JPMorgan sees custom AI chips overtaking GPUs in 2027

JPMorgan Chase expects the custom AI ASIC market to enter a new growth cycle as major cloud and technology companies look for lower computing costs, better power efficiency and less dependence on general-purpose GPUs. The bank estimates the digital AI ASIC market will reach around $600 billion to $700 billion by 2026 and maintain a compound annual growth rate of over 40% to 50% in the coming years.

Broadcom is positioned as the dominant supplier, with about 80% to 85% of the high-end ASIC market, while Marvell ranks second at about 10% to 12%. JPMorgan projects Broadcom’s AI revenue will rise from around $20 billion in FY2025 to over $600 billion in FY2026 and over $1.5 trillion in FY2027, supported by programs tied to Google, Meta, ByteDance, OpenAI, SoftBank/Arm and Anthropic-related TPU rack-scale systems.

By 2027, total AI accelerator shipments are expected to reach 23.3 million units, with GPUs at 10.9 million units, representing 47%, and ASICs/XPUs at 12.5 million units, representing 53%. GPUs are still expected to grow, but cloud providers’ in-house chips may gain more ground in large, stable and predictable workloads.