Illustration = ChatGPT DALL-E 3

Broadcom posted results that exceeded market expectations, driven by demand for customized artificial intelligence (AI) chips in a move away from Nvidia. Despite the stock price of Broadcom dropping nearly 23% this year amid concerns about a bubble in the AI craze, the stock surged more than 13% in after-hours trading shortly after the company mentioned that demand for AI computing remained solid on June 6.

Broadcom reported that its revenue for the first quarter of fiscal year 2025 (Nov. 1, 2024, to Feb. 2, 2025) increased by 25% year-over-year to $14.92 billion (about 21.6 trillion won). Net income reached $9.1 billion (about 13 trillion won), exceeding market expectations by 24%. Revenue from the semiconductor institutional sector increased by 11% year-over-year to $8.2 billion (about 12 trillion won), while revenue from the software institutional sector rose by 47% year-over-year to $6.7 billion (about 9.7 trillion won).

During the earnings conference call, Hock Tan, Broadcom's chief executive officer (CEO), noted, "Customer expenditures related to AI were the key growth driver in the first quarter," and added that "in the second quarter, AI-related revenue is expected to reach $4.4 billion (about 6.37 trillion won)." In the first quarter, AI-related revenue was $4.1 billion (about 5.93 trillion won), marking a 77% increase from the same period last year.

The demand for application-specific integrated circuits (ASICs), which drive Broadcom's AI business, is notable. ASICs are customized chips specializing in 'specific functions' such as learning and inference, and they are gaining attention as alternatives to graphics processing units (GPUs) in generative AI data centers due to lower prices, power consumption, and total investment costs. Broadcom is developing ASICs in partnership with several major corporations, including Google, Meta, and China's ByteDance, which operates the world's largest short-form video platform, TikTok. Broadcom is also securing four additional customers beyond its three existing hyper-scalers, with sales expected soon from two of these clients.

On that day, CEO Tan expressed confidence that Broadcom's ASICs outperform Nvidia's AI accelerator chips. He emphasized, "We are increasing production of AI semiconductors for hyper-scaler customers," stating that "in some cases, our customized chips are more efficient than Nvidia's general-purpose accelerator chips." Tan added, "Hyper-scaler customers are consistently pursuing aggressive investments."

Industry analysts assess that Broadcom is establishing a unique position in the ASIC market. Companies like Google and Meta are continuing to collaborate with Broadcom, and there are predictions that they will equip their data centers with over 10 million AI chips by 2027. Kinhai Chan, a researcher at market research firm Summit Insights, noted, "Broadcom has a diverse client base in the AI market, having secured numerous of the world's largest data center investment corporations as clients for AI-specific ASICs, giving it a favorable position compared to competitors."

The outlook is also bright. Broadcom expects its second-quarter revenue to reach $14.9 billion, exceeding market expectations of $14.6 billion. Although it previously provided guidance for AI-related revenue to be between $60 billion and $90 billion (about 87 trillion to 130 trillion won) by 2027, Broadcom representatives mentioned that the rapid expansion of the business may result in even larger revenue figures.

※ This article has been translated by AI. Share your feedback here.