TrendForce reports that surging AI server demand is driving major global cloud service providers (CSPs) — including Google, AWS, Meta, Microsoft, Oracle, Tencent, Alibaba, and Baidu — to ramp up investments in NVIDIA rack-scale GPU solutions, data center expansion, and in-house AI ASIC development.
Combined Capex for these eight CSPs is projected to exceed $420 billion in 2025, a 61 percent year-over-year increase, and is expected to reach $520 billion in 2026 with the adoption of rack-scale solutions like GB/VR systems. Investment focus is shifting from immediate revenue-generating assets to short-lived infrastructure such as servers and GPUs, emphasizing long-term competitiveness.
NVIDIA’s GB200/GB300 rack systems are expected to dominate CSP deployments in 2025, with emerging AI cloud customers like Tesla/xAI, CoreWeave, and Nebius joining established North American and Oracle CSPs. By late 2026, CSPs are likely to transition to the NVIDIA Rubin VR200 rack platform.
On the custom AI chip front, North American CSPs are expanding AI ASIC production. Google leads with its TPU v7p (Ironwood), growing over 40 percent in 2026. AWS’s Trainium v2 and upcoming Trainium v3 are set for mass production, doubling ASIC shipments in 2025 and growing further in 2026. Meta’s MTIA v2 and MTIA v3 aim to improve inference efficiency, while Microsoft’s Maia v2 is scheduled for 2026 production, though Maia v3 delays limit near-term shipments, putting it behind competitors.
Meta Platforms
Meta Platforms is significantly increasing its capital expenditures to bolster its artificial intelligence infrastructure. In 2025, the company plans to invest between $66 billion and $72 billion, marking a substantial increase from the $39.2 billion spent in 2024. This strategic allocation is aimed at enhancing AI capabilities, including the development of custom AI chips and the expansion of data centers. Meta’s commitment to AI infrastructure underscores its ambition to strengthen its position in the competitive AI landscape.
Amazon Web Services
Amazon Web Services (AWS) is set to invest approximately $100 billion in capital expenditures (CapEx) in 2025, up from $83 billion in 2024, to scale its AI infrastructure and maintain its leadership in cloud computing. This substantial investment underscores AWS’s commitment to expanding its data center capacity and enhancing its AI capabilities.
A significant portion of this Capex is allocated to the development of AI infrastructure, including the construction of new data centers and the expansion of existing facilities. Notably, AWS is investing $20 billion in Pennsylvania, focusing on AI innovation campuses in Salem and Falls Townships, and $10 billion in North Carolina to build out data center capacity in Richmond County. Additionally, AWS is expanding its presence in Ohio with an additional $10 billion investment, bringing its total commitment in the state to over $23 billion by the end of 2029 AP News reports.
In response to the growing demand for AI services, AWS is also enhancing its AI offerings. The company is introducing Trainium 2, an AI chip that offers a 30–40 percent better price-performance ratio than Nvidia solutions, and integrating advanced models like Claude 3.7 Sonnet and Llama 4 into its AI-enhanced Bedrock service.
Despite these aggressive investments, AWS faces challenges, including potential capacity constraints and delays in hardware acquisition. However, AWS remains focused on its long-term strategy to dominate the AI infrastructure market.
Alphabet Inc., Google’s parent company, is significantly increasing its capital expenditures to bolster its AI infrastructure and cloud services. In 2025, Alphabet plans to invest $85 billion, up from $75 billion previously announced, to meet the growing demand for AI and cloud computing capabilities. This investment is primarily directed toward expanding data centers, enhancing server capacity, and developing AI-specific hardware.
In the second quarter of 2025, Alphabet reported a 32 percent year-over-year increase in Google Cloud revenue, reaching $13.6 billion, driven by strong demand for AI infrastructure. The company’s AI initiatives, including the Gemini chatbot and AI Mode Search tool, have contributed to this growth, with AI Mode Search amassing 100 million users in the U.S. and India. Despite concerns over escalating AI-related costs, Alphabet’s strategic investments aim to strengthen its position in the competitive cloud and AI markets.
Microsoft
Microsoft is making significant capital expenditures to expand its AI infrastructure and cloud services. In fiscal year 2025, the company plans to invest approximately $80 billion, marking a substantial increase from the $53 billion spent in 2023. This investment is primarily directed toward building AI-enabled data centers to train AI models and deploy AI and cloud-based applications globally.
A notable portion of this expenditure includes a $30 billion investment in the United Kingdom over four years, which encompasses $15 billion in capital expenditures to develop the country’s cloud and AI infrastructure. Additionally, Microsoft has secured access to over 100,000 Nvidia GB300 GPUs through a $19.4 billion deal with Nebius, a cloud provider, as part of a broader $33 billion investment in emerging cloud infrastructure companies.
Furthermore, Microsoft has launched the world’s first supercomputer-scale GB300 NVL72 Azure cluster, powered by 4,608 Nvidia GB300 GPUs, capable of delivering 92.1 exaFLOPS of FP4 inference performance. This deployment underscores Microsoft’s commitment to advancing AI capabilities and enhancing its cloud services.
Rajani Baburajan
