can send emails, book travel, fill out forms and work across a person's apps, asking for approval before sensitive steps such as a purchase. Each Muse runs on what Meta calls Muse Secure VM, a dedicated virtual machine in the cloud with its own browser, and it is powered by Meta's Muse Spark model. Reuters reported that the launch came despite internal concerns flagged in testing. Muse is free with usage limits, and Meta's help center lists paid plans at $20 and $100 a month.
Why the CPU matters, and why it does not
Every agent in the screenshot made the same point. The CPU it reported belongs to the machine that runs its tools, browser and code, not the accelerators that run the AI model itself. So the answer says nothing about which chips do the thinking.
It still matters to chipmakers. Agents that browse, run code and keep working in the background need a lot of ordinary server computing, and that is EPYC and Xeon territory. AMD's 5th Gen EPYC line, launched in 2024, scales up to 192 cores. One reply to Shrout summed up the trend, saying agents mostly wait on tools, so CPUs finally matter again.
For users, the practical difference is small at the time of this writing. Cloud providers can move agents between machines at any time, so today's answer could change tomorrow. Still, it is a fun question to ask your own agent.
Comments