Broadcom is once again drawing attention following an artificial intelligence (AI) summit held over the past weekend in San Francisco. Hock Tan, Broadcom's chief executive officer (CEO) and an immigrant from Malaysia, has grown the company through aggressive mergers and acquisitions (M&A) into the largest force leading the anti-Nvidia front.
According to companiesmarketcap.com, a website that tracks U.S. market capitalizations, Broadcom's market cap stood at $1.82 trillion (about 2,669.03 trillion won) as of the 26th (local time). That ranks seventh, following Nvidia, Apple, Google, Microsoft, Amazon and TSMC. Excluding TSMC, a Taiwanese company, Broadcom ranks as the second-largest U.S. semiconductor company after Nvidia ($5 trillion).
Broadcom traces its roots to the semiconductor products division that Hewlett-Packard (HP) operated starting in 1961. This organization continued for nearly 40 years before being absorbed into Agilent, which was spun off from HP's semiconductor business in 1999. Kohlberg Kravis Roberts (KKR), the world's largest private equity fund (PEF), and Silver Lake Partners acquired Agilent for $2.66 billion at the end of 2005 and launched it as Avago Technologies, headquartered in Singapore. Avago acquired Broadcom for $37 billion in 2015 and changed its name to that of the acquired company, Broadcom. The CEO of Avago at the time was Hock Tan, who has been in charge of management for 20 years, first at Avago from 2006 and then at Broadcom.
Tan was born in Penang, Malaysia, in 1953 and moved to the United States in 1971 at the age of 18 as a first-generation immigrant. He graduated from the Massachusetts Institute of Technology (MIT) and Harvard Business School.
Like a private equity fund, he grew Broadcom by attaching companies from the same industry. He acquired LSI, a system semiconductor specialist, for $6.6 billion in 2014; Brocade, a maker of network equipment such as routers and switches, for $5.9 billion in 2017; and VMware, a virtualization platform company, for $69 billion in 2023. Broadcom's market cap, which stood at around $2.7 billion in the Avago era, surpassed $1 trillion in 2024.
Tan also flexibly resolved conflict with the Donald Trump administration. In 2017, early in Trump's first term, Broadcom attempted to swallow Qualcomm for $117 billion. The aim was to acquire Qualcomm, which was competing with China's Huawei in the fifth-generation (5G) mobile communications infrastructure market, and expand its business into mobile chips. However, President Trump rejected the deal on national security grounds, calling Broadcom a Singaporean company close to China. A few months later, in November 2017, Tan appeared alongside President Trump at the White House and announced that Broadcom would relocate its headquarters to Silicon Valley in the United States.
At the time of the announcement, Tan said, "My mom would never have imagined that I would stand next to the U.S. president at the White House," drawing a smile from President Trump. The scene in which President Trump, watching the announcement from behind, replied, "My mom too (would never have known I would become president and enter the White House)," became a major talking point.
Though it lost Qualcomm, the AI boom was an enormous opportunity for Broadcom. Broadcom's core business had long been networking, but as demand for application-specific integrated circuits (ASICs) optimized for inference surged from last year, the chip capabilities Broadcom had built over 60 years began to shine. To counter Nvidia's graphics processing units (GPUs), big tech companies lined up to commission Broadcom to design their own chips. Broadcom oversees silicon manufacturing, power and thermal management, and memory integration, as well as final foundry (contract semiconductor manufacturing) preparation.
Google, the leader in this field, also chose Broadcom. Broadcom is participating in the development of Google's ASIC, the Tensor Processing Unit (TPU). OpenAI also announced in October last year that it would cooperate with Broadcom to develop custom chips, and in June this year, OpenAI's first proprietary inference-only chip, "Jalapeno," was unveiled. Anthropic, OpenAI's rival, signed a contract with Google and Broadcom in April this year to secure gigawatt-scale TPU capacity. Meta also entrusted the design to Broadcom in March this year when it unveiled four products in its Meta Training and Inference Accelerator (MTIA) lineup, including the MTIA 300, 400, 450 and 500.
As Broadcom leads the custom semiconductor market, it is putting Nvidia on edge. According to Forbes, Nvidia's top-spec GPUs are priced at $30,000 to $40,000 each, while custom TPUs in which Broadcom is involved can be purchased for $12,000, and the energy efficiency of TPUs in inference tasks is 67% higher than comparable GPUs. At the unveiling of Jalapeno last month, Tan boasted of the company's technology, saying it "has performance on par with Nvidia's Blackwell chips or Google's TPUs."
