# "GPU demand" — X 热门讨论 (2026-09-28 21:19 UTC)
## @IvanaSpear (IvanaSPEAR) · 09-28 16:14 · ♥37 ↻6 💬2 $META's Muse is scaling much faster than investors appreciate.
And the implications for AI hardware demand could be significant.
Meta launched Muse on September 8.
Within roughly a week, it had already crossed 500K users, 250K+ daily active users and 2M+ prompts.
Third-party estimates put U.S. mobile DAUs at ~642K.
Now put that against Meta’s installed base: 3.6 BILLION people use at least one Meta app every day.
That distribution advantage is enormous.
For context, rival agent Instinct reportedly has 100K+ total users, and is already facing compute-capacity constraints.
Muse is still very early. But if agentic AI adoption continues at anything close to this pace, the compute implications could be substantial.
We’re already seeing a spike in GPU demand, with prices as high as $24/GPU/hr for B300 vs. $9/GPU/hr just two months ago.
The bigger point: agents consume a lot more compute than traditional chatbots. They browse, reason, call tools, run multi-step workflows and can keep working in the background.
Scale that across Meta’s distribution and the hardware demand curve could look very different.
Very positive for the AI hardware stack, particularly areas with outsized exposure to agentic inference.
We’ll go deeper on this in our webinar tomorrow.
Link in comments to sign up.
#Meta #Muse #AgenticAI https://x.com/IvanaSpear/status/2104605704308830710