JPMorgan Chase: AI custom chip shipments may surpass GPU by 2027, with Broadcom and Marvel on the rise BlockBeats News: On June 22nd, JPMorgan Chase announced that as large cloud computing companies and tech giants seek to reduce AI computing costs, improve energy efficiency, and break free from single path dependence on general-purpose GPUs, the custom chip ASIC market is entering a new growth cycle, and Broadcom and Marvel are expected to be the biggest beneficiaries of this trend. In a recently released semiconductor industry research report, Morgan Stanley analysts Harlan Sur and Mayur Ramdhani estimated that the digital AI ASIC market will reach approximately $60 billion to $70 billion by 2026 and maintain a compound annual growth rate of over 40% to 50% in the coming years. The report states that Broadcom currently holds approximately 80% to 85% of the high-end ASIC market share, with Marvel ranking second with a share of around 10% to 12%. The rapid growth in demand for AI computing is changing the structure of chip procurement. JPMorgan Chase believes that customers such as Google, Amazon, Meta, Microsoft, OpenAI, and SoftBank/Arm are accelerating the development of self-developed or customized 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 customers, software stacks, or platforms, making them more suitable for ultra large scale cloud vendors with large-scale internal workloads. The report predicts that Broadcom's AI revenue will significantly increase from approximately $20 billion in fiscal year 2025 to over $60 billion in fiscal year 2026, and track to reach over $150 billion in fiscal year 2027. Its project pipeline includes Google TPU, Meta MTIA, ByteDance AI video and network chips, OpenAI XPAmericaoftBank/Arm XPU, and Anthropic related TPU rack level solutions. On the Marvel side, JPMorgan expects its data center revenue to increase from approximately $6.1 billion in 2025 to approximately $9.3 billion in 2026, and reach approximately $14.6 billion in 2027. The growth drivers include Amazon Trainium 3 and Trainium 4, Microsoft Maia, Google SmartNIC/DPU, CXL controllers, as well as 800G/1.6T optical DSP, coherent lite, and initial CPO solutions. The report also makes a key judgment: by 2027, the unit shipment volume of AI ASIC/XPU will exceed that of GPU. JPMorgan predicts that the total shipment of AI accelerators will reach 23.3 million by 2027, of which 10.9 million will be GPUs, accounting for 47%; The number of ASIC/XPUs is 12.5 million, accounting for 53%. This means that although GPUs will continue to grow, custom chips may occupy a larger share in the deployment of new AI computing power. JPMorgan cited Google/Roadcom TPU7x Ironwood and Nvidia Blackwell as examples, stating that AI ASICs are competitive in terms of cost-effectiveness and power efficiency. The report shows that the FP8 computing power of TPU7x Ironwood is close to Nvidia B200/B300, but the estimated selling price is about $13000, lower than B200's $35000 and B300's $40000; Its computing power per dollar and per watt is also better than the control GPU. This judgment does not mean that Nvidia demand will rapidly weaken. On the contrary, it points to the differentiation of AI infrastructure investment: GPUs continue to serve general training and inference needs, while cloud vendors' self-developed ASICs will gain higher penetration in large-scale, stable, and predictable internal workloads. For investors, JPMorgan's report strengthens the AI hardware chain from GPUs to ASICs, advanced packaging, HBM interfaces, and more SerDes、 The logic of optical interconnection and CPO diffusion. If the predictions in the report are fulfilled, Broadcom and Marvel will not only be AI network or connectivity chip suppliers, but will also become core platform companies in the next stage of AI computing architecture migration.