A certain type of business succeeds covertly. Among them is Broadcom. It supplied the radio-frequency chips that allowed iPhones to connect to towers, and for years it remained hidden behind the scenes of the smartphone boom. Almost no one outside the industry knew what it was.
Asking a customer who makes the silicon in their phone in any electronics store in San Jose or Karachi would result in blank looks. It turns out that anonymity was a feature rather than a flaw.
| Field | Detail |
|---|---|
| Company Name | Broadcom Inc. |
| Ticker Symbol | NASDAQ: AVGO |
| CEO | Hock Tan |
| Headquarters | Palo Alto, California, USA |
| Founded | 1991 (as HP spin-off Avago); merged with Broadcom Corp. in 2016 |
| Industry | Semiconductors & Infrastructure Software |
| Key Segments | Semiconductor Solutions (~60%), Infrastructure Software (~40%) |
| FY2025 Revenue | ~$64.0 billion |
| AI Semiconductor Revenue (FY2025) | Over $20 billion |
| Adjusted EBITDA Margin | ~68% |
| Free Cash Flow (FY2025) | $26.9 billion |
| Market Capitalization | ~$1.5 trillion |
| Major Acquisitions | Brocade (2017), CA Technologies (2018), Symantec Enterprise (2019), VMware (2023, $69B) |
| 5-Year Stock Return | Over 850% |
| Stock Split | 10-for-1 in July 2024 |
The change was not announced in a single press release. It built up. Hock Tan, the well-known frugal CEO from Malaysia who handles acquisitions the same way private equity handles cheap real estate, continued to purchase items that appeared unattractive at first glance. Brocade in 2017. CA Technologies the following year. Symantec’s business division in 2019. Every transaction silently increased profits while drawing the same chorus of skepticism from analysts: “What is a chipmaker doing in mainframe software?” As you watch it happen, you get the impression that Tan was putting together something that the rest of the market had not yet given a name to.
VMware followed. After a protracted regulatory process, the $69 billion deal closed in late 2023 and fundamentally altered the company. All of a sudden, Broadcom was no longer a “collection of franchises” but rather a hybrid that combined the unattractive plumbing of high-speed Ethernet switching with the equally unattractive but essential software found in every Fortune 500 data center. Consumers complained about price adjustments. Boost Mobile turned to leave. Nokia used Red Hat as a hedge. However, the money continued to flow in.

Naturally, the AI buildout was what altered everything. A different reality was emerging among the engineers who were actually wiring up these clusters, while headlines focused on GPUs and Nvidia’s meteoric rise. Instead of training on a single chip, models train across thousands or even hundreds of thousands of accelerators connected by optical interconnects and switches. Broadcom is in charge of that. Although the company may not have fully anticipated this windfall in 2020 or 2021, it was undoubtedly prepared when the orders began coming in.
The FY2025 figures provide a clear picture of the situation. $64 billion in revenue, a 24% increase. Revenue from AI-related semiconductors alone exceeded $20 billion, almost doubling from the previous year. 68% adjusted EBITDA margins—software-like profitability for a company that still ships actual chips. Although it’s still unclear if the custom-silicon contracts with Google and Meta will continue to grow at the same rate or if the hyperscalers will eventually attempt to bring more of that work in-house, investors appear to think this is sustainable.
Broadcom introduced VMware Telco Cloud Platform 9 at Mobile World Congress 2026 in Barcelona, offering carriers sovereign AI environments, GPU-as-a-service, and a 40% TCO reduction over five years. On paper, the pitch is strong. It remains to be seen if telcos, battered, cautious, and still recovering from the hangover from 5G capex, will actually purchase it at scale. It’s obvious that some people dislike the new prices. Others are moving their workloads covertly.
It’s difficult to ignore the irony. The company that began by selling parts that no one could see is now selling the entire backbone that all people rely on. Tesla was questioned. Apple was questioned. For the most part, Broadcom continued to cut, integrate, and buy. As you watch this develop, you get the impression that the next ten years of computing, including edge inference, AI agents, sovereign clouds, and whatever else is created, will eventually pass through a Broadcom switch. That type of dominance is peculiar. Strong, quiet, and difficult to move.
