Meta Q3 2026 Earnings: AI Spend, Muse Adoption Test

Meta Platforms reports third-quarter 2026 results after market close on Wednesday, October 21, with a call scheduled the same day, the company confirmed in a October 12 release. The timing puts two numbers on a collision course: Wall Street’s roughly $41.68 billion estimate for Meta’s quarterly AI capital spending, and the still-murky adoption curve of Muse, the personal AI agent Meta introduced in September 2026. Reuters, Investing.com, MarketWatch and Ars Technica have all flagged the same question heading into the report: can Meta show that its AI bill is buying real user demand, or just bigger data centers.

The stakes are not abstract. Meta’s 2026 capital expenditure guidance already sits at $130 billion to $145 billion, more than double what the company spent two years earlier. Muse, meanwhile, has gone from a model codename inside Meta Superintelligence Labs to a subscription product with monthly plans, business connectors, and a privacy controversy, all within about six weeks. This earnings cycle is the first real test of whether that spending and that product can be mentioned in the same sentence as “return.”

Google · Preferred Sources

Don't miss new tech stories on Google

Add TrendinTech once in the Google app and our stories appear in your news suggestions.

Add Now

Meta’s Q3 2026 Earnings Date and What Wall Street Expects

Meta announced on October 12 that its third-quarter 2026 financial results would post after market close on October 21, with a conference call to follow. That is one of the few facts in this cycle that is not up for debate. Everything else being discussed by analysts, from capital spending to Muse subscriber counts, is an estimate dressed up as near-certainty.

Analysts tracked by Visible Alpha project Q3 revenue of $63.42 billion, up 23.8% year over year, landing inside Meta’s own guidance band of $61 billion to $64 billion issued back in July. The same analyst pool estimates capital expenditure of about $41.68 billion for the quarter, more than double the year-earlier level, and projects free-cash-flow burn of roughly $8.52 billion. None of those three figures are confirmed numbers. They are forecasts that Meta’s actual filing will either validate or embarrass on October 21.

A separate wrinkle sits in the cost line. According to reporting from moomoo, Meta has signaled it expects to accrue roughly $10 billion in Q3 legal expense tied to an agreement with a group of bipartisan state attorneys general. That is a cost that has nothing to do with GPUs or data centers, but it still lands on the same income statement investors are using to judge whether AI spending is paying off. It is the kind of detail that can make a quarter look worse than the underlying AI business actually performed, and analysts flagged it as a factor that could muddy the read on Meta’s core margins.

The AI Spending Numbers Behind the Headlines

Meta’s own numbers from earlier in 2026 give the clearest baseline. First-quarter 2026 revenue came in at $56.31 billion, up 33% year over year, with capital expenditures of $19.84 billion and free cash flow of $12.39 billion, according to the company’s official first-quarter release. By the second quarter, revenue had grown to $60.8 billion, a 28% year-over-year increase, but free cash flow had fallen sharply to about $784 million as spending accelerated. That drop, from $12.39 billion in free cash flow in Q1 to under $1 billion in Q2, is the single clearest signal that Meta’s AI buildout is now eating directly into the cash the company generates.

If the Visible Alpha estimate of $41.68 billion in Q3 capital expenditure holds, it would mark another step up from the pace set in the first half of the year, and it would push Meta further toward the top of its full-year $130 billion to $145 billion guidance range. That range already more than doubles the capital spending Meta was running at in 2024, a shift that has drawn scrutiny from institutional investors watching AI capital allocation across the sector, not just at Meta.

Metric	Q1 2026 (actual)	Q2 2026 (actual)	Q3 2026 (estimate/guidance)
Revenue	$56.31 billion	$60.8 billion	$63.42 billion est. (guided $61B-$64B)
Revenue growth, year over year	33%	28%	23.8% estimated
Capital expenditure	$19.84 billion	Not separately confirmed	$41.68 billion estimated
Free cash flow	$12.39 billion	$784 million	$8.52 billion burn estimated
Full-year 2026 capex guidance	$130 billion to $145 billion

Every figure in the Q3 column is an analyst estimate or a company guidance range, not a reported result. Meta’s actual numbers land on October 21, and the gap between estimate and reality, especially on capital expenditure, is likely to move the stock in either direction more than the revenue line will.

Muse: From Model Codename to Paid Assistant

Muse did not start as a consumer product. Meta Superintelligence Labs introduced Muse Spark on April 8, 2026, describing it at the time as Meta’s most powerful model yet. By July 24, an updated Muse Spark 1.1 was powering planning, email and calendar connections, slide creation, and task execution inside the Meta AI app. Meta also shipped Muse Image, its first dedicated image-generation model from the same lab, on July 7.

The turn from model family to product came in September 2026, when Meta introduced Muse itself: a personal AI agent meant to act on tasks, manage projects, and turn stated goals into action plans across phone, computer and the web. Internally, Muse is also known by the codename Hatch. Meta positioned it as a shift away from the conversational chatbot format of earlier Meta AI releases and toward something closer to an assistant that actually does things, including sending emails, helping sell a car or other assets, and booking travel, on the user’s behalf and with approval required before anything publishes, sends, or spends.

Meta did not stop at the consumer version. On September 29, with an update on October 8, the company expanded Muse into a small-business agent with connectors for Asana, Box, Canva, Dropbox, Figma, Intuit QuickBooks, Notion, Shopify, Slack, Stripe, Zoom, and Meta’s own Facebook Pages, Instagram professional accounts, and ad accounts. That expansion matters for the earnings narrative because it gives Meta a second monetization lever beyond consumer subscriptions, business workflow automation, that it can point to even if individual-user Muse adoption proves slower than hoped.

How Much Muse Actually Costs

Meta’s own subscription help page lists Muse pricing plainly: a free tier with a usage limit, a Power plan at $20 per month that includes 500 million Muse tokens per week, and a Maximum plan at $100 per month with 3 billion Muse tokens per week. That token-based structure is unusual for a consumer AI assistant, and it signals Meta is trying to capture heavy users, the ones running Muse against business connectors all day, rather than casual chatbot users who might never hit the free tier’s limit.

Tier or model	Price	What it includes	Status as of Oct. 20, 2026
Muse free tier	Free	Basic agent tasks, usage limited	Live, US and Canada
Muse Power	$20/month	500 million Muse tokens per week	Live
Muse Maximum	$100/month	3 billion Muse tokens per week	Live
Muse for Small Business	Free allowance, paid tiers beyond it	Connectors for Shopify, QuickBooks, Stripe, Slack, and more	Live since late September
Muse Wonder / Muse Wonder Full	Bundled with Muse Code subscriptions	Frontier agent models for developers and heavy subscribers	Live since Oct. 12
Muse Spark Open	Free, open weights	816B total / 42B active parameters, 1M token context	Live since Oct. 15

For more on how Meta is positioning those paid tiers against rival coding and agent subscriptions, see the earlier breakdown of Meta’s Muse Code pricing strategy, which lands squarely in the middle of what OpenAI and Anthropic charge for comparable developer access.

Muse Wonder and the Frontier Model Bet

On October 12, Meta’s newsroom introduced Muse Wonder and Muse Wonder Full, describing them as the company’s most capable models to date, built for agentic, multi-step tasks rather than single-turn chat. Muse Wonder is available through the Muse app’s paid plans and through Meta’s model API for developers. Muse Wonder Full, the larger sibling, is reserved for Muse Code subscribers, mirroring how OpenAI and Anthropic gate their most capable models behind developer tiers.

CNBC’s coverage of the October 12 launch framed it directly as “a test of users’ willingness to pay,” a framing that lines up with the broader question Meta faces on October 21: not whether Muse can do agentic tasks in a demo, but whether enough people will pay $20 or $100 a month, quarter after quarter, to make the underlying infrastructure spend worthwhile. Meta’s stock already moved on the Muse Wonder announcement once, a reminder that the market is pricing in adoption well ahead of confirmed numbers.

Muse Spark Open and the Enterprise Platform Push

Three days after Muse Wonder, on October 15, Meta released Muse Spark Open, an open-weight, natively multimodal model with 816 billion total parameters, 42 billion active parameters, and a 1-million-token context window. Giving away an open-weight model of that scale, even as Muse Wonder stays locked behind paid tiers, is Meta repeating the playbook it ran with Llama: open models build developer goodwill and ecosystem lock-in, while the frontier model stays commercial. Inference framework SGLang added day-zero support for Muse Spark Open, a sign the open-source AI tooling world is treating the release as a serious entrant rather than a Llama afterthought.

The same week brought a structural change that has gotten less attention than the model releases: Meta introduced a Meta Enterprise Platform and named Chirantan “CJ” Desai, the outgoing CEO of MongoDB, as its chief enterprise platform officer, a newly created role reporting directly to Mark Zuckerberg. Pulling a sitting database-company CEO into a brand-new enterprise role two days after Muse Spark Open shipped suggests Meta is building out a business-software arm around Muse for Small Business rather than treating it as a side feature of the consumer app. That kind of hire rarely happens unless a company expects enterprise AI revenue to show up on an income statement within a few quarters.

Adoption Numbers: Why the Estimates Don’t Agree

This is where the earnings preview gets genuinely uncertain. Meta has not published an official Muse user count, and third-party estimates vary enough to call each other into question. Yahoo Finance reported Muse crossing roughly 3 million weekly users and 5 million downloads within about three weeks of its September launch, while separate analyst tracking cited elsewhere put daily active users anywhere from 1.3 million to 2.5 million, alongside roughly 300,000 daily US downloads and about 20% 14-day retention. None of those figures come from Meta directly, and they do not agree with each other closely enough to treat any single one as the real number.

What that spread tells investors is less about Muse’s actual popularity and more about how little visibility outsiders have into it. Meta’s Q3 call is the first point where the company could, if it chooses, offer an official figure. Historically Meta has been selective about which product metrics it discloses on earnings calls, and AI assistant adoption is exactly the kind of number a company highlights when it is strong and omits when it is not. Yahoo Finance’s own reporting carried a warning from one analyst that monetization of a product like Muse could prove “notoriously slow,” even if download and engagement numbers look respectable in the first few months.

The Muse Privacy Problem Investors Can’t Ignore

Adoption numbers are not the only overhang on Muse heading into earnings. Ars Technica reported that Muse exposed users’ personal and business secrets when conversations were shared, a finding serious enough that The Verge confirmed Meta paused Muse’s conversation-sharing feature after researchers found shared AI chats sitting exposed on the open web. For a product Meta is actively trying to sell into small businesses through connectors to Stripe, QuickBooks and Shopify, a sharing feature that leaked business data is not a cosmetic bug. It is the kind of story that can slow enterprise sales cycles exactly when Meta is trying to accelerate them with a new chief enterprise platform officer in place.

The episode also feeds a broader data-collection concern around Muse, after reporting earlier in October suggested Muse was compiling detailed profiles on millions of users and the people they interact with. Meta has not confirmed a specific user count tied to that reporting, but the combination of a paused sharing feature and questions about data retention gives analysts another reason to ask pointed questions on the October 21 call, beyond the usual revenue and margin questions.

Zuckerberg’s Bet: What He’s Told Investors

Mark Zuckerberg has been unusually direct about the scale of the wager he is asking shareholders to make. In comments reported by the BBC, Zuckerberg acknowledged the scrutiny directly, saying, “I get that this is a big bet across the industry.” He framed his own confidence in long-term terms rather than quarterly ones, telling the outlet, “My personal bet is that the people who invest in this will feel very good and be rewarded over time.”

Zuckerberg has also argued the spending is already paying off inside Meta’s existing businesses, saying AI spending “was accelerating every part of our core business,” according to the same BBC report. On where that spending leads next, he pointed past Muse’s current feature set toward autonomous agents, saying “such agents will be the next wave of our product line in the months and years to come,” and describing a future in which “soon, we’ll have agents that can work 24/7 on your behalf.” Those comments, taken together, suggest Zuckerberg sees Muse less as a finished product and more as the first public step in a much longer agent roadmap, one that will keep requiring the kind of capital spending analysts are now estimating at $41.68 billion a quarter.

Competitive Field: Muse Against the Rest of the Field

Muse is not launching into an empty market. Meta is competing for agent and assistant mindshare against OpenAI’s consumer and developer products, Google’s Gemini line, and Anthropic’s Claude family, all of which have shipped major updates in the same two-month window Muse has existed. Meta’s decision to give away Muse Spark Open as an open-weight model, even while gating Muse Wonder Full behind paid subscriptions, mirrors the split strategy already playing out across the open-source AI field, where open models like Kimi K2, DeepSeek and Qwen3 compete heavily on price against closed frontier offerings.

The capital spending question is not unique to Meta either. Investors pushing back on AI infrastructure budgets has become a pattern across the sector this earnings season, visible in the pressure New York City’s pension fund comptroller put on Anthropic’s IPO plans over AI spending concerns. Meta’s Q3 report lands in the same stretch of the calendar as other major tech earnings, and the market’s reaction to Meta’s AI capex figure is likely to set a tone for how harshly investors judge other companies’ AI spending in their own earnings previews this quarter.

Historical Context: How Meta Got Here

It is worth remembering how fast this shift happened. Meta spent much of 2023 through 2025 building its AI identity around Llama, an open-weight model family that won developer loyalty but never directly generated subscription revenue. Zuckerberg’s own language at Meta’s Q1 2026 earnings, when he said the company had released its first model from the newly formed Meta Superintelligence Labs and was pursuing what he called “personal superintelligence,” marked the pivot point. Muse Spark followed in April, Muse Image in July, and the full Muse assistant product in September, each release narrowing the gap between Meta’s AI research output and an actual revenue-generating product.

That is a remarkably compressed timeline for a company Meta’s size. Google, OpenAI and Anthropic each took longer to move from frontier research models to paid consumer agent products with business connectors. Meta’s willingness to compress that cycle, launching Muse, then Muse for Small Business, then Muse Wonder, then Muse Spark Open, then a new enterprise executive hire, all within about five weeks, is itself a signal of how much pressure Meta feels to show AI revenue before the next earnings cycle after this one.

Market Reaction and Analyst Sentiment Heading Into Oct. 21

Sentiment going into the report is split. Meta’s AI spending and the uncertainty around Muse’s true adoption rate were cited by the BBC as a source of growing investor frustration even as the company’s stock enjoyed strong gains tied to the September Muse launch. At the same time, Yahoo Finance characterized Meta’s recent run as one of its best quarters in more than two years on a share-price basis, suggesting markets have already priced in substantial optimism about Muse’s prospects, optimism that the October 21 numbers will either confirm or puncture.

That gap, between a stock move that already assumes Muse succeeds and adoption data that nobody outside Meta can actually confirm, is the single biggest risk factor into this earnings report. If Meta’s disclosed numbers fall meaningfully short of the unconfirmed third-party estimates floating around, the stock reaction could be sharper than the underlying revenue beat or miss would normally justify.

What to Watch For on the Call

Whether Meta discloses any official Muse user, download, or subscriber figure, something it has not done publicly to date
Whether reported Q3 capital expenditure lands near, above, or below the $41.68 billion analyst estimate
Whether free cash flow stabilizes after Q2’s drop to $784 million or continues falling toward the roughly $8.52 billion burn some analysts are modeling
Whether Meta raises, holds, or narrows its $130 billion to $145 billion full-year capex guidance
Any commentary on the Muse sharing-feature pause and what it means for the Muse for Small Business rollout
How Zuckerberg frames the Meta Enterprise Platform and CJ Desai’s new role relative to Muse’s business ambitions

Predictions: Where This Goes From Here

Meta will likely report capital expenditure near or slightly above the $41.68 billion estimate, given the pace of Muse Wonder, Muse Spark Open, and the new enterprise platform all shipping within the same quarter.
Meta will probably continue declining to publish a precise Muse user count on the call itself, instead offering directional language about engagement and growth, leaving third-party estimates as the only public data points for at least another quarter.
Expect Meta to lean on the Muse for Small Business connectors and the new Meta Enterprise Platform as the monetization story, since enterprise contracts are easier to point to than consumer subscription churn that Meta has not disclosed.
The Muse privacy episode will likely prompt at least one direct analyst question on the call about data handling and enterprise trust, given how directly it touches the business connectors Meta is now selling.
Regardless of the Q3 numbers, Meta is unlikely to slow its release cadence. A Muse Wonder successor or a Muse Spark Open update is a reasonable bet before the next earnings cycle, continuing the pattern set since April.

Frequently Asked Questions

When does Meta report Q3 2026 earnings?
Meta confirmed on October 12 that its third-quarter 2026 results will be released after market close on Wednesday, October 21, 2026, followed by a conference call the same day.

How much is Meta expected to spend on AI in Q3 2026?
Analysts tracked by Visible Alpha estimate roughly $41.68 billion in capital expenditure for the quarter, more than double the year-earlier level. That is an estimate, not a confirmed Meta figure, and sits inside Meta’s full-year guidance of $130 billion to $145 billion.

What is Muse?
Muse is Meta’s personal AI agent, introduced in September 2026 and internally known as Hatch. It is designed to complete tasks such as sending emails, helping sell assets, and booking travel, with user approval required before it publishes, sends, or spends anything.

How much does Muse cost?
Muse offers a free tier with usage limits, a Power plan at $20 per month with 500 million Muse tokens per week, and a Maximum plan at $100 per month with 3 billion Muse tokens per week, according to Meta’s own subscription help documentation.

What are Muse Wonder and Muse Spark Open?
Muse Wonder and Muse Wonder Full are Meta’s newest frontier AI models, introduced October 12, built for agentic tasks and gated behind paid Muse and Muse Code plans. Muse Spark Open, released October 15, is a separate open-weight model with 816 billion total parameters and a 1-million-token context window, available for free.

How many people actually use Muse?
Meta has not published an official figure. Third-party estimates disagree significantly, ranging from roughly 3 million weekly users and 5 million downloads reported by Yahoo Finance to separate analyst estimates of 1.3 million to 2.5 million daily active users, so any specific adoption number should be treated as unconfirmed until Meta discloses its own data.

Is Muse coming to Meta’s AI glasses?
Meta announced in September that Muse would come to its AI glasses, including the newly announced Ray-Ban Meta Audio line, in the coming months. As of October 20, that integration had not yet shipped to the public.

Was there a privacy issue with Muse?
Yes. Ars Technica and The Verge both reported that Meta paused a conversation-sharing feature in Muse after researchers found shared AI chats, including personal and business information, exposed on the open web.

Meta Q3 Earnings: $41.68B AI Spend, Muse Test [2026]