TSMC CoWoS Capacity Forecast Raised as Mizuho Sees AI Server CPU Demand Surging Through 2027 
Kioxia Bets on AI Memory Boom With Next-Gen NAND Production in Japan 
OpenAI Proposes 5% U.S. Government Stake Amid AI Policy Talks 
Trump Administration to Launch Voluntary AI Standards for Frontier Models 
Kuaishou Stock Jumps as Kling AI Secures $2 Billion Funding Round 
EU Chip Industry Faces Growing Risks From China Export Controls and U.S. Technology Dependence: Report 
Meta Stock Jumps as AI Cloud Expansion Challenges AWS, Microsoft, and Google 
Apple Expands iPhone Lineup, Boosts Foldable iPhone Production Plans Through 2027 
SoftBank’s LY Corp, Bain Raise Kakaku.com Bid to ¥670 Billion, Intensifying Takeover Battle 
Anthropic Brings Claude AI Models to Microsoft Azure Foundry With NVIDIA Blackwell GPUs 
Anthropic Tightens AI Access Controls After Reports of China-Based Workarounds 
Baige Online Shares Soar 333% in Hong Kong IPO Debut as AI Insurance Demand Lifts Chinese Listings 
Northern Star Appoints New CEO as Activist Elliott Pushes for Leadership Overhaul 
Lockheed Martin Emerges as Frontrunner to Acquire Ultra Maritime in $3.5 Billion Defense Deal 
Tesla Q2 Deliveries Lift Chinese Auto Suppliers as EV Demand Improves 
DOJ Seeks Dismissal of Fraud Charges Against Gautam Adani in U.S. Court 
Smartphones are helping filmmakers tell the stories the movie industry overlooks 
Meta Platforms (NASDAQ: META) is preparing to expand into the cloud computing market, a move that could position the company as a future rival to Amazon Web Services (AWS), Microsoft Azure, and Google Cloud. According to Morgan Stanley, Meta is evaluating two possible business models: a hosted AI API and model-access platform similar to AWS Bedrock, or a "neocloud" offering that leases raw AI computing capacity powered by its own infrastructure.
Morgan Stanley believes the neocloud strategy is the more practical path. Analysts argue that building a full AI platform with hosted models and developer tools would require significant hiring, technical execution, and product development. Meta’s Muse AI models have yet to demonstrate competitive performance on industry benchmarks such as TerminalBench and SWE Bench Verified, while also facing strong competition from advanced AI models like Google’s Gemini.
The investment bank estimates Meta will add roughly 2 gigawatts of owned IT capacity in 2026 and another 3.5 gigawatts in 2027, compared with an estimated 3-gigawatt base by the end of 2025. This growing infrastructure could allow Meta to lease excess compute capacity to businesses facing AI chip shortages and high demand for computing resources.
While Meta currently rents approximately 2.5 gigawatts of compute from providers including CoreWeave, Nebius, Google Cloud, and Oracle, Morgan Stanley notes that the company could monetize portions of its own infrastructure as new capacity comes online.
The bank also sees meaningful earnings potential from the neocloud model. Based on current market pricing, leasing 250 megawatts of compute capacity for one year at around $40 per watt could increase Morgan Stanley’s 2028 earnings-per-share estimate for Meta by approximately $3, representing roughly 8% upside.
Despite this opportunity, Morgan Stanley views cloud leasing as a temporary earnings driver rather than Meta’s long-term growth strategy. The firm maintains its Overweight rating, citing higher expectations for MetaAI, AI-powered business messaging, subscription services, diffusion models, and other AI products that could generate sustainable revenue growth.
Morgan Stanley projects Meta’s capital expenditures to increase to approximately $175 billion in 2027 and $205 billion in 2028, up from an estimated $145 billion in 2026, as the company expands its AI data center footprint. The analysts believe additional investments may be required if Meta ultimately pursues a full-scale cloud platform, making future capital spending likely to trend even higher.
