Oct 21, 2026 earnings call — Muse personal agent rollout and AI strategy
More than 10 million people in the US and Canada have set up Muse personal agents in the first six weeks.
Revenue $64.6B (est $63.3B) · EPS $7.40 (est $6.54) · Next-quarter revenue outlook $73.0B–$76.0B
Key quotes
"In just a month and a half, more than 10 million people in the US and Canada have set up Muse personal agents,"
"Muse is coming to Meta Glasses starting next month, and that will be an important step towards this."
Analyst Q&A 7
limits to user growth (compute, models)
Q (Brian Nowak · Morgan Stanley): He asked what limits Muse's user growth now (capabilities, model efficiency, capacity) and asked Mark to describe Meta's enterprise product vision and value drivers.
A: Management said compute efficiency and capacity are key constraints now, and they need to make Muse more efficient to scale globally. They plan to develop better models (Muse Wonder and future models) and build enterprise offerings like the model API, Muse code, and business agents to drive value for small businesses and enterprises.
"We wanna make sure that we have built enough compute to be able to serve what we think is going to be a billion plus people." — Mark Elliot Zuckerberg (CEO)
agent commerce friction and capital plans
Q (Eric Sheridan · Goldman Sachs): He asked how Meta will reduce friction with businesses that resist agent interactions, and he asked about the new share class and capital plans.
A: Management said many businesses want to enable agent interactions and adoption will vary, but Meta is working on permissioned, structured integrations and open standards (personal agent protocol) with partners. On capital, they created a new nonvoting share class to expand funding tools while preserving voting power, but have no specific financing plans now.
"we created this new class of shares with the purpose of creating a publicly traded nonvoting stock usable for, you know, capital raising employee compensation acquisitions, other general corporate purposes" — Susan J. Li (CFO)
permissioned agent-to-agent interactions
Q (Douglas Anmuth · JPMorgan): He asked how the ecosystem will shift to more permissioned agent-to-agent transactions and what that unlocks, and he asked about 2027 expense and CapEx plans.
A: Management said the move to agent-to-agent integration will take time and be uneven, but proactivity from consumer and business agents can enable more transactions and new services like ongoing searches and notifications. On 2027 budget and capacity, they are still early, evaluating investments, and will provide more specifics at the Q4 call.
"it is clear that there are a lot of parts of the experience that would be even better if Muse was able to interact more seamlessly with business agents on the other side." — Susan J. Li (CFO)
AI impact on ads and revenue sources
Q (Justin Post · BofA): He asked how the new model could improve the ad business and whether Susan's strong revenue growth comment includes incremental AI revenue next year.
A: Management said ad improvements come from three areas: ongoing ML work, new LLM research to better reason about preferences, and generative AI tools (eg, Advantage Plus) to help advertisers create personalized creative. They expect new AI businesses can begin contributing incremental revenue in 2027 while core ads and engagement investment continues.
"we expect that generative AI tools, which have already been very valuable for advertisers, will continue to grow in terms of the features that we make available to advertisers primarily through the Advantage Plus suite." — Susan J. Li (CFO)
company focus on Muse and SMB scaling
Q (Mark Shmulik · Bernstein): He asked whether Muse is now central to Meta's strategy and how mandates changed, and he asked what is needed to scale Muse for small businesses.
A: Management said personal superintelligence has been the long-term North Star and they built models and agents end-to-end to deliver personal agent experiences and new hardware. For SMBs, Muse adds skills and connectors (Instagram, ad accounts, Shopify, QuickBooks, Canva) to help with operations, ads, messaging, inventory, and more.
"we need to build models specifically to be good as personal agents." — Mark Elliot Zuckerberg (CEO)
Muse per-user cost and lowering costs
Q (Ross Sandler · Barclays): He asked for metrics on per-user cost today and how Meta will bend the cost curve down going forward.
A: Management said they cannot share specific cost numbers but confirmed Muse is compute-expensive and compute efficiency is a key focus. Near term levers include using spare/off-peak capacity, smaller models, inference efficiency, and product efficiency; longer term they expect LLM advances and AI tools to reduce tokens and improve software efficiency.
"we are not we do not have sort of specific numbers to share around the cost structure of Muse today, but, of course, it is, as Mark said, pretty compute expensive." — Susan J. Li (CFO)
monetization timing and commerce detail
Q (Mark Stephen Mahaney · Evercore): He asked if Meta will monetize Muse faster than usual due to high costs and whether the 25% commerce stat includes travel purchases.
A: Management said they expect to work on monetization sooner than usual because of costs and commerce prevalence, and they plan to take a percent of value created while building the Muse network. They confirmed the commerce figure does include travel, and travel is one of Muse's top use cases.
"My guess is, I think that we are going to work on monetization somewhat sooner on this than our normal playbook." — Mark Elliot Zuckerberg (CEO)