Capital Spending by Eight Global CSPs to Top $710 Billion in 2026, Accelerating AI and ASIC Expansion

The rapid buildout of generative AI infrastructure has turned high-bandwidth memory, or HBM, and DRAM into major cost drivers for cloud service providers. Rising memory prices are lifting the cost of AI accelerators and servers while forcing operators to reconsider how computing capacity, memory and power are balanced. The shift matters because memory supply and pricing increasingly determine data-center investment returns and the competitiveness of AI hardware.

TrendForce expects memory to account for 68% of capital expenditure at major global cloud service providers by 2027 as they accelerate AI infrastructure deployments. The research firm said persistently rising HBM and DRAM contract prices are feeding into higher AI chip prices. Those costs are also prompting operators to redesign AI system memory architectures and step up development of custom AI ASICs to improve efficiency and curb infrastructure expenses.

Demand for generative AI and large language model computing is driving global cloud service providers to accelerate investment in data-center infrastructure. Rack-scale AI servers allow operators to deploy computing capacity more quickly, while custom ASICs can improve performance and cost efficiency. The shift is making both technologies central to cloud capital-allocation plans and strengthening the outlook for server makers and semiconductor suppliers.

TrendForce said in its latest research that combined capital expenditure by the world’s nine largest cloud service providers is expected to rise nearly 90% in 2026 from 2025 as they expand generative AI capacity. With operators showing greater willingness to buy rack-scale AI servers and internally developed ASICs, the research firm raised its forecast for global AI server shipment growth in 2026 to nearly 31%.

Amazon, Google, Meta and Microsoft are racing to expand their “AI factories,” pouring money into data centers, servers, computing chips and power infrastructure. The four major U.S. cloud service providers are seeking to secure computing capacity for generative AI, reshaping competition across the global semiconductor, energy and cloud markets.

Amazon expects capital spending to rise to $200 billion in 2026, with the funds focused on AI data centers and infrastructure. Its AI business already generates more than $15 billion in annualized revenue, and the company is considering selling its in-house chips to outside customers. Amazon also expects AI to boost AWS sales, which it forecasts will double to $600 billion by 2036.