Citi: OpenAI to become Broadcom's fourth largest customer, expected to deliver in the second half of next year

Author: Bu Shuqing

Source: Hard AI

Ultraman's dream of building a $7 trillion chip empire has taken another step forward. It is reported that OpenAI is negotiating with chip design companies such as Broadcom to develop new AI chips.

Citi believes that following Google, Meta, and ByteDance, OpenAI will become Broadcom's fourth largest ASIC (Application-Specific Integrated Circuit) customer, with Broadcom expected to deliver to OpenAI after the second half of 2025.

On Thursday local time, Citi analyst Christopher Danely released a report reiterating a buy rating on Broadcom with a target price of $175, about 9% higher than the current stock price. This is roughly 33 times the expected earnings per share in 2025, in line with the industry average. Citi believes that its earnings per share (including stock compensation) could reach a peak of $6.

Transition from GPU to ASIC, Is OpenAI Reshaping the Chip Market?

It is worth noting whether the cooperation with Broadcom poses a threat to OpenAI's "old partner" NVIDIA?

According to a previous article by Hard AI, the competition between ASIC and general-purpose graphics cards has been going on for a long time, intensifying with the entry of cloud vendors and large OEM manufacturers.

Currently, the main manufacturers of general-purpose graphics cards are NVIDIA, holding nearly 70% of the AI computing market share; the main manufacturers of ASIC are Broadcom and Marvell, with the two holding over 60% of the ASIC market share.

ASIC sacrifices versatility for high performance in specific scenarios; while general-purpose graphics cards are versatile, their performance may not be as good as ASIC in specific scenarios.

In fact, the demand varies for different computing card customers.

Cloud vendors may value elastic computing more, while enterprises may focus more on cluster computing, etc. Faced with specific needs, ASIC has more advantages over standard computing cards, fitting more closely to the customer's own usage scenarios.

Currently, companies like Google, Meta, Microsoft, and Amazon are leading the trend of ASIC.

For example, Google's TPU, Meta's MTIA, Microsoft's Maia, Amazon's Trainium2, etc.

It is important to note that the cost of ASIC may be higher than that of general-purpose graphics cards. According to Morgan Stanley's calculations, the TCO (Total Cost of Ownership) of GB200 is 44% lower than TPUv5 and 30% lower than Trainimium 2

Broadcom's management expects that in the next 3-5 years, the potential market for the company's AI business will reach between $30 billion and $50 billion annually, while the company's target for the fiscal year 2024 is only $11 billion.

Broadcom's main business focuses on ASIC chips needed by large consumer AI platform customers (such as Google, Meta, ByteDance, etc.).

Management believes that these large consumer AI customers will not slow down their investment in xPU clusters in the next 3-5 years, with the scale increasing from 100,000 clusters this year to 1 million clusters. Broadcom emphasizes that the most important thing is that there are no signs of demand slowing down before reaching this scale.

With a capital expenditure of $40 billion for each cluster, Broadcom's services can account for $25 billion out of the $40 billion (including computing, networking, excluding power, etc.), which means that in the next 3-5 years, the potential market for the company's AI business will reach between $30 billion and $50 billion annually