{"content_id":"w3y85hbxu6","slug":"meta-stock-september-2026-ai-monetization","locale":"en","schema_type":"Article","category":"report","category_name":"Report","title":"Meta Stock and AI Monetization in September 2026","summary":"Meta's September stock rebound is linked to expectations for consumer AI subscriptions. Muse and Meta One expanded direct payment channels, but their profit contribution must be assessed based on paid conversion and service operating costs.","sponsorship_disclosure":null,"affiliate_disclosure":null,"commerce_disclosure":null,"author":{"name":"Injoys Editorial Team","url":"https://injoys.com/ko/about"},"key_points":["In September 2026, Meta announced Muse and Meta One, laying out clearer paths for charging consumers for AI.","Meta's capital expenditures were $31.08 billion in Q2 2026, including principal payments on finance leases.","Meta is also monetizing AI in its existing advertising business.","Free app rankings and subscription figures that include trials do not represent the number of paying customers.","MTIA's cost savings must be assessed based on actual deployment and operating costs by service."],"content_markdown":"Meta's September rebound was driven by expectations for consumer AI subscriptions. Muse and Meta One expanded direct payment channels. However, product launches and investment returns should be considered separately.\n\nThe figures are based on disclosures for the second quarter of 2026 and announcements and reports from March and September of the same year.\n\n## What Changed in September?\n\nWhat changed was that consumers gained specific ways to pay for AI. Meta announced Muse, a personal AI agent, on September 8, 2026. On September 15, it unveiled the subscription options for Meta One.\n\nMuse's early popularity was also linked to expectations for the stock price. On September 18, Axios reported that it ranked No. 1 among free iPhone apps in the United States. This ranking came ten days after launch. [Axios report on Muse's early usage trends](https://www.axios.com/2026/09/18/meta-muse-personal-agent-make-money)\n\nPetaPixel reported on September 21 that Meta's stock had risen more than 11%. However, the monthly increase of about 30% requires separate price verification. The claim that it ranked first in returns among the M7 must also be checked by comparing returns over the same period. [PetaPixel's September 22 stock price report](https://petapixel.com/2026/09/22/metas-stock-price-surges-as-muse-ai-tops-app-store-charts/)\n\nM7 is a term for a group of large U.S. technology stocks. The companies are Apple, Microsoft, Alphabet, and Amazon. Meta, NVIDIA, and Tesla are also included. Return rankings vary depending on the trading dates and whether dividends are included.\n\n## Why AI Investment Appeared Burdensome\n\nCapital expenditures accounted for most of the cash generated by operations. CAPEX is money spent on long-term assets such as servers and data centers. Meta's disclosed CAPEX also includes principal payments on finance leases.\n\n| Metric | Q2 2026 amount | Meaning |\n|---|---:|---|\n| Cash flow from operating activities | $31.86 billion | Net cash generated by operating activities |\n| CAPEX | $31.08 billion | Capital expenditures, including principal payments on finance leases |\n| Free cash flow | $784 million | The company's disclosed measure of cash flow after investment |\n\nFree cash flow is different from net income. Meta calculates it by subtracting purchases of property and equipment and principal payments on finance leases from operating cash flow. The amounts in the table are the disclosed figures from the earnings release. [Meta's Q2 2026 earnings release](https://investor.atmeta.com/investor-news/press-release-details/2026/Meta-Reports-Second-Quarter-2026-Results/)\n\nMeta issued $25 billion in corporate bonds in May 2026. It did not repurchase shares during the first half of the same year. However, the absence of repurchases is different from ending the program. At the end of June, some of the repurchase authorization approved by the board still remained. [Meta's Q2 2026 Form 10-Q](https://www.sec.gov/Archives/edgar/data/1326801/000162828026050705/meta-20260630.htm)\n\n## Did Existing AI Investments Produce No Results?\n\nAI monetization is also progressing in the existing advertising business. During its second-quarter earnings call, Meta explained improvements in ad prediction and ranking. It also discussed work to incorporate language models into recommendation systems. These are the results of using AI in the existing business as described by the company. [Transcript of Meta's Q2 2026 earnings call](https://s21.q4cdn.com/399680738/files/doc_financials/2026/q2/META-Q2-2026-Earnings-Call-Transcript.pdf)\n\nThe September changes can therefore be seen as an expansion of monetization channels. In advertising, advertisers pay the costs. With consumer subscriptions, users pay for additional features and usage. However, the existence of new products alone does not prove that the investment has been recovered.\n\n## Muse and Meta One Compared\n\nMuse focuses on performing tasks, while Meta One focuses on paid features in existing apps. Both products connect AI use to paid revenue. However, users pay for them for different reasons.\n\n| Category | Muse | Meta One |\n|---|---|---|\n| Main function | Performing tasks on behalf of users | Additional features in existing apps and expanded AI usage |\n| How it is used | Works in a dedicated virtual computer and browser | Uses features in Facebook, Instagram, WhatsApp, and Meta AI |\n| Main basis for charges | More agent usage | Paid options for individual apps and bundled subscriptions |\n| Performance metrics | Task completion, repeat use, and paid conversion | Use of paid features and subscription retention |\n\nMeta's announcement for Muse explains its dedicated virtual computer architecture. Users decide the scope of access allowed for the agent. Subscriptions are also available for additional usage. [Meta's September 8, 2026 Muse announcement](https://about.fb.com/news/2026/09/introducing-muse-personal-ai-agent/)\n\nAt launch, Muse Power cost $20 per month. Maximum cost $100 per month. Both prices are in U.S. dollars. [TechCrunch's September 8, 2026 launch report](https://techcrunch.com/2026/09/08/meta-debuts-its-muse-ai-agent-will-consumers-trust-it/)\n\nMeta One is not a product with the same package for every user. It is divided into subscriptions for individual apps and bundled subscriptions. Prices and benefits may vary by region, app, and account. [Meta's September 15, 2026 Meta One announcement](https://about.fb.com/news/2026/09/introducing-meta-one-subscription-service-more-features-ai/)\n\n## The Difference Between Subscription Counts and Paid User Counts\n\nA figure combining subscriptions and trials is different from the actual number of paying customers. In its September 15 announcement, Meta reported a cumulative total of 15 million. This figure combines subscriptions and trials. Counting all of them as paid users could inflate revenue.\n\nThe same announcement explains that subscriptions for individual apps were launched at the beginning of the year. The cumulative figure therefore cannot be viewed as new sign-ups after the September announcement. Subscription counts must also be distinguished from the number of people.\n\n| Reported metric | What it shows | What it does not immediately show |\n|---|---|---|\n| Free app ranking | App popularity at a specific point in time | Number of paying customers and revenue |\n| Combined subscription and trial count | Cumulative participation in the product | Paid retention rate after trials end |\n| Paid subscription revenue | Actual amount charged | Profit after operating costs |\n\nThis distinction is necessary when calculating the value of the AI business. Multiplying a count that includes trials by the monthly fee does not produce actual revenue. The price and billing period for each product must also be checked.\n\n## Does Profit Also Increase When Free Usage Grows?\n\nAn increase in free users alone is not enough to determine whether profit is rising. Muse performs tasks using a browser and tools. These tasks require computing resources. The conclusion that computing resources are needed is based on the service's structure.\n\nProcessing costs may rise along with subscription revenue. Conversely, lower costs per task can help improve profitability. Revenue from paid users alone does not account for the cost of serving free users.\n\nThe following metrics should be considered together when assessing profitability.\n\n- Paid conversion rate: The percentage of free users who pay\n- Subscription retention rate: The percentage who remain through the next billing period\n- Revenue per paid user: Revenue earned from each customer\n- Processing cost per task: Computing costs required to complete a task\n- Cost of free usage: Operating costs for supporting nonpaying users\n\n## Does MTIA Solve the Cost Problem?\n\nMTIA is a way to improve costs, but its cost-saving effect has not been confirmed. MTIA is a family of AI accelerators developed by Meta. Inference is the computation through which a trained model produces an answer or judgment. Meta is designing MTIA 450 and 500 for generative AI inference.\n\n| Product | Mass deployment plan in the official announcement | Distinction to make when interpreting it |\n|---|---|---|\n| MTIA 450 | Early 2027 | This is a planned schedule, and actual deployment must be confirmed separately |\n| MTIA 500 | Second half of 2027 | The announcement does not limit it to the second half of the year |\n\nThis schedule is based on the technical announcement dated March 11, 2026. Details on each product are available in Meta AI's MTIA technical documentation. Separate evidence is needed for the cost reduction rate or the share dedicated to Muse. [Meta AI's MTIA development and deployment plan](https://ai.meta.com/blog/meta-mtia-scale-ai-chips-for-billions/)\n\nCustom chips do not eliminate all service costs. Resources such as browser execution and storage are also needed. Chip performance and the profitability of the overall service must therefore be evaluated separately.\n\n## Common Mistakes When Reading Stock Prices and Earnings\n\nCaution is needed when interpreting product popularity as proof that an investment has been recovered. Stock prices may reflect future expectations first. Actual performance is confirmed later through revenue and costs.\n\n| Common interpretation | Facts that must be distinguished |\n|---|---|\n| The launch of an AI product means the investment has been recovered | A launch and the recovery of cumulative investment costs are separate matters |\n| A high free app ranking means high profitability | Payment rates and operating costs must be checked |\n| No share repurchases mean the repurchase program has ended | Repurchase activity during a period and the authorized limit are separate matters |\n| A custom chip deployment plan is the same as actual cost savings | Completed deployment and actual costs must be checked |\n| The September rebound was caused solely by the new products | Broader market movements and other company-specific factors must also be reviewed |\n\nUse the same basis when comparing stock returns.\n\n1. Check the closing price on the last trading day of August 2026.\n2. Check the closing price on the comparison end date using the same price data.\n3. Divide the ending closing price by the starting closing price, then subtract 1.\n4. Multiply the result by 100 to calculate the return for the period.\n5. Apply the same trading dates and dividend treatment to the entire M7.\n\nFor year-to-date returns, the final trading day of the previous year is the starting point. The claim that the stock had fallen more than 12% through the end of August must also be checked using this basis. A return without a reference price is difficult to cite as a confirmed figure.","content_html":"\u003cp\u003eMeta's September rebound was driven by expectations for consumer AI subscriptions. Muse and Meta One expanded direct payment channels. However, product launches and investment returns should be considered separately.\u003c/p\u003e\n\u003cp\u003eThe figures are based on disclosures for the second quarter of 2026 and announcements and reports from March and September of the same year.\u003c/p\u003e\n\u003ch2\u003e\n\u003ca href=\"#what-changed-in-september\" class=\"anchor\" id=\"what-changed-in-september\"\u003e\u003c/a\u003eWhat Changed in September?\u003c/h2\u003e\n\u003cp\u003eWhat changed was that consumers gained specific ways to pay for AI. Meta announced Muse, a personal AI agent, on September 8, 2026. On September 15, it unveiled the subscription options for Meta One.\u003c/p\u003e\n\u003cp\u003eMuse's early popularity was also linked to expectations for the stock price. On September 18, Axios reported that it ranked No. 1 among free iPhone apps in the United States. This ranking came ten days after launch. \u003ca href=\"https://www.axios.com/2026/09/18/meta-muse-personal-agent-make-money\"\u003eAxios report on Muse's early usage trends\u003c/a\u003e\u003c/p\u003e\n\u003cp\u003ePetaPixel reported on September 21 that Meta's stock had risen more than 11%. However, the monthly increase of about 30% requires separate price verification. The claim that it ranked first in returns among the M7 must also be checked by comparing returns over the same period. \u003ca href=\"https://petapixel.com/2026/09/22/metas-stock-price-surges-as-muse-ai-tops-app-store-charts/\"\u003ePetaPixel's September 22 stock price report\u003c/a\u003e\u003c/p\u003e\n\u003cp\u003eM7 is a term for a group of large U.S. technology stocks. The companies are Apple, Microsoft, Alphabet, and Amazon. Meta, NVIDIA, and Tesla are also included. Return rankings vary depending on the trading dates and whether dividends are included.\u003c/p\u003e\n\u003ch2\u003e\n\u003ca href=\"#why-ai-investment-appeared-burdensome\" class=\"anchor\" id=\"why-ai-investment-appeared-burdensome\"\u003e\u003c/a\u003eWhy AI Investment Appeared Burdensome\u003c/h2\u003e\n\u003cp\u003eCapital expenditures accounted for most of the cash generated by operations. CAPEX is money spent on long-term assets such as servers and data centers. Meta's disclosed CAPEX also includes principal payments on finance leases.\u003c/p\u003e\n\u003cdiv class=\"overflow-x-auto\"\u003e\u003ctable\u003e\n\u003cthead\u003e\n\u003ctr\u003e\n\u003cth\u003eMetric\u003c/th\u003e\n\u003cth\u003eQ2 2026 amount\u003c/th\u003e\n\u003cth\u003eMeaning\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Metric\"\u003eCash flow from operating activities\u003c/td\u003e\n\u003ctd data-label=\"Q2 2026 amount\"\u003e$31.86 billion\u003c/td\u003e\n\u003ctd data-label=\"Meaning\"\u003eNet cash generated by operating activities\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Metric\"\u003eCAPEX\u003c/td\u003e\n\u003ctd data-label=\"Q2 2026 amount\"\u003e$31.08 billion\u003c/td\u003e\n\u003ctd data-label=\"Meaning\"\u003eCapital expenditures, including principal payments on finance leases\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Metric\"\u003eFree cash flow\u003c/td\u003e\n\u003ctd data-label=\"Q2 2026 amount\"\u003e$784 million\u003c/td\u003e\n\u003ctd data-label=\"Meaning\"\u003eThe company's disclosed measure of cash flow after investment\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\u003c/div\u003e\n\u003cp\u003eFree cash flow is different from net income. Meta calculates it by subtracting purchases of property and equipment and principal payments on finance leases from operating cash flow. The amounts in the table are the disclosed figures from the earnings release. \u003ca href=\"https://investor.atmeta.com/investor-news/press-release-details/2026/Meta-Reports-Second-Quarter-2026-Results/\"\u003eMeta's Q2 2026 earnings release\u003c/a\u003e\u003c/p\u003e\n\u003cp\u003eMeta issued $25 billion in corporate bonds in May 2026. It did not repurchase shares during the first half of the same year. However, the absence of repurchases is different from ending the program. At the end of June, some of the repurchase authorization approved by the board still remained. \u003ca href=\"https://www.sec.gov/Archives/edgar/data/1326801/000162828026050705/meta-20260630.htm\"\u003eMeta's Q2 2026 Form 10-Q\u003c/a\u003e\u003c/p\u003e\n\u003ch2\u003e\n\u003ca href=\"#did-existing-ai-investments-produce-no-results\" class=\"anchor\" id=\"did-existing-ai-investments-produce-no-results\"\u003e\u003c/a\u003eDid Existing AI Investments Produce No Results?\u003c/h2\u003e\n\u003cp\u003eAI monetization is also progressing in the existing advertising business. During its second-quarter earnings call, Meta explained improvements in ad prediction and ranking. It also discussed work to incorporate language models into recommendation systems. These are the results of using AI in the existing business as described by the company. \u003ca href=\"https://s21.q4cdn.com/399680738/files/doc_financials/2026/q2/META-Q2-2026-Earnings-Call-Transcript.pdf\"\u003eTranscript of Meta's Q2 2026 earnings call\u003c/a\u003e\u003c/p\u003e\n\u003cp\u003eThe September changes can therefore be seen as an expansion of monetization channels. In advertising, advertisers pay the costs. With consumer subscriptions, users pay for additional features and usage. However, the existence of new products alone does not prove that the investment has been recovered.\u003c/p\u003e\n\u003ch2\u003e\n\u003ca href=\"#muse-and-meta-one-compared\" class=\"anchor\" id=\"muse-and-meta-one-compared\"\u003e\u003c/a\u003eMuse and Meta One Compared\u003c/h2\u003e\n\u003cp\u003eMuse focuses on performing tasks, while Meta One focuses on paid features in existing apps. Both products connect AI use to paid revenue. However, users pay for them for different reasons.\u003c/p\u003e\n\u003cdiv class=\"overflow-x-auto\"\u003e\u003ctable\u003e\n\u003cthead\u003e\n\u003ctr\u003e\n\u003cth\u003eCategory\u003c/th\u003e\n\u003cth\u003eMuse\u003c/th\u003e\n\u003cth\u003eMeta One\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Category\"\u003eMain function\u003c/td\u003e\n\u003ctd data-label=\"Muse\"\u003ePerforming tasks on behalf of users\u003c/td\u003e\n\u003ctd data-label=\"Meta One\"\u003eAdditional features in existing apps and expanded AI usage\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Category\"\u003eHow it is used\u003c/td\u003e\n\u003ctd data-label=\"Muse\"\u003eWorks in a dedicated virtual computer and browser\u003c/td\u003e\n\u003ctd data-label=\"Meta One\"\u003eUses features in Facebook, Instagram, WhatsApp, and Meta AI\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Category\"\u003eMain basis for charges\u003c/td\u003e\n\u003ctd data-label=\"Muse\"\u003eMore agent usage\u003c/td\u003e\n\u003ctd data-label=\"Meta One\"\u003ePaid options for individual apps and bundled subscriptions\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Category\"\u003ePerformance metrics\u003c/td\u003e\n\u003ctd data-label=\"Muse\"\u003eTask completion, repeat use, and paid conversion\u003c/td\u003e\n\u003ctd data-label=\"Meta One\"\u003eUse of paid features and subscription retention\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\u003c/div\u003e\n\u003cp\u003eMeta's announcement for Muse explains its dedicated virtual computer architecture. Users decide the scope of access allowed for the agent. Subscriptions are also available for additional usage. \u003ca href=\"https://about.fb.com/news/2026/09/introducing-muse-personal-ai-agent/\"\u003eMeta's September 8, 2026 Muse announcement\u003c/a\u003e\u003c/p\u003e\n\u003cp\u003eAt launch, Muse Power cost $20 per month. Maximum cost $100 per month. Both prices are in U.S. dollars. \u003ca href=\"https://techcrunch.com/2026/09/08/meta-debuts-its-muse-ai-agent-will-consumers-trust-it/\"\u003eTechCrunch's September 8, 2026 launch report\u003c/a\u003e\u003c/p\u003e\n\u003cp\u003eMeta One is not a product with the same package for every user. It is divided into subscriptions for individual apps and bundled subscriptions. Prices and benefits may vary by region, app, and account. \u003ca href=\"https://about.fb.com/news/2026/09/introducing-meta-one-subscription-service-more-features-ai/\"\u003eMeta's September 15, 2026 Meta One announcement\u003c/a\u003e\u003c/p\u003e\n\u003ch2\u003e\n\u003ca href=\"#the-difference-between-subscription-counts-and-paid-user-counts\" class=\"anchor\" id=\"the-difference-between-subscription-counts-and-paid-user-counts\"\u003e\u003c/a\u003eThe Difference Between Subscription Counts and Paid User Counts\u003c/h2\u003e\n\u003cp\u003eA figure combining subscriptions and trials is different from the actual number of paying customers. In its September 15 announcement, Meta reported a cumulative total of 15 million. This figure combines subscriptions and trials. Counting all of them as paid users could inflate revenue.\u003c/p\u003e\n\u003cp\u003eThe same announcement explains that subscriptions for individual apps were launched at the beginning of the year. The cumulative figure therefore cannot be viewed as new sign-ups after the September announcement. Subscription counts must also be distinguished from the number of people.\u003c/p\u003e\n\u003cdiv class=\"overflow-x-auto\"\u003e\u003ctable\u003e\n\u003cthead\u003e\n\u003ctr\u003e\n\u003cth\u003eReported metric\u003c/th\u003e\n\u003cth\u003eWhat it shows\u003c/th\u003e\n\u003cth\u003eWhat it does not immediately show\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Reported metric\"\u003eFree app ranking\u003c/td\u003e\n\u003ctd data-label=\"What it shows\"\u003eApp popularity at a specific point in time\u003c/td\u003e\n\u003ctd data-label=\"What it does not immediately show\"\u003eNumber of paying customers and revenue\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Reported metric\"\u003eCombined subscription and trial count\u003c/td\u003e\n\u003ctd data-label=\"What it shows\"\u003eCumulative participation in the product\u003c/td\u003e\n\u003ctd data-label=\"What it does not immediately show\"\u003ePaid retention rate after trials end\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Reported metric\"\u003ePaid subscription revenue\u003c/td\u003e\n\u003ctd data-label=\"What it shows\"\u003eActual amount charged\u003c/td\u003e\n\u003ctd data-label=\"What it does not immediately show\"\u003eProfit after operating costs\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\u003c/div\u003e\n\u003cp\u003eThis distinction is necessary when calculating the value of the AI business. Multiplying a count that includes trials by the monthly fee does not produce actual revenue. The price and billing period for each product must also be checked.\u003c/p\u003e\n\u003ch2\u003e\n\u003ca href=\"#does-profit-also-increase-when-free-usage-grows\" class=\"anchor\" id=\"does-profit-also-increase-when-free-usage-grows\"\u003e\u003c/a\u003eDoes Profit Also Increase When Free Usage Grows?\u003c/h2\u003e\n\u003cp\u003eAn increase in free users alone is not enough to determine whether profit is rising. Muse performs tasks using a browser and tools. These tasks require computing resources. The conclusion that computing resources are needed is based on the service's structure.\u003c/p\u003e\n\u003cp\u003eProcessing costs may rise along with subscription revenue. Conversely, lower costs per task can help improve profitability. Revenue from paid users alone does not account for the cost of serving free users.\u003c/p\u003e\n\u003cp\u003eThe following metrics should be considered together when assessing profitability.\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePaid conversion rate: The percentage of free users who pay\u003c/li\u003e\n\u003cli\u003eSubscription retention rate: The percentage who remain through the next billing period\u003c/li\u003e\n\u003cli\u003eRevenue per paid user: Revenue earned from each customer\u003c/li\u003e\n\u003cli\u003eProcessing cost per task: Computing costs required to complete a task\u003c/li\u003e\n\u003cli\u003eCost of free usage: Operating costs for supporting nonpaying users\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch2\u003e\n\u003ca href=\"#does-mtia-solve-the-cost-problem\" class=\"anchor\" id=\"does-mtia-solve-the-cost-problem\"\u003e\u003c/a\u003eDoes MTIA Solve the Cost Problem?\u003c/h2\u003e\n\u003cp\u003eMTIA is a way to improve costs, but its cost-saving effect has not been confirmed. MTIA is a family of AI accelerators developed by Meta. Inference is the computation through which a trained model produces an answer or judgment. Meta is designing MTIA 450 and 500 for generative AI inference.\u003c/p\u003e\n\u003cdiv class=\"overflow-x-auto\"\u003e\u003ctable\u003e\n\u003cthead\u003e\n\u003ctr\u003e\n\u003cth\u003eProduct\u003c/th\u003e\n\u003cth\u003eMass deployment plan in the official announcement\u003c/th\u003e\n\u003cth\u003eDistinction to make when interpreting it\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Product\"\u003eMTIA 450\u003c/td\u003e\n\u003ctd data-label=\"Mass deployment plan in the official announcement\"\u003eEarly 2027\u003c/td\u003e\n\u003ctd data-label=\"Distinction to make when interpreting it\"\u003eThis is a planned schedule, and actual deployment must be confirmed separately\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Product\"\u003eMTIA 500\u003c/td\u003e\n\u003ctd data-label=\"Mass deployment plan in the official announcement\"\u003eSecond half of 2027\u003c/td\u003e\n\u003ctd data-label=\"Distinction to make when interpreting it\"\u003eThe announcement does not limit it to the second half of the year\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\u003c/div\u003e\n\u003cp\u003eThis schedule is based on the technical announcement dated March 11, 2026. Details on each product are available in Meta AI's MTIA technical documentation. Separate evidence is needed for the cost reduction rate or the share dedicated to Muse. \u003ca href=\"https://ai.meta.com/blog/meta-mtia-scale-ai-chips-for-billions/\"\u003eMeta AI's MTIA development and deployment plan\u003c/a\u003e\u003c/p\u003e\n\u003cp\u003eCustom chips do not eliminate all service costs. Resources such as browser execution and storage are also needed. Chip performance and the profitability of the overall service must therefore be evaluated separately.\u003c/p\u003e\n\u003ch2\u003e\n\u003ca href=\"#common-mistakes-when-reading-stock-prices-and-earnings\" class=\"anchor\" id=\"common-mistakes-when-reading-stock-prices-and-earnings\"\u003e\u003c/a\u003eCommon Mistakes When Reading Stock Prices and Earnings\u003c/h2\u003e\n\u003cp\u003eCaution is needed when interpreting product popularity as proof that an investment has been recovered. Stock prices may reflect future expectations first. Actual performance is confirmed later through revenue and costs.\u003c/p\u003e\n\u003cdiv class=\"overflow-x-auto\"\u003e\u003ctable\u003e\n\u003cthead\u003e\n\u003ctr\u003e\n\u003cth\u003eCommon interpretation\u003c/th\u003e\n\u003cth\u003eFacts that must be distinguished\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Common interpretation\"\u003eThe launch of an AI product means the investment has been recovered\u003c/td\u003e\n\u003ctd data-label=\"Facts that must be distinguished\"\u003eA launch and the recovery of cumulative investment costs are separate matters\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Common interpretation\"\u003eA high free app ranking means high profitability\u003c/td\u003e\n\u003ctd data-label=\"Facts that must be distinguished\"\u003ePayment rates and operating costs must be checked\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Common interpretation\"\u003eNo share repurchases mean the repurchase program has ended\u003c/td\u003e\n\u003ctd data-label=\"Facts that must be distinguished\"\u003eRepurchase activity during a period and the authorized limit are separate matters\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Common interpretation\"\u003eA custom chip deployment plan is the same as actual cost savings\u003c/td\u003e\n\u003ctd data-label=\"Facts that must be distinguished\"\u003eCompleted deployment and actual costs must be checked\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Common interpretation\"\u003eThe September rebound was caused solely by the new products\u003c/td\u003e\n\u003ctd data-label=\"Facts that must be distinguished\"\u003eBroader market movements and other company-specific factors must also be reviewed\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\u003c/div\u003e\n\u003cp\u003eUse the same basis when comparing stock returns.\u003c/p\u003e\n\u003col\u003e\n\u003cli\u003eCheck the closing price on the last trading day of August 2026.\u003c/li\u003e\n\u003cli\u003eCheck the closing price on the comparison end date using the same price data.\u003c/li\u003e\n\u003cli\u003eDivide the ending closing price by the starting closing price, then subtract 1.\u003c/li\u003e\n\u003cli\u003eMultiply the result by 100 to calculate the return for the period.\u003c/li\u003e\n\u003cli\u003eApply the same trading dates and dividend treatment to the entire M7.\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eFor year-to-date returns, the final trading day of the previous year is the starting point. The claim that the stock had fallen more than 12% through the end of August must also be checked using this basis. A return without a reference price is difficult to cite as a confirmed figure.\u003c/p\u003e\n","tags":["Generative AI","AI Agents","AI chips","Investor psychology","Platform Business"],"faqs":[{"question":"What is Meta's September stock price rebound linked to?","answer":"It is linked to Muse's strong initial performance and expectations for consumer AI subscriptions. However, the new product alone cannot explain the entire stock price movement."},{"question":"Are Muse and Meta One the same product?","answer":"They are different products. Muse is an agent that performs personal tasks on the user's behalf. Meta One is a subscription system that provides add-on features for existing apps and AI usage."},{"question":"Was Meta unable to make money from AI before launching Muse?","answer":"That cannot be stated conclusively. Meta also uses AI in its existing ad prediction and recommendation systems. Consumer subscriptions are an additional path beyond its existing advertising monetization."},{"question":"How much was Meta's capital expenditure in the second quarter of 2026?","answer":"It was $31.08 billion, including principal repayments on finance leases. Cash flow from operating activities in the same quarter was $31.86 billion."},{"question":"Does being No. 1 in the free app rankings mean having the largest number of paying customers?","answer":"The number of paying customers cannot be determined from the free app rankings alone. The actual payment volume and subscription retention rate must be checked separately."},{"question":"Are all of Meta One's cumulative 15 million paid subscriptions?","answer":"The figure announced on September 15, 2026 combines subscriptions and trials. Counting all of them as paying customers could overstate actual revenue."},{"question":"When will MTIA 450 and MTIA 500 be deployed?","answer":"According to the official plan from March 11, 2026, MTIA 450 is scheduled for early 2027. MTIA 500 is scheduled for large-scale deployment in 2027. Plans must be distinguished from the actual completion of deployment."},{"question":"How is the profitability of AI subscriptions assessed?","answer":"The paid conversion rate and subscription retention rate should be checked first. Revenue per user and processing cost per task should also be considered. The cost of supporting free users also affects profitability."},{"question":"Has Meta's share repurchase program been discontinued?","answer":"It is confirmed that no share repurchases were made in the first half of 2026. However, authorized repurchase capacity remained at the end of June. Not executing repurchases and discontinuing the program are different facts."}],"sources":[{"url":"https://about.fb.com/news/2026/09/introducing-muse-personal-ai-agent/","title":"Meta, Introducing Muse: The World’s First Personal AI Agent Built for Everyone, September 8, 2026","type":"source"},{"url":"https://about.fb.com/news/2026/09/introducing-meta-one-subscription-service-more-features-ai/","title":"Meta, Introducing Meta One: A Subscription Service With More Features and AI to Create, Connect, and Stand Out, September 15, 2026","type":"source"},{"url":"https://investor.atmeta.com/investor-news/press-release-details/2026/Meta-Reports-Second-Quarter-2026-Results/","title":"Meta Reports Second Quarter 2026 Results","type":"data_point"},{"url":"https://www.sec.gov/Archives/edgar/data/1326801/000162828026050705/meta-20260630.htm","title":"Meta Platforms, Form 10-Q for the Quarter Ended June 30, 2026","type":"data_point"},{"url":"https://s21.q4cdn.com/399680738/files/doc_financials/2026/q2/META-Q2-2026-Earnings-Call-Transcript.pdf","title":"Meta, Q2 2026 Earnings Call Transcript","type":"source"},{"url":"https://ai.meta.com/blog/meta-mtia-scale-ai-chips-for-billions/","title":"Meta AI, Four MTIA Chips in Two Years: Scaling AI Experiences for Billions, March 11, 2026","type":"source"},{"url":"https://www.axios.com/2026/09/18/meta-muse-personal-agent-make-money","title":"Axios, Meta AI gains momentum with Zuckerberg's agent for the masses, September 18, 2026","type":"data_point"},{"url":"https://techcrunch.com/2026/09/08/meta-debuts-its-muse-ai-agent-will-consumers-trust-it/","title":"TechCrunch, Meta debuts its Muse AI agent. Will consumers trust it?, September 8, 2026","type":"source"},{"url":"https://petapixel.com/2026/09/22/metas-stock-price-surges-as-muse-ai-tops-app-store-charts/","title":"PetaPixel, Meta’s Stock Price Surges as Muse AI Tops App Store Charts, September 22, 2026","type":"data_point"}],"images":[{"id":1475,"url":"https://injoys.com/rails/active_storage/blobs/proxy/eyJfcmFpbHMiOnsiZGF0YSI6MjE0NDksInB1ciI6ImJsb2JfaWQifX0=--6bcac5b2184b975f5a3d1e28c97b26f7ab39dcf3/ai-df03724f.webp","is_representative":true,"generation_method":"ai_photo","license":"ai_generated","mime_type":"image/webp","width":1536,"height":1024,"translations":{"ko":{"alt":"카페 창가에서 한 남성이 태블릿의 주가 그래프에 손가락을 댄 채 재무 자료를 살펴본다.","caption":"AI 구독에 대한 기대가 실제 이익으로 이어지는지는 유료 전환과 운영비로 확인해야 합니다.","description":null},"en":{"alt":"At a café window, a man pauses with his finger over a tablet chart beside a financial worksheet.","caption":"Whether AI subscription hopes become profits depends on paid conversions and operating costs.","description":null},"ja":{"alt":"カフェの窓際で、男性が財務資料を見ながらタブレットの株価チャートに指をかざしている。","caption":"AIサブスクへの期待が利益につながるかは、有料転換と運営費で見極める必要がある。","description":null},"es":{"alt":"Junto a la ventana de una cafetería, un hombre observa una hoja financiera con el dedo sobre una gráfica en una tableta.","caption":"Que las suscripciones de IA generen beneficios depende de la conversión a planes de pago y los costes operativos.","description":null},"id":{"alt":"Di dekat jendela kafe, seorang pria menatap lembar keuangan sambil menahan jari di atas grafik pada tablet.","caption":"Harapan laba dari langganan AI perlu diuji lewat konversi berbayar dan biaya operasional.","description":null},"pt":{"alt":"À janela de um café, um homem observa uma folha financeira com o dedo sobre um gráfico num tablet.","caption":"O lucro esperado das assinaturas de IA depende da conversão em clientes pagantes e dos custos operacionais.","description":null},"zh-hant":{"alt":"咖啡館窗邊，一名男子看著財務資料，手指停在平板電腦的股價圖表上方。","caption":"AI 訂閱能否帶來利潤，仍須看付費轉換與營運成本。","description":null},"de":{"alt":"Am Caféfenster betrachtet ein Mann Finanzunterlagen, während sein Finger über einem Diagramm auf dem Tablet schwebt.","caption":"Ob KI-Abos Gewinne bringen, hängt von zahlenden Kunden und den Betriebskosten ab.","description":null}}},{"id":1476,"url":"https://injoys.com/rails/active_storage/blobs/proxy/eyJfcmFpbHMiOnsiZGF0YSI6MjE0NTksInB1ciI6ImJsb2JfaWQifX0=--543cc016fc4e83e1b373c67a8af95627f788e2cb/ai-60a0c596.webp","is_representative":false,"generation_method":"ai_semi","license":"ai_generated","mime_type":"image/webp","width":1536,"height":1024,"translations":{"ko":{"alt":"데이터센터에서 엔지니어가 서버랙의 가속기 트레이를 잡고 있고, 옆 카트의 태블릿에는 차트가 표시돼 있다.","caption":"MTIA의 비용 절감 효과는 실제 배치와 서비스별 운영비로 확인해야 합니다.","description":null},"en":{"alt":"An engineer steadies an accelerator tray in a server rack, with charts on a tablet atop a nearby cart.","caption":"MTIA cost savings must be verified through actual deployment and operating costs by service.","description":null},"ja":{"alt":"データセンターで技術者がサーバーラックのアクセラレータートレイを支え、近くの台車のタブレットにはグラフが映る。","caption":"MTIAのコスト削減効果は、実際の導入とサービス別の運用費で確かめる必要があります。","description":null},"es":{"alt":"Un técnico sujeta una bandeja de aceleradores en un rack; una tableta cercana muestra gráficos.","caption":"El ahorro de costes de MTIA debe comprobarse con su despliegue real y los costes operativos de cada servicio.","description":null},"id":{"alt":"Teknisi menahan baki akselerator di rak server, sementara tablet di troli dekatnya menampilkan grafik.","caption":"Penghematan biaya MTIA perlu dibuktikan lewat penerapan nyata dan biaya operasional tiap layanan.","description":null},"pt":{"alt":"Um técnico segura uma bandeja de aceleradores num rack de servidores; um tablet próximo exibe gráficos.","caption":"A redução de custos com MTIA precisa ser comprovada pela implantação real e pelos custos operacionais de cada serviço.","description":null},"zh-hant":{"alt":"工程師在資料中心扶著伺服器機櫃中的加速器托盤，旁邊推車上的平板顯示圖表。","caption":"MTIA的成本節省效果，仍須透過實際部署及各項服務的營運成本來驗證。","description":null},"de":{"alt":"Ein Techniker hält einen Beschleuniger-Einschub im Serverrack; ein Tablet auf einem Wagen zeigt Diagramme.","caption":"Ob MTIA Kosten spart, muss sich im tatsächlichen Einsatz und an den Betriebskosten der einzelnen Dienste zeigen.","description":null}}}],"published_at":"2026-09-23T19:55:16+09:00","updated_at":"2026-09-23T19:55:16+09:00","license":"cc_by","translation_status":"reviewed","available_locales":["ko","en","ja","es"],"data_locales":["ko","en","ja","es","id","pt","zh-hant","de"],"url":"https://injoys.com/en/articles/meta-stock-september-2026-ai-monetization"}