Hello, this is Jimza, an editor at Money Never Sleeps. Recently, there is a stock that domestic investors are showing a lot of interest in, which is Broadcom. It has risen to the top in terms of net buying over the past month, attracting many people's attention. There was an interesting news about this company recently, which is about Broadcom forming a partnership with OpenAI. Specifically, various foreign media outlets reported that Broadcom has started discussions with OpenAI for the development of new AI semiconductors. It was mentioned that this is to reduce dependence on NVIDIA and to further facilitate the supply of AI semiconductors. Since the release of the Nvidia PT, investments in AI companies have increased, leading to a natural increase in demand for high-performance GPUs for large-scale AI training and operation.

There has been a surge in prices, and in fact, NVIDIA has almost monopolized the AI accelerator market, leading to a bottleneck where supply cannot keep up with demand. As a result, the prices of AI semiconductors have skyrocketed, and even if you manage to make a purchase, it has become difficult to receive the product on time. In this situation, big tech companies have started to seek other outlets. Google, which has been developing TPUs and using its own AI semiconductors for a long time, as well as Microsoft, Meta, Amazon, Tesla, and many other companies have started to bring semiconductor design in-house. Even OpenAI, which was at the center of the AI craze, was no exception. In fact, the idea of OpenAI bringing semiconductor design in-house has been discussed for a long time. It was reported last October that OpenAI was looking for potential acquisition targets.

After nearly 10 months of exploration, the company selected by OpenAI is Broadcom. For those who are familiar with Broadcom, this decision by OpenAI may seem somewhat surprising. This is because Broadcom is not a company that designs accelerators for GPU AI operation, but rather specializes in communication semiconductors, such as Wi-Fi chipsets and RF chipsets. Communication chips and semiconductors for AI processing are distinctly different products. In fact, just by looking at the product portfolio that Broadcom has released related to AI, you can see that the area the company is focusing on is networking. In simple terms, the semiconductors designed by Broadcom are not for actually processing AI workloads, but rather for quickly transmitting and receiving those workloads in network infrastructure. So, if that's the case, then OpenAI...

Why did AI choose Broadcom as a potential partner to develop Open AI semiconductors? It's not about acceleration or communication chips. It's because Broadcom has excellent capabilities in ASIC design. ASIC stands for Application-Specific Integrated Circuit. In simple terms, it refers to semiconductors tailored for specific purposes according to customer requests. Broadcom is a company with various core technologies in custom semiconductor design. One of its notable technologies is FPGA. FPGA stands for Field-Programmable Gate Array. What is FPGA in the simplest terms? It can be likened to Lego blocks. Unlike conventional semiconductors where once a circuit is imprinted, it cannot be changed, FPGA chips can have their circuits altered through programming, similar to disassembling and reassembling Lego bricks according to the user's needs. In other words, FPGA chips allow for circuit changes through programming, unlike typical physical semiconductor circuits.

FPGAs are chips that can be reprogrammed through software. When a company wants to create custom semiconductors, it is essential to test whether the semiconductor operates according to their design intentions. Nevertheless, a semiconductor is hardware. As with hardware, unexpected defects may be discovered after the actual product is produced during the design phase. However, it is not feasible in terms of cost and time to fabricate wafers and engrave circuits to create semiconductors from the initial R&D stage to verify if the chip actually functions well every time. Developing a prototype assuming mass production and product development involves making such prototypes after the product design is in the final stage. Therefore, the approach that companies aspiring to make chips typically choose is using FPGAs from the early stages of development.

To verify if the design will function as intended when implemented in actual hardware, companies utilize FPGA technology, among others. Broadcom is a company that possesses various essential capabilities in this field, being recognized as a top-tier player in the semiconductor industry. For example, Google's TPU, currently in use, was created through collaboration with Broadcom. In addition to Google, other customers like Meta and Microsoft are also clients of Broadcom. Just by looking at Broadcom's list of partners, one can infer how well-regarded their capabilities are in the industry. In fact, Broadcom's revenue related to AI custom chip design has been steadily increasing. Broadcom announced its financial results for the second quarter of the fiscal year 2024 on June 12th. During this period, revenue from AI products reached 3.1 billion.

It was said to represent a record level of dollars. Broadcom's total quarterly revenue was about $12.5 billion, with about 25% coming from AI semiconductors. In the last quarter earnings announcement, Broadcom projected that this year's revenue related to AI semiconductors would reach $10 billion. Compared to last year's AI chip sales of $3.5 billion calculated by JP Morgan, Broadcom has already caught up with last year's annual revenue in just one quarter, showing a steep growth trend. Broadcom's second-quarter revenue increased by 43% compared to the same period last year, reaching $3.75 billion. In fact, it can be said that AI semiconductors are currently driving the growth of this company to a significant extent. In summary of the situation so far, Broadcom is enhancing its global top-tier capabilities through chip design.

It is being adopted as a partner of big tech companies that want to reduce their dependence on NVIDIA, and OpenAI, which, like other big tech companies, currently wants a custom AI semiconductor, is discussing a partnership. Broadcom, which is steadily growing based on partnerships with big tech companies, could emerge as a new competitor threatening NVIDIA in the AI semiconductor industry. I wonder if this is the case. Subscribe to Wonny Never Sleep Newsletter and Naver Premium Content now.

오픈 AI는 왜 엔비디아가 아닌 브로드컴을 선택했을까?

가장 빠르고 정확하게 미국장 소식을 정리해드림 🇺🇸

비즈니스/협업 문의 : snowballlabs.official@gmail.com
머니네버슬립의 모든 채널 : https://litt.ly/moneyneversleeps

#머니네버슬립 #미국주식 #주식투자
