JPMorgan: Shipments of AI custom chips may overtake GPUs in 2027, with Broadcom and Marvell riding the wave

BlockBeats News, June 22 — JPMorgan Chase stated that as large cloud computing companies and tech giants seek to reduce AI computing costs, improve energy efficiency, and break free from reliance on a single pathway of general-purpose GPUs, the custom chip ASIC market is entering a new growth cycle, with Broadcom and Marvell expected to be the biggest beneficiaries of this trend.

In a recent semiconductor industry research report, JPMorgan analysts Harlan Sur and Mayur Ramdhani estimate that the digital AI ASIC market will reach approximately $60 billion to $70 billion by 2026, maintaining a compound annual growth rate of over 40% to 50% in the coming years. The report states that Broadcom currently holds about 80% to 85% of the high-end ASIC market share, with Marvell in second place, holding about 10% to 12%.

The rapid growth in AI computing demand is changing chip procurement structures. JPMorgan believes that clients such as Google, Amazon, Meta, Microsoft, OpenAI, and SoftBank/Arm are accelerating the development or customization of AI processors to achieve better performance, power consumption, and total cost of ownership. Unlike Nvidia and AMD’s general-purpose GPUs, ASICs are typically designed for specific clients, software stacks, or platforms, making them more suitable for hyperscale cloud providers with large internal workloads.

The report projects that Broadcom’s AI revenue will significantly increase from about $20 billion in fiscal year 2025 to over $60 billion in fiscal year 2026, and track to over $150 billion in fiscal year 2027. Its project pipeline includes Google TPU, Meta MTIA, ByteDance AI video and network chips, OpenAI XPU, SoftBank/Arm XPU, and TPU rack-level solutions related to Anthropic.

On the Marvell side, JPMorgan expects its data center revenue to grow from approximately $6.1 billion in 2025 to about $9.3 billion in 2026, reaching around $14.6 billion in 2027. Growth drivers include Amazon Trainium 3 and Trainium 4, Microsoft Maia, Google SmartNIC/DPU, CXL controllers, as well as 800G/1.6T optical DSPs, coherent lite, and early-stage CPO solutions.

The report also presents a key judgment: by 2027, the unit shipments of AI ASICs/XPU will surpass those of GPUs. JPMorgan forecasts that in 2027, total AI accelerator shipments will reach 23.3 million units, with GPUs accounting for 10.9 million (47%) and ASICs/XPU for 12.5 million (53%). This indicates that although GPUs will continue to grow, custom chips may capture a larger share of new AI compute deployments.

JPMorgan cites Google/Broadcom TPU7x Ironwood and Nvidia Blackwell as examples, stating that AI ASICs are competitive in cost-performance and power efficiency. The report shows that TPU7x Ironwood’s FP8 computing power is close to Nvidia B200/B300, but its estimated price is about $13k, lower than B200’s $35k and B300’s $40k; its cost per dollar of compute and per watt of compute also outperform the comparison GPUs.

This judgment does not mean Nvidia’s demand will decline rapidly. Instead, it points to a divergence in AI infrastructure investment: GPUs will continue serving general training and inference needs, while cloud providers’ in-house ASICs will achieve higher penetration in large-scale, stable, and predictable internal workloads.

For investors, JPMorgan’s report reinforces the logic of AI hardware transitioning from GPUs to ASICs, advanced packaging, HBM interfaces, SerDes, optical interconnects, and CPO diffusion. If the forecasts in the report materialize, Broadcom and Marvell will not only be AI network or connectivity chip suppliers but will become core platform companies in the next phase of AI computing architecture migration.