From 'Cowboy' to the dairy industry giant, Niugen Sheng entered the cowshed at the age of 20 and worked at a cattle farm. Five years later, he entered Yili and started as a bottle washer. He was later promoted to workshop director. Starting in 1992, he served as the vice president of operation at Yili. In 1999, he left and created Mengniu.$NVIDIA (NVDA.US)$GPU switches to$Broadcom (AVGO.US)$ASIC will OpenAI start another brutal battle in the chip market?
The dream of Ultraman to build a chip empire worth 7 trillion dollars has taken another step forward. It is reported that OpenAI is negotiating with chip design companies such as Broadcom to develop new AI chips.
Citigroup believes that after Tesla and ByteDance, OpenAI will become the fourth largest ASIC (custom chip) customer of Broadcom, and Broadcom is expected to deliver to OpenAI after the second half of 2025.$Alphabet-A (GOOGL.US)$,$Meta Platforms (META.US)$Currently, Google, Meta,
On Thursday local time, Citigroup analyst Christopher Danely released a report reaffirming the buy rating on Broadcom with a target price of $175, which is about 9% higher than the current stock price and roughly in line with the expected EPS of 2025, about 33 times, roughly consistent with the industry average. Citigroup believes that its EPS (including stock compensation) can reach a peak of $6.
Is OpenAI reshaping the chip market by moving from GPUs to ASICs?
It is worth noting whether the cooperation with Broadcom will pose a threat to OpenAI's "old partner" Nvidia?
The competition between ASICs and general-purpose computing cards has existed for a long time, and it has intensified with the entry of cloud providers and large OEMs.
Currently, the main manufacturer of general-purpose computing cards is Nvidia, which holds nearly 70% of the AI computing market share; the main manufacturers of ASICs are Broadcom and Marvell, which together hold over 60% of the ASIC market share.
ASIC sacrifices versatility for high performance in specific scenarios, while general-purpose computing cards have versatility, but their performance is not as good as ASICs in specific scenarios.
In fact, different computing card customers have different requirements.
Cloud providers may focus more on elastic computing, while enterprises may be more concerned about cluster computing power, etc. In the face of specific needs, ASIC is more advantageous than standard computing cards and more in line with customers' usage scenarios.
Currently, Google, Meta,$Microsoft (MSFT.US)$And.$Amazon (AMZN.US)$Broadcom's customer base not only includes Google and Meta, but also Cisco, Fujitsu, LM Ericsson Telephone, Nokia OYJ, HPE, NEC, Juniper Networks, Ciena, and Western Digital, among many others from various industries.
For example, Google's TPU, Meta's MTIA, Microsoft's Maia, Amazon's Trainium2, etc.
It should be noted that the cost of ASICs may be higher than that of general-purpose computing cards. According to Deutsche Bank's calculations, the TCO (total ownership cost) of GB200 is 44% lower than that of TPUv5 and 30% lower than that of Trainimium 2. Product structure, 10-30 billion yuan products operating income of 401/1288/60 million yuan respectively.
Broadcom's management expects that the potential market for the company's AI business will reach between 30 billion and 50 billion dollars per year in the next 3-5 years, while the company's target for the 2024 fiscal year is only 11 billion dollars.
Broadcom's main business focuses on ASIC chips for large consumer AI platform customers (Google, Meta, ByteDance, etc.).
The management believes that these large consumer AI customers will not slow down their investment in xPU clusters in the next 3-5 years, and the scale will increase from 0.1 million clusters this year to 1 million clusters. Broadcom emphasizes that the most important thing is that there is no sign of demand slowing down before reaching this scale.
Each cluster has a capital expenditure of $40 billion, and Broadcom's services can account for $25 billion in the $40 billion (including computing, networking, excluding electrical utilities), which means that Broadcom's potential market for AI business will reach between $30 billion and $50 billion per year in the next 3-5 years.
Editor/ping