The Zhitong Finance App learned that Meta (META.US)'s total revenue for the third quarter was US$64.6 billion, an increase of 26% over the previous year, with revenue guidance of US$73 billion to US$76 billion for the fourth quarter. The call focused on the Muse personal smart device: over 10 million users in the US and Canada in 1.5 months, with a daily usage rate of about 25%. Shopping is the biggest use case, with an average purchase amount exceeding $200; management expects computing power efficiency to increase by more than 10 times next year, and plans multiple monetization paths such as subscriptions, transaction credits, and corporate APIs.
Meta's total revenue for the third quarter was US$64.6 billion, up 26% year over year, of which advertising revenue was US$63.1 billion, up 26% year over year, and other app family revenue was US$1.1 billion, up 59% year over year. GAAP operating profit for the quarter was US$13 billion, down 37% year on year, and operating profit margin was 20%, mainly hampered by US$9.9 billion in multi-state attorney general settlement legal fees; after excluding this fee, operating profit increased 11% year over year.
Net profit of $9.9 billion, earnings per share of $3.83. Capital expenditure was US$34.9 billion, free cash flow was negative US$2.1 billion, cash and marketable securities at the end of the quarter were US$82.4 billion, and debt was US$83.7 billion. The number of employees exceeded 66,000, down 12% from the second quarter. At the same time, the company announced that it will create a new class of non-voting shares and distribute them to existing shareholders through stock dividends. The actual effect is equivalent to a 2:1 split. Trading is expected to begin in late January 2027.
The core focus of the call was on the progress and monetization path of Muse's personal intelligence. Since Muse was launched 1.5 months ago, more than 10 million people in the US and Canada have set it up, and the daily usage rate is about 25%, which is more than double the initial level of Threads; shopping is the most common use case, with over a quarter of users participating, and the average purchase amount exceeds $200. Management made it clear that Muse has not yet been launched globally, and intensive computing power is the main constraint, but it is expected that next year's computing power cost efficiency will increase by more than 10 times. At that time, profit conditions will be in place after revenue reaches scale.
The business model is based on the free tier and monetized through multiple channels such as transaction extraction, advanced subscriptions, enterprise APIs, and Muse Code. The Meta One unified subscription service has expanded globally, with over 15 million subscriptions and trials. On the enterprise side, the company targets 500 million self-identified small businesses, launched Muse for Small Business, previewed personal intelligence agreements, and cooperated with Walmart, Stripe, Shopify, etc. to establish agency business standards.
In terms of outlook, Meta indicates fourth-quarter revenue of $73 billion to $76 billion, annual capital expenditure of $135 billion to $145 billion, and total annual expenses of $175 billion to $179 billion (including $9.9 billion in legal fees). Reality Labs' losses are expected to narrow to around $15 billion next year. Management did not provide specific revenue guidance for 2027, but indicated that AI investment has increased core business through increased participation, optimized recommendations, and improved monetization efficiency, and is expected to maintain strong revenue growth throughout 2027.
In terms of the development of advertising AI, the generative search architecture has covered about a quarter of unconnected feed views and half of Reels views. The GEM model expansion has increased off-site conversions by 1% to 1.7% and application ad conversions by 9.6%. Nearly 10 million advertisers use generative AI creative tools, and the annual revenue operating rate of cooperative advertising exceeds 15 billion US dollars.
Below are the transcripts of Meta Platforms' third quarter 2026 earnings conference call:
Executive speeches
Mark Zuckerberg
Founder, Chairman and CEO
OK. Hi everyone, thanks for attending today. The past few months have been busy and exciting. Our Family of Apps (Family of Apps) is now used by 3.64 billion users every day, and Facebook and Instagram have both recorded record levels of engagement, thanks to strong video growth. Threads has more than X monthly active users and continues to grow rapidly. WhatsApp has reached a record high in message delivery and is increasingly becoming an important platform for smart devices.
Over the past few months, we've achieved several major milestones in achieving our goal of “providing everyone with personal superintelligence.” Our Muse personal intelligence was one of the most successful product launches ever. We launched the Muse Wonder as our first cutting-edge model, and released a number of new glasses and accessories equipped with Muse integration, including our new audio glasses coming soon.
We've been discussing the vision of personal superintelligence for over a year now. We want to help everyone have a highly capable personal intelligence that knows your goals, helps you achieve them, and improves your life in any way you want — whether it's improving relationships, helping you save time, money, start a business, create new things, or anything else you can think of. Muse is the product form of this vision. So we now have the basic technology to achieve this goal. There is still a lot of work ahead, but our initial position is very strong.
In just 1.5 months, more than 10 million people in the US and Canada have set up the Muse personal smart device. Growth continues to be strong, and we're at the top of the App Store almost every day. Engagement and retention rates are also off to a strong start. Approximately 25% of users who have set up Muse will use it every day. For our earliest adopters, the retention rate seemed to improve the longer they used Muse. For comparison, this is more than double what Threads was when it started, and it's also higher than what we think other AI apps started.
The most common use case for Muse is shopping, where more than a quarter of Muse users participate in shopping, and the average purchase amount is over $200. Plaid tells us they've never seen this level of activity for their connectors on any other service. In the first week, we had to work with Stripe to increase our credit limit because the transaction volume exceeded all of our expectations.
Other than shopping, the most common connectors are personal productivity and connecting Muse to your other services, such as email and calendar. This shows the importance of Muse Secure VM, which we built to provide the most advanced privacy and security, and we look forward to launching Muse Confidential VM soon, when even Meta won't be able to see your content.
We've also seen Muse Wonder further improve quality, engagement, and retention, which indicates that Muse will continue to perform better as we continue to improve our level of intelligence. Now that Muse has a clear product market fit, we are entering a new phase — one that is more operationally focused on scale. The goal for the next phase will be to leverage Meta's many strengths. We need to do three things: expand product scale, increase revenue, and optimize computing power costs.
We expect to expand Muse to serve billions of people. This is both our mission and our strategy. As far as our mission is concerned, we believe in putting the power of personal superintelligence in everyone's hands. We believe this is the path to ensuring a positive future for all, with personal empowerment as a source of prosperity and a balance of power as the foundation for security.
At the value level, we believe making this capability widely available is the best we can do. This is also our strategy because we believe that a large-scale network of multi-billion Muses will be able to do more for individuals and businesses by working together. Therefore, we hope to create a large-scale Muse network as soon as possible.
From a commercial perspective, the average revenue of our current Muse users — sorry, the average revenue per user in our current app family is hundreds of dollars per year, and we expect Muse to reach higher levels over time as we've seen a lot of commercial activity happening. This will take some time, but let's get started now. Our business model will have several components. For most people, the product should be free, and we'll build commercial tools that charge fees based on the deals we help facilitate. We expect this will be paid by the business.
For users who want the most advanced model or use additional tokens, they subscribe and pay directly. For business owners who want extra features to grow their business with Muse, or companies that want employees to use Muse, they can also subscribe and pay directly. For users who want Muse to help improve the personalization and quality of content and ads on the Meta platform, they can also connect to these experiences. To clarify, the average revenue I mentioned earlier refers to the app family income of our current Muse users.
In terms of computing power, we are building the capacity to serve more than 1 billion people over the next few years. Recently, Muse's operation has consumed a lot of computing power, but we expect to increase its cost efficiency by more than 10 times next year, which will make it profitable once the revenue line reaches scale. If computing power efficiency continues to improve rapidly over the next few years—we think this is very possible—then Muse will be highly profitable over many years. Recently, the strength of our core business has enabled us to expand the Muse network before making a profit, and we believe we are the best positioned company to advance this strategy.
Another point about the computing power economy is that since training is a huge fixed cost, now that we have a leading model, we want to serve as many high-profit inference needs as possible. That's why we're building the Meta enterprise platform led by CJ Desai. Muse Wonder and the more advanced models we're training can service additional high-margin services outside of the Muse network, including our API, Muse Code, and other enterprise services.
A fundamental characteristic of our business is that we have many good uses for computing power. It's hard to accurately predict how the Muse Network, our enterprise platform, and Muse Wonder and subsequent models will accelerate the growth rate of our core business. But by building a variety of ways to deploy these advanced models, we've increased the chance of one or more of them significantly exceeding our expectations.
Other than serving other businesses, we haven't discovered the limits of using computing power to accelerate our core business. We've implemented an evaluation of Muse Wonder to improve its effectiveness in independently improving our recommendation and advertising systems. As our upcoming models become more advanced, I expect we'll see these systems improve faster.
OK. Now let's talk about models. Muse Wonder is cutting-edge in many dimensions important to personal superintelligence. This validates the importance of autonomous training models to provide the best experience. In addition to Muse, we've also seen that Meta AI's per capita participation has more than doubled since launching the new model in April.
While Muse Wonder is our most important release, I'm very happy with how fast we executed it. Over the past few months, we've released Muse Spark 1.2 and 1.3, Muse Code, two open weight models Muse Glimmer and Muse Spark Open, and Muse Voice. Looking ahead, as this field progresses rapidly, we plan to simultaneously advance the cutting edge of intelligence and distill these models into smaller, more efficient product models optimized for personal intelligence. My opinion is that we will be able to build superintelligence at scale. As a result, we are methodically climbing up this ladder. We're already working on larger models after Muse Wonder, and they're going to be very impressive and capable. We are also developing a new small model that will be launched soon, which will provide Muse Wonder level intelligence with greater efficiency.
Next up. Our work on personal superintelligence is now also the focus of Reality Labs. Muse will launch on Meta Glasses starting next month, which will be an important step towards this goal. People already love the many features of glasses — music, calls, capturing moments, hearing aids, etc. But over time, we expect Muse and having your personal superintelligence around the clock will be the most important part of the experience.
Muse is also at the core of the eyewear business model. Once again, our mission and strategy is to serve everyone. We believe in putting the power of this technology in everyone's hands. So instead of making big profits on hardware, we're pricing our devices to build large-scale software and super-smart businesses over time. Eyewear sales continue to grow rapidly, and now with Muse, we expect paid subscriptions to increase, and this business will improve significantly from then on.
I also shared at the Connect conference that we're changing our approach to virtual reality, only releasing new devices when we make a major breakthrough, like the Meta VR glasses we launched this year — of course, the first VR experience in the form of glasses ever. I still believe that the path to providing holographic experiences and integrating the physical and digital worlds into everyday glasses will first be realized through VR, but we will advance this investment more rapidly to prioritize delivery of personal superintelligence. Therefore, combining these factors, we expect that Reality Labs' profit and loss will improve from then on, and that operating losses will drop to about $15 billion next year, and trend towards profit from then on.
Before I finish, I'd like to take a moment to talk about the amazing speed we've seen. An important part of our approach has always been to act fast to learn quickly and better serve large-scale users. But recently, things have accelerated. Over the past year, we released Edits, which has over 50 million active users. New apps such as Forum and Seller are being built by small teams and are rapidly expanding to serve heavy users. Muse took only about 8 months from the idea on the whiteboard to a successful product. We're about to release Muse Charm, which will take less than a year from concept to hardware in people's pockets.
I'm proud of what we're building. People have different opinions about the future, but at Meta, we believe the future belongs to everyone, and putting power into people's hands is how we achieve this goal. I think that's important. It takes a company to represent these ideas. But that's not just an idea; it's an execution. So I thank everyone who has worked so hard to deliver all of this to billions of people around the world. As always, thank you to every one of our partners who have traveled with us.
Now please Susan.
Susan Li
chief financial officer
Thanks Mark and good afternoon everyone. Let's start with segment results. All comparisons are YoY unless otherwise stated. The total revenue of the app family in the third quarter was $64.2 billion, up 26%. App Family ad revenue for the third quarter was $63.1 billion, up 26% in both reported and fixed exchange rates. In the third quarter, the total number of ad impressions across our services increased by 10%. Display growth in all regions was healthy, thanks to engagement and user growth as well as ad load optimization. The global average advertising price increased 14% year over year, thanks to improved advertising effectiveness and improved macro conditions compared to the third quarter of last year. This was partly offset by strong display growth, particularly from low monetization surfaces and regions.
The app family's other revenue was $1.1 billion, up 59%, mainly driven by WhatsApp paid messaging and subscription revenue. In the Reality Labs division, third-quarter revenue was $443 million, down 6%, partly offset by continued strong growth in AI glasses revenue due to a decline in Quest headphone sales.
Now let's look at the consolidated results. Total revenue for the third quarter was $64.6 billion, up 26% in both reported and fixed exchange rates. Total expenses for the third quarter were $51.6 billion, up 68% year over year, including $9.9 billion related to the August 2026 multi-state attorney general agreement. The year-on-year increase was mainly driven by law-related costs, infrastructure costs, and third-party AI token costs. Infrastructure cost growth is driven by third-party cloud spending, depreciation, and increased data center operating costs. There were over 66,000 employees at the end of the third quarter, down 12% from the second quarter, after layoffs in May 2026.
GAAP operating revenue for the third quarter was $13 billion, down 37% year over year, and operating margin was 20%. Excluding the third-quarter legal fees associated with the above legal settlement, operating revenue for the third quarter will increase 11% year-on-year. The tax rate for the quarter was 24%, above our guideline range, mainly driven by the non-deductible portion of legal fees. Net profit was $9.9 billion, or $3.83 per share.
Capital expenditure (including financial lease principal payments) was US$34.9 billion, driven by investments in server, data center and network infrastructure. Free cash flow was negative $2.1 billion. Cash and marketable securities at the end of the quarter were US$82.4 billion and debt was US$83.7 billion.
Now on to business performance. There are two main factors driving our revenue performance: our ability to deliver engaging experiences to the community, and our ability to effectively monetize this engagement over time. Regarding the first point, we continue to see significant benefits from content recommendation programs. On Instagram, global usage time per user reached a record high in the third quarter, with double-digit year-on-year growth, and both Reels and Feed performed strongly. On Facebook, the time spent using Reels reached record highs globally and in the US, increasing by double digits year over year. Threads also continued to expand, and the usage time per user increased by more than 20% year over year. The growth in engagement across our services continues to benefit from improvements in the relevance, personalization, and timeliness of our recommendations.
Our investment in recommendation engines in search and ranking continues to yield positive results on Instagram and Facebook. For example, on Instagram Reels, our third-quarter correlation release increased Reels usage time by more than 4%, thanks in part to our upgraded model architecture—ranking a series of Reels at a time rather than evaluating each Reel independently — and improvements we've made to add richer watch time signals.
On Facebook Feed and Reels, we've completed the global launch of a new generative search architecture, inspired by LLM, that uses semantic representation to mark content fragments and generate potential recommendations to help our system present relevant content before it accumulates significant engagement. The new architecture has now been rolled out to around 1/4 of unconnected feed views and half of Reels views, increasing Reels viewing time by 1%.
Our recommendation work also helps us deliver fresher, more personalized content. Currently, nearly 45% of recommended Instagram Reels were posted in less than 1 day, an increase of more than 11 percentage points over the second quarter. On Facebook Reels, we doubled the amount of history the ranking model can consider, increasing the viewing time of American Reels by 1%.
The LLM also helped us understand the content more deeply. We continue to process public Instagram posts every day through an LLM-based content understanding system. By the end of the quarter, these signals had increased engagement in Instagram search results by 1%.
This quarter, we also began migrating our content understanding model to our first-party Muse model series. In offline evaluations, these models outperformed our previous production models and reduced the per-unit content processing cost for Instagram, Threads, Facebook, and ads by more than 80%. In addition to improving quality, having more of this technology stack gives us more flexibility to develop models around the signals that are most important for personalization.
The second driver for turning to our revenue performance—improving monetization efficiency. We see significant opportunities to continue improving the performance of businesses using our services. Within our advertising system, we continue to drive effectiveness by deploying deeper predictive models that better understand users, provide advertising experiences that are more in line with their unique preferences, and make our infrastructure more efficient. Last quarter, we launched Meta Generative Recommender, which uses LLM to infer ad content and user preferences to predict the best ads for everyone.
In the third quarter, we launched this generative search method on Facebook Feed, increasing the conversion rate for the target market segment by nearly 2%. Furthermore, we expanded the GEM model of ad ranking to more service surfaces, increasing the conversion rate of off-site conversion activities by 1% to 1.7% and the conversion rate of application ads by 9.6%.
We continue to efficiently expand our search system. In the third quarter, Andromeda's improvements — such as scaling model complexity and throughput — increased Facebook's conversion rate by 1.4%. We're also using AI to make it easier for businesses to manage activities, develop ideas, and seize opportunities for results.
Meta AI Business Assistant is becoming a more proactive and capable partner for advertisers. We estimate that advertisers who adopted its recommendations increased their conversion rate by nearly 5% and increased their ad investment by 9% at the same cost per conversion. Compared to those who did not adopt it, the situation was similar. We're rolling out more capabilities, including creative insight, stronger business context, and more agent-based actions. These features will enable marketers to plan, create, and optimize campaigns through conversational experiences, and schedule tasks and reminders while maintaining control through approvals and previews.
I'm sharing an example of Meta AI business assistant empowering advertisers. Simply Organic Maple, a family owned producer of maple syrup in Vermont, uses Meta AI commercial assistant to improve its campaign information. After Business Assistant recommended testing longer, more descriptive ad copy, the company increased purchases by 57% and reduced the cost per purchase by 36%. Its strongest revised ad achieved a 3x return on ad spend, a 52% increase over the original.
The adoption of our generative AI ad creative tools is also growing. Nearly 10 million advertisers used at least 1 of these tools in the third quarter. Advertisers' adoption of our video generation feature increased 25% month-on-month, thanks in part to pre-made videos shortening the generation time from uploading a single image to just a few seconds. We're also expanding opportunities at the intersection of creators and advertising. Collaborative advertising, which allows advertisers to run ads with creators, brands, and other businesses, continued to expand rapidly in the third quarter. Our cooperative advertising products grew by more than 97% year over year, and the annualized revenue operating rate exceeded $15 billion.
In the third quarter, we also completed the global promotion of the Creator Marketing Hub to all advertisers, bringing creator discovery, content organization, and collaborative ad activation in one place. Currently, more than 10 million creators can be discovered. We're integrating Meta AI commercial assistant into the hub to enable advertisers to search for creators in a conversational way, identify content for collaborative ads, and get guidance in their creator marketing workflows.
Beyond advertising, we're also building more ways to monetize engagement in the ecosystem. In September, we announced the global expansion of Meta One, our unified subscription service for consumers, creators, and businesses. Meta One is gradually being launched, covering more than 50 features of Instagram, Facebook, and WhatsApp, including Meta Business Agent and Meta AI, and has so far surpassed 15 million subscriptions and trials. Our app and the core experience of Meta AI will remain free, while subscriptions provide more specialized capabilities and expanded AI usage.
In addition to monetizing engagement with existing services, we're also building new experiences and revenue streams around consumer intelligence. Mark highlighted Muse's impressive early momentum. Users have found many reasons to use Muse; the top three categories are shopping, personal productivity, and creativity. We're encouraged by the breadth of use cases, particularly shopping, which is used by more than 25% of Muse weekly active users.
While our immediate priority is to expand the adoption and reach of Muse, we're seeing multiple monetization paths, starting with subscriptions. We're also seeing significant opportunities to expand Muse to small businesses. According to a recent user survey, we estimate that at least 500 million self-identified small businesses use Meta's family of apps to reach customers and generate revenue. Muse helps these businesses acquire customers and operate more efficiently, while helping people complete transactions. Over time, consumers and business intelligence can directly connect and transact, creating opportunities around paid usage, commerce, and enterprise services.
In addition to growth, we focus on building sustainable unit economics. Muse's computational power intensity means that expanding its user base will require significant investment, and we expect total service costs to rise. At the same time, we expect to increase the cost efficiency of computing power by more than 10 times next year, thanks to better use of off-peak capacity and more efficient models and inference optimization. As revenue expands, these efficiency improvements give us confidence in our ability to build a profitable business over time.
Now on to our financial outlook. We expect total revenue for the fourth quarter of 2026 to be between $73 billion and $76 billion. Our guidance assumes a headwind of around 1% year-on-year growth in foreign exchange for total revenue, based on current exchange rates.
Move to an expense and capital expenditure outlook. We raised our cost outlook to include the $9.9 billion in legal litigation costs confirmed in the third quarter. We now expect the total cost for the full year 2026 to be between $175 billion and $179 billion. Excluding legal fees for the third quarter, we continue to expect operating revenue in 2026 to be higher than in 2025. At the segment level, we now expect Reality Labs' losses to fall to between $18 billion and $19 billion compared to 2025.
We expect capital expenditure (including financial lease principal payments) to be between US$135 billion and US$145 billion in 2026, which is narrower than the previous outlook of US$130 billion to US$145 billion. In the absence of changes in the tax environment, we expect the tax rate to be between 15% and 17% for the fourth quarter of 2026.
Additionally, we continue to focus on active legal and regulatory matters that may have a significant impact on our business and financial results. For example, we continue to see scrutiny of youth-related issues in multiple markets, and several juvenile-related trials are scheduled in the US in the next few months.
Looking beyond 2026, we are still developing next year's budget and capacity plans, but we expect to continue to actively invest to support core business and emerging AI initiatives. Our current production capacity and planned launch capacity give us the flexibility to allocate resources between model training, core business growth, and new product development and support.
Our investment in AI has already generated significant incremental ad revenue through stronger engagement, better recommendations, and more efficient monetization. Building on this, our MSL investment is creating the technology and infrastructure needed to unlock significant new revenue opportunities and enhance overall business capabilities. While these investments will take longer to pay off than core business investments, the breadth of monetization opportunities and internal efficiency give us confidence in their long-term potential. Looking ahead, although we are not providing a full-year revenue outlook for 2027, the benefits our AI investment brings to our core business, combined with model advancements and emerging opportunities for product launches, will enable us to maintain strong overall revenue growth throughout 2027.
Additionally, we continue to focus on improving Reality Labs' profitability. As Mark mentioned, we expect our losses to narrow to around $15 billion next year as we reallocate our investments to other priority areas, mainly AI.
Finally, today we announced plans to create a new class of non-voting capital stock that will be distributed to all existing shareholders through stock dividends. For every 1 existing share held, shareholders will receive 1 new share without voting rights. The actual effect is equivalent to a 2 to 1 stock split. We expect trading to begin in late January 2027.
Overall, our business momentum continued in the third quarter, with strong execution of core advertising and engagement plans, and we're excited about progress in achieving personal superintelligence.
Now let's open the questioning section.
Q & A session
Brian Nowak
Morgan Stanley Research Division
I have two questions, one about Muse and one about enterprise business. First of all, congratulations on Muse's success. But as you analyze user behavior and products, can you talk about any of the constraints on the forward user growth slope that you are most concerned about — whether it's capacity, model efficiency, production capacity, etc. — so we can understand the predictive user growth constraints you want to address?
The second one relates to corporate business. Mark, can you outline your vision for Meta enterprise product supply and the key value drivers you want to provide to businesses that adopt the service?
Mark Zuckerberg
Founder, Chairman and CEO
Yes, I think I can address these two issues and then Susan can add. OK. So now for Muse, what I'm pointing out is that we have over 10 million people set it up in the US and Canada, but I think one of the biggest constraints so far is that since it's a computationally intensive service, we haven't launched it globally. Part of this is due to the desire to optimize for every location, and some work on connectors. But that's partly because we want to make sure this is done efficiently.
So there are pros and cons to this. The benefit is, of course, that made it even more impressive for 10 million people to use and sign up in the first 6 weeks, as these are all Americans and Canadians, and are concentrated in a small number of markets. If you were to launch it more widely, you'd expect that number to be significantly larger than what we're seeing today. But on the other hand, we actually have to do a good job to raise efficiency to a reasonable level.
So that's an important part of what we need to do. Computing power efficiency work is also important in this regard, because we need to ensure that we have built enough computing power to serve what we think is more than 1 billion users (if we can meet certain computing power efficiency goals), but we clearly still have a lot of work to do. This will be one of the limiting factors for next year.
We want to build a business around it. So I talked about that. I think going along with reducing computing power — or improving computing power efficiency — is important to make all of this work from a commercial perspective, and I'm optimistic about this, but I think this is one of the things we need to do.
Then we continue to increase the value of our products, which I think we mainly do by continuing to develop new models. That's why I mentioned that we first launched Muse on Muse Spark, which is currently the default model for the free tier. We launched Muse Wonder, and people can access it today at the paid tier. We know it performs much better, so we're working on a more efficient version of Muse Wonder so it can be rolled out more widely. But as you can imagine, as we iterate two or three models forward, I think this will unlock more value for people when you have significantly more intelligence and the ability to solve more problems, and deliver at a more affordable price point. So we're really excited about it.
Moving on to the enterprise part because I think there is some connection between these two parts, and the main enterprise products in the near future will be Model API and Muse Code. We've had enterprise intelligence companies expanding for a while, and all three are growing rapidly. So we're excited about it. Over time, we hope to establish a complete Muse Smart Commerce Network. There are business agents, which we've talked about in our previous earnings call, to help businesses with sales and customer support.
But we've seen a lot of people actually using Muse to help run small businesses in particular. We believe there is an opportunity to help create an intelligent body that people can use to actually automate many of the difficult, time-consuming, but not core, parts of starting a small business. So I think it's a big deal.
Then make a version of Muse that people can bring into different corporate environments and work there, and we think that's going to be a big deal. Other than that, building what I call a Muse network means making Muse more than just a single player game — you have your Muse — but rather your Muse can collaborate with your friends and family. If you want to buy something, don't just navigate to that company's website, but talk to their smart body. I think having all this well collaborated would unlock tremendous value at scale.
So I'm really excited about all of this. I think enterprise use cases will be an important part of this, but it will also have a very valuable utility boost for consumers. So I'm really excited about both.
Susan Li
chief financial officer
I'd like to add Mark's review about the business. AI is creating a lot of disruption in businesses today. From what we've seen, we really don't think traditional enterprise companies have locked in this. In fact, some of the leading companies in this field today didn't exist a few years ago. Those companies have invested heavily in model development, and we have done the same. We have increasingly competitive models that can benefit all users, small businesses and businesses alike. Today, 500 million businesses already use our products.
So we really believe AI will reshape the way businesses operate. We expect businesses of all sizes to use AI agents for software development, productivity, sales, marketing, customer service, etc. We have differentiated assets to serve this market, combining models, infrastructure, global distribution, and long-term relationships with businesses. We hope this will be a new diversified recurring revenue stream for us, supporting multiple business models, including subscriptions and usage-based APIs and messaging, and other revenue lines. We anticipate this will be an opportunity that will take years to develop, and we intend to expand our investment based on demonstrated customer needs and product market fit.
Eric Sheridan
Goldman Sachs Group Research Division
The first question continues Brian's question. When you think about the friction points of agency-based commerce, how do you, as a company, address the friction that aggregators or suppliers don't want agents to interact with their platforms or services, and how do you think about reducing this friction or creating the right balance between existing services and how agency behavior might change in the medium to long term?
Second question, with tonight's earnings report announcing new stock classes at the same time, are there any updated ideas on how you view capital sources and optimizing capital structures to ensure proper investment to take advantage of the opportunities presented by AI in the next few years?
Mark Zuckerberg
Founder, Chairman and CEO
I can answer the first one. So what we're seeing now is that the vast majority of businesses are very excited to support this, as they recognize that this is not only a new opportunity to increase sales and quality of experience for existing customers, but also bring in new customers. So, across the business ecosystem, different people will adopt this at different rates, and how our strategy evolves in this regard will depend on how it progresses. But overall, I'd say we were very pleasantly surprised at how quickly a number of businesses not only wanted to figure out how to make Muse use more effective use of their services, but also took the initiative to reach out to us to integrate around this. I think that's a very good sign.
Obviously, not everyone wants to be an early adopter. It doesn't matter. We'll work with people on terms that make sense to them. But I do think that getting this all to work well over time will be very valuable to everyone, and I think we're currently on a good track in this regard.
Susan Li
chief financial officer
Yes. I'd like to add that I'll also answer your second question. I think smart devices will be an important part of the evolution of reality, and I think everyone must adapt to what the new era of agency commerce means. We clearly have a number of businesses that are very excited to work with us. There are also businesses that understandably focus on protecting customer relationships. We expect the market to eventually move away from smart devices trying to navigate every digital experience like a human being. I think in the long run, this will lead to more permissioned, structured interactions, including certified APIs, approved agent partnerships, or direct communication between individual consumer agents and the enterprise's own agents.
We're helping to establish the standards needed to support this model. Earlier this month, we previewed the Personal Agent Agreement, an open standard across Muse and Meta business agents that we're developing with industry partners such as Sierra, Walmart, Stripe, and Shopify. The agreement aims to define how consumer personal agents can safely and reliably interact with business agents and preserve appropriate controls for businesses and customers.
So, again, I think adoption may be uneven and take time, and we anticipate any experience—many of which will remain AI-assisted rather than fully autonomous in the near future. But as trusted integrations evolve and agents prove they can generate incremental value and demand and operate safely, we expect more and more businesses to participate.
Your second question is about announcements of new stock classes. I should explain that our purpose in creating this new stock class is to create a publicly tradable, non-voting stock that can be used for capital raising, employee compensation, acquisitions, and other general corporate purposes, while retaining the voting rights of existing Class A and Class B shareholders. We believe this stock class will maintain our founder's leadership structure. It will expand our financing tools and give us the flexibility to invest in the unprecedented AI-driven momentum we're seeing across all parts of our business. We currently have no specific financing plans to announce, but we believe it would be prudent to establish an appropriate capital structure framework to ensure that we have access to diverse funding sources.
Douglas Anmuth
J.P. Morgan Research Division
I have two questions. First, I'd like to follow up with Susan's last comment about the evolution of Muse and the adoption of agents by more businesses. I want to know your views on how things are changing and transforming when moving towards more licensing and agent-to-agent transactions. What exactly is this releasing over time?
Second, with regard to expenses and capital expenses, think about 2027. I know you guys haven't really gotten to that point yet. But if I go back to some of my previous reviews, you guys talked about maximizing computing power in 2026 and 2027, then having more flexibility in 2028. A year ago, you talked about a significant increase in the US dollar in 2026 compared to 2025, and then you also used the increase in annual operating revenue. I'm just curious if any of these can be linked to help think about 2027.
Susan Li
chief financial officer
Thanks Doug. Regarding your first question about how the proxy ecosystem will evolve, I wish I could predict it more perfectly. I think this will take some time. Again, I think it would be a bit uneven. But I definitely think that when I use Muse today, it's clear that many parts of the experience would be better if Muse could interact more seamlessly with enterprise agents on the other side. For example, if Muse had to call a business on my behalf, I think it would be easier for the business if that call was handled by a business entity representing the business.
What I think is really unique about Muse, and what makes the product experience so great, is that it represents the level of initiative I have shown, right? I can give it a task, and Muse will perform that task. Even if I run into an obstacle or can't find the right thing, I need to open up a new path to help me find what I'm looking for. So I think this proactive dimension represented by consumer intelligence or business intelligence will enable more transactions or more activities to occur more widely than today when more actions are taken — when potential needs express themselves — compared to when Muse continues to search for things for me, regularly updates me what I'm looking for, and sends me 10 new books every month. Intelligent entities should do similar things for enterprises. I think this creates an opportunity. Thirdly, it creates a new, more active framework for how agents interact with agents over time.
Regarding your second question about 2027, we are in the early stages of developing next year's budget and capacity plans. So I'd like to say that we're in the process of evaluating a number of investment proposals with attractive return characteristics over different time frames, and we want to figure out how to fund as many of these proposals as possible within the framework of our overall budget and financial goals. But it's very dynamic, and it's still pretty early. So we don't have more specific or quantitative information to share until the fourth quarter conference call, which is usually when we share our guidance for the next year.
Justin Post
Bank of America Securities Research Division
OK, I have two questions, one for Mark and one for Susan. Mark, you have a powerful new model in your hands, and it seems like there's a lot you can do internally in addition to the outside. But when it comes to the ad business specifically, you've restructured the system around AI. How much room for improvement as you deploy that new model? How do you think?
Then to Susan, you mentioned strong revenue growth next year. Wondering if that's advertising or if it includes some incremental AI revenue to be realized next year?
Susan Li
chief financial officer
I'd be happy to answer both of these questions. OK. Regarding your first question, I think our advertising business has been benefiting from our continued investment in so-called traditional AI machine learning models and recent generation models for many years. I'd like to say there are three topics worth outlining. The first is a traditional roadmap based on performance improvements in our pipeline — I'd describe it more as business-as-usual machine learning work. This continues to be a strong and healthy pipeline. As we look to 2027, again, we didn't provide an outlook, but I'd say our investments in engagement and advertising programs and product improvements gave us confidence that there is meaningful room for growth throughout 2027. This also relates to your second question. But I should say, basically, I think the sequencing and recommendation pipeline is healthy.
The second theme is that we're excited to bet on some new research towards more LLM-oriented AI infrastructure to help us more deeply infer people's preferences and what they're interested in in organic content and recommendations, and we expect we'll be able to unlock the benefits of ranking and recommendation engines from this research.
Third, we expect generative AI tools — already very valuable to advertisers — to continue to grow in the features we provide to advertisers through the Advantage+ suite, which make it easier for advertisers to quickly and seamlessly generate ideas, or generate video or sound from an image, all of which I think are meant to serve to make ads more and more personalized to the recipient. So I'd like to say that we think all three areas have meaningful room for growth.
Your second question concerns next year's revenue growth. As I mentioned, I think we're feeling good about the budget investments expected to be made next year in core advertising and engagement plans. We're also building several new AI businesses, and we expect these to start contributing incremental revenue in 2027, but more importantly, can meaningfully expand over time. These will include monetizing our AI products through subscriptions, transactions, and advertising, and developing corporate products, as we mentioned in the previous question. So I'd like to say that while many of these opportunities are still early, and our immediate focus is really on building high-quality products and achieving product market fit and large-scale product market fit, I think our record in reaching billions of people and building businesses on a large scale makes us feel good that AI will become a meaningful source of diversified revenue over time.
Mark Shmulik
Bernstein Institutional Services LLC Research Division
Mark, coming out of the Connect conference, I have a feeling that we are entering a new era for the company, where the business is repositioned around Muse. Is this a fair assessment of how internal matters are progressing? If so, can you talk more about how some of the business mandates and outcomes might change?
Then, Susan, it's interesting to see the pace of the launch. The business announcements preceded some small business announcements. I appreciate some of your comments on SMB smart opportunities. But can you share more details about services and expanded SMB offerings?
Mark Zuckerberg
Founder, Chairman and CEO
Yes. I can talk briefly about the first one. I mean, over a year ago I wrote a letter about personal superintelligence as our working North Star. It's basically in line with the work we've been doing at the company for over 20 years, focusing on empowering individuals, giving people the chance to make a voice, and giving people the tools to connect the way they want. As we extend this to our current work around AI, we think the most important thing we can build for the next generation of technology—basically putting the power of this technology in everyone's hands is a personal vision of superintelligence.
So we worked on this for a while, and that's when we restarted and set up a lab around MSL. It's largely the idea that we need to build models that specialize in being personal intelligence, and we'll be able to build the best experience by owning, building, and designing end-to-end technology. This includes models and agents. This has been in the works for a while.
The same is true of the vision surrounding glasses. We wanted to build a next-generation hardware platform that would provide the best end-to-end experience for people. I think this is starting to take shape, and I think it will be even more so in the next few years. These are long-term bets. I think that's what we're here to do, make long-term technology bets, and design the best end-to-end experience for billions of people around the world. I think as long as we keep doing this, the business will take care of itself over time, and as long as we continue to execute well, I'm very proud of the team's performance in this area.
Susan Li
chief financial officer
Thanks Mark. Your second question is about Muse for small businesses. So we launched it in late September, and you can think of it as an extension of Muse rather than a separate product. It's a set of new skills and connectors within existing Muse apps to help SMB owners operate and find new customers. It only takes a few clicks to connect to a business's Instagram account, analytics or Facebook page or Meta ad account, as well as dozens of third-party connectors that the business already runs, such as Shopify, QuickBooks, Canva, etc. It already knows what the company sells, what the brand sounds like, and what customers are asking.
Muse can take action, including analyzing ad and content performance, drafting campaigns, customer responses, tagging questions, and we're actively listening to SMB users to understand their key use cases. We've seen them use it for ad optimization, direct message organization, growing sales footprints to keep the site up to date based on inventory and new products, and handle employee load balancing. So we're watching how businesses use Muse pretty closely to ensure we carefully consider the new features Muse releases to make them as useful and efficient as possible for SMBs.
I'd also like to say that we have unique touchpoints. Again, in my opening remarks, I mentioned 500 million self-identified small businesses. We think much of our investment in consumer products will continue to help drive these businesses. We have several excellent channels to showcase these opportunities to small businesses around the world. So this is a place where we feel we already have a very strong existing platform, and we're really excited about it. We also believe that the transition to AI has broadly unlocked new opportunities for businesses as us.
Ross Sandler
Barclays
OK. Mark, you mentioned this before, but can you elaborate on some of the high-level metrics of Muse's cost structure per user today, and then how do you think the cost curve has declined over time? What are you doing to reduce that cost curve?
Susan Li
chief financial officer
I'd be happy to answer. Mark, you should add anything you think I'm missing. We—we don't have specific figures on Muse's cost structure to share today. But of course, as Mark said, it's quite expensive, and making sure we focus on computing power efficiency will be one of the key vectors to be able to scale Muse and grow it into the business we want.
I think there are several dimensions to achieving this kind of efficiency. I think some of the more recent parts include using off-peak or backup computing capacity to run certain task sets and use smaller models. There's a lot of work to be done on different inference efficiency plans, all of which are actually for higher GPU utilization and other product efficiency work we can do.
I think in the long run, I think this is something we can expect to be able to do over time — as LLM expands, uses fewer tokens to complete more tasks, and relies on the AI model itself to help us improve the efficiency of the software we use to run the model. So I think this is a place where we both feel that there is still a lot of work on the efficiency roadmap that will help us curve the curve, and we also anticipate that AI-driven technological advancements in this application will help us achieve greater savings over time.
Mark Stephen Mahaney
Evercore ISI Institutional Equity Research Division
Two questions. Over the years, Meta has had a routine where you present a new surface and build user growth and engagement, then superimpose monetization. What do you think is a similar routine here? Or does it have to be a little faster because the costs associated with Muse are arguably much larger than when those surfaces were launched in the past?
Then clarify the commercial data points you provided, approximately 25% users, and $200+ ASP. Are you including tour purchases? Or is it just a retail data point?
Mark Zuckerberg
Founder, Chairman and CEO
My guess is that I think we'll start to monetize sooner than our normal routine for a few reasons. One is the digital scale we're talking about — whether it's cost or opportunity — I think making progress in this area is important to be able to scale it up. Another is that a big part of what people do is commerce.
To answer your question, that also includes travel. I think we might think of travel as a different situation, but travel is actually one of Muse's other top use cases. Basically, people plan different kinds of things, and they may get it through our family of apps. So I think this is going to be an important thing for us to deliver the consumer value people want, and in doing so, I think we'll be able to build a model where we get some of the value we help create for people. So I think this is going to be -- it's important to figure out all of this math so we can continue building the computing power we want to build to reach everyone's scale.
In addition to everything Susan said in response to the previous question, I'd like to add that we're just seeing this trend, and on a year-over-year basis, I expect we'll be able to deliver the same intelligence a year from now with a smaller model at a fraction of the cost of the current Muse Wonder, or frankly any current cutting-edge model. So, if this represents what people seem to be — the kind of baseline level sufficient for the Muse product market fit, I think we'll make progress in efficiency through continued model development.
What's unclear, of course, is that I don't think most people will want to use the current model level of intelligence in a year, because that will also improve. So I think we need to work on all of these things, including the revenue portion and expanding various aspects, and advancing model development and all efficiency, so that we can both fund and support computing power construction, and reach billions of people. Again, we want to get there as soon as possible.
Part of the reason we don't just pay everyone upfront to use this is because we wanted to build this Muse network, and we think when there are lots of people using Muse, it will be more valuable to people in terms of scale. That way, they can collaborate and work together. I think this is a really important state of affairs, and we're really excited about it. Again, I don't think there are many companies in the world that are as well positioned to do this as we are. I think we're in a good position to do this, and we plan to move forward quite aggressively.