Piper Sandler covers the AI chip sector for the first time: Nvidia (NVDA.US) and Broadcom (AVGO.US) are the “first choice”, and the computing power market is expected to reach 2.2 trillion US dollars in 2030

The Zhitong Finance App learned that Piper Sandler published a research report covering the artificial intelligence (AI) chip and semiconductor sector for the first time. The bank initiated ratings for seven major semiconductor stocks. Among them, AMD (AMD.US), Arm (ARM.US), Broadcom (AVGO.US), Maywell Technology (MRVL.US), and Nvidia (NVDA.US) received “additional” ratings, while Intel (INTC.US) and Qualcomm (QCOM.US) received “neutral” ratings. The bank pointed out that Broadcom and Nvidia are listed as “top picks” in the field of AI accelerators, while AMD and Arm are receiving a lot of attention for seizing share in the server CPU field.

In terms of target prices, Piper Sandler gave AMD 600 dollars, Arm 320 dollars, Broadcom 460 dollars, Intel 110 dollars, Mwell Technology 270 dollars, Nvidia 300 dollars, and Qualcomm 190 dollars.

Analysts David O'Connor and Zackary Altman said in the report that computing power can be described as the core cornerstone of AI development. With the implementation and explosion of intelligent AI, the AI industry is in urgent need of more computing power support.

The bank indicated that production capacity was sold out in 2027, and the 2028 order book was full. Piper Sandler predicts that the computing power market will reach 2.2 trillion US dollars by 2030, of which the accelerator market is 2 trillion US dollars and the CPU market is 200 billion US dollars; this means that the computing power market will expand about 5 times by 2030 and drive the revenue expansion of the computing power industry chain, which translates into a compound annual growth rate of about 45% of the average earnings per share in this group of stocks.

The bank believes that these companies are the most important participants in the construction of hyperscale cloud vendors and new cloud vendors. The report also points out that the current scarcity of computing power is clearly visible in the GPU hourly pricing of new cloud vendors. Silicon Data's H100 pricing index shows that Nvidia H100 GPUs cost $2.60 per hour, compared to $2.00 in December, before the introduction of OpenClaw.

The report also mentioned that in February 2025, there was also a surge in demand due to increased demand brought about by DeepSeek's deployment; falling computing power costs will further drive usage, which in turn will continue to drive demand for computing power. Piper Sandler stressed that it is important to note that Nvidia mass-produced the H100 as early as the fourth quarter of 2022, which means that this chip has been iterating on technology for almost 4 years. Additionally, component costs are also rising, thereby boosting hardware vendor revenue; the bank found that GPU chip costs have risen from 30% to 40% in the past two generations of products.

At the individual stock level, O'Connor called Nvidia “the absolute leader in AI computing power, with 80% market share.” He said that the explosion of smart AI workloads this year has brought about stepwise changes in demand, and supply may remain tight for the next two to three years, adding that the stock is “one of the lowest-valued stocks in the AI field” calculated at about 14 times the expected valuation for the 2028 fiscal year.

O'Connor called Broadcom the leader in custom ASIC chips, with a 75% share, and pointed to tight supply issues. “Demand is double the supply, so securing additional supply is key,” he wrote. He also pointed out that the company's predictable market demand for the 2027 fiscal year has reached 12 gigawatts.

O'Connor called AMD “the best beneficiary target for intelligent AI,” pointing out that AMD continues to gain share in the server CPU field, and that its Helios GPUs are also increasing, and anchor customers include OpenAI, Meta (META.US), and Anthropic. He expects AMD's profits to grow at a CAGR of 65% by 2030.

Analysts are also optimistic about Maywell Technology's data center business and say the $120 billion agreement with Google (GOOGL.US) is transformative. He also mentioned that Analyst Day on October 6th could be a potential catalyst.

Finally, O'Connor emphasized Arm's dominant position in the CPU intellectual property field and expansion into custom chip and accelerator IP, which he believes could double Arm's profits.