AI agent adoption may be constrained by compute bottlenecks; Morgan Stanley remains bullish on Meta's (META.US) long-term commercialization prospects, raises target price to $820 and maintains "Overweight" rating.

Morgan Stanley's latest research report points out that as AI agent applications gradually become widespread, the GPU computing power required to support their operation may become an important factor constraining user scale expansion. The bank estimates that for every additional 100 million daily active users of Meta Platforms' (META.US) AI agent Muse, approximately 1 to 4 gigawatts (GW) of GPU computing capacity may be required. Although the massive infrastructure demand means AI agent adoption may be slower than market expectations, Morgan Stanley remains bullish on Meta's long-term commercialization potential, raising its target price from $775 to $820 and maintaining an "Overweight" rating. In a report published on October 18, Morgan Stanley stated that the market is currently focused on Muse's CPU resource requirements, but as agent usage scales up, GPUs actually consume more power, making them more likely to become the key constraint determining product expansion speed. Since the specific parameter scale, interaction frequency, and token processing speed of the Spark and Wonder models behind Muse remain unclear, the bank conducted sensitivity analysis across different scenarios. Assuming 100 million daily active users use an average of 25% of their weekly free token allowance, Muse's required GPU computing capacity could reach approximately 1 to 4 gigawatts, depending on the model's token processing efficiency per second. Morgan Stanley noted that model inference efficiency will directly affect the infrastructure costs of AI agent services. If frontier models with massive parameter scales are adopted, computing demand will be significantly higher than in scenarios using lightweight models. Different agent tasks may also be handled by models of different scales, so actual demand depends on Meta's future model architecture and task allocation approach. This estimate implies that AI agents still face considerable infrastructure challenges in achieving scaled application with hundreds of millions or even billions of users. Morgan Stanley expects Meta to add approximately 13 gigawatts of total computing capacity between 2027 and 2028 to support its various AI products and services. However, under the assumption that each GPU processes approximately 10,000 tokens per second, if Muse has 500 million daily active users and each person uses 50 million tokens per week, the corresponding service could require approximately 13 gigawatts of computing capacityequivalent to Meta's entire expected new computing capacity over the next two years. This means that even if Meta continues large-scale investment in data centers, the rapid adoption of a single agent product could consume substantial new computing resources, thereby limiting development space for other AI businesses. Morgan Stanley believes that improving model efficiency and reducing per-user service costs will be key to driving agent adoption. Tech giants such as Alphabet Inc. Class C (GOOGL.US) and Amazon.com, Inc. (AMZN.US) similarly need to balance computing investment against return on investment, while smaller competitors may face greater pressure to expand infrastructure. At the same time, the bank reminds investors that AI agent user growth curves may be longer than expected, and may even flatten at certain stages. The report uses the development trajectory of e-commerce as a reference, noting that in 2000, when e-commerce reached its fifth year of development, its share of adjusted retail sales was only about 4%; even by 2026, this proportion will only be about 28%. Although AI agents do not need to build physical logistics infrastructure on the same scale as e-commerce, they still require continuous increases in computing capacity, development of more efficient models, and efforts to drive enterprises to change existing business systems and user interaction methods. However, massive computing demand also means global AI capital expenditure may remain elevated for years to come. Morgan Stanley noted that chip and memory capacity available to major tech companies in 2027 has essentially been locked up in advance. Although AI capital expenditure annual growth may gradually slow due to factors such as base expansion, supply chain constraints, and actual construction capability, long-term expansion demand from agent applications is still expected to support sustained high levels of investment over several years. From a commercialization perspective, Morgan Stanley believes AI agents still have enormous long-term market space. The bank estimates that approximately $30 trillion in global consumer spending could potentially be digitized through agent services, of which about 55% is related to e-commerce and travel. As AI agents gradually acquire capabilities such as product search, shopping decisions, travel planning, and transaction execution, the related technology is expected to transform how consumers and businesses interact. To achieve this, enterprises will also need to upgrade existing internet systems and establish technical architectures that support secure interaction and transactions between agents. Morgan Stanley believes this will create business opportunities for Meta's developing Meta Enterprise Platform, enabling enterprise customers to access agent-related tools through the platform. In terms of near-term product performance, Muse still faces challenges with insufficient user retention. According to Sensor Tower data and Morgan Stanley's analysis, Muse currently still gains approximately 300,000 new downloads per day in the United States, but daily active users are only about 2.5 million, with 14-day post-install retention at approximately 20%significantly lower than mature applications such as Instagram, Facebook, and Google Search. Morgan Stanley believes this reflects that while Muse can attract users to try it, it has not yet established sufficiently stable daily usage habits. Whether it can add practical use cases, launch more convenient app connection features, and support commercial activities such as shopping and travel planning will become important factors driving daily active user growth. Although Muse's commercialization remains at an early stage, Morgan Stanley is more optimistic about Meta's core advertising business and emerging AI revenue opportunities. The bank raised Meta's target price from $775 to $820, which based on the October 16 closing price of $709.33 implies approximately 15.6% upside potential, and maintained its "Overweight" rating. The bank stated that the target price increase is mainly based on improved performance in Meta's core advertising business and the incremental revenue that Muse and enterprise AI subscription services may bring in the future. The bank raised Meta's 2027 and 2028 revenue forecasts by approximately 1% and 2%, respectively, and raised GAAP earnings per share forecasts for the same periods by approximately 1% and 3%, respectively, to $34.04 and $35.76. Among these, AI-driven advertising products are expected to continue improving advertisers' return on investment and drive Meta's advertising revenue growth. Morgan Stanley raised 2027 and 2028 advertising revenue forecasts by approximately 1% each, while also raising payment and other business revenue expectations to reflect the potential contribution from Muse and enterprise AI subscription services. Although Meta is expected to increase sales and related expenses to promote its enterprise AI platform, Morgan Stanley believes the company remains capable of achieving profit growth while maintaining large-scale AI investment. The bank currently expects Meta's total revenue in 2027 and 2028 to reach approximately $316.1 billion and $381.8 billion, respectively, with GAAP earnings per share of approximately $34 and $36. In terms of valuation, the $820 target price corresponds to a P/E ratio of approximately 23 times Meta's average expected earnings per share for 2027 and 2028. Morgan Stanley believes this valuation is reasonable given the long-term commercialization potential of new businesses such as Muse, subscription services, and APIs. Under a bull case scenario, if Muse can continue to launch more practical features and drive growth in user scale, usage frequency, and commercialization revenue, Meta's stock price could reach $1,000, implying approximately 41% upside potential from the October 16 closing price. However, Morgan Stanley also cautions that mass adoption of AI agents will not happen overnight. Whether Meta can reduce computing costs through improved model efficiency and convert user growth into actual revenue will be key factors determining its AI investment returns and long-term valuation level.