Can surging data center capex keep AI server growth marching ahead?

Bernstein reports that global hyperscalers are significantly accelerating capital expenditure for AI infrastructure, with 2026 capex estimates raised by nearly 20% and total investment projected to reach $500 billion by 2027, growing at a 26% CAGR. This surge, evidenced by $770 billion in data center projects and increased capex guidance from major cloud providers, is driving a substantial increase in global server shipments, forecast to hit $450 billion by 2026, encompassing strong growth in both high-end GPU servers and expanding ASIC adoption. The robust investment, fueled by positive AI demand and easing supply constraints, indicates sustained growth and significant revenue opportunities across the AI server supply chain.

Analysis

A recent Bernstein report indicates a significant acceleration in AI-related capital expenditures, underpinning a robust growth outlook for the server supply chain. Consensus estimates for 2026 capex have been revised upward by nearly 20% since July, with total investment now projected to grow at a 26% compound annual rate to approximately $500 billion by 2027. This spending is substantiated by a pipeline of upcoming and under-construction data centers valued at $770 billion and specific upward revisions to capex guidance from key hyperscalers, including Alphabet (+$10B to $85B for 2025), Amazon (to $118.5B annually), and Meta ($66-72B for 2025). The direct consequence is a forecasted expansion of the global server market to $450 billion in 2026. This growth is dual-pronged: high-end GPU server shipments are expected to grow about 55% in 2025, driven by Nvidia's upcoming GB200/300 platforms, while ASIC adoption is also rapidly expanding. ASICs are projected to comprise nearly 45% of CoWoS-based AI accelerator shipments by 2026, a trend reinforced by Broadcom securing over $10 billion in new AI orders. Critically, prior market concerns regarding supply chain constraints, such as T-glass for Nvidia's chips, are reportedly easing, de-risking the execution of this build-out and supporting strong revenue growth forecasts for downstream players like Taiwanese ODMs.

Key Decisions for Investors

• Investors should consider exposure across the entire AI infrastructure value chain, as the secular capex trend benefits both semiconductor designers like Nvidia (NVDA) and Broadcom (AVGO) as well as the hyperscalers funding the investment, such as Microsoft (MSFT), Alphabet (GOOGL), and Meta (META).
• Portfolio construction should acknowledge the dual growth vectors in AI hardware; while Nvidia maintains GPU dominance, the rapid rise of the custom ASIC market, benefiting players like Broadcom, presents a distinct and complementary investment thesis.
• Forward-looking capex guidance from hyperscalers remains a primary bull-case indicator; investors should monitor future earnings reports for continued upward revisions, which signal sustained demand momentum.
• The easing of supply-chain bottlenecks reduces execution risk for the sector, potentially warranting a re-evaluation of positions in companies whose growth was previously perceived to be constrained, including downstream manufacturers and component suppliers.