Broadcom

Broadcom (AVGO) is a publicly traded Networking & Custom Silicon company founded in 1961 and based in San Jose, USA.

Broadcom designs custom AI accelerators for hyperscalers (Google TPU) and supplies networking silicon critical to AI cluster interconnects. Google TPU v4/v5 manufactured by Broadcom; also leads in Ethernet switching and SerDes for AI networking. Market cap ~$700B post-VMware acquisition.

Broadcom is the dominant supplier of custom ASICs for Google, Meta, and others, enabling vertical integration in AI compute.

broadcom.com Networking & Custom SiliconUSApublic · AVGO

At a glance

Total raised

$648M

last round ipo

Headcount

—

no public estimate on record

Headquarters

San Jose

USA

Founded

1961

65 years active

Status

Public

Ticker · AVGO

Patents on file

USPTO grants

Funding history

Total raised

$648M

Rounds tracked

Latest stage

IPO

Date	Stage	Amount	Valuation	Lead investors
Aug 6, 2009	IPO	$648M	—	—

Stock performance

AI bull / bear

Bull case

Broadcom is the most credible counterbalance to Nvidia in AI silicon, and it is winning on two fronts simultaneously. On custom ASICs, its partnership with Google on TPU v4 and v5 positions it at the center of the largest alternative AI compute ecosystem, and the explosive growth of Anthropic — which trains on TPU pods — directly drives semiconductor and infrastructure demand through Broadcom's supply chain. On networking, Broadcom's Ethernet switching and SerDes are the backbone of AI cluster interconnects, giving it a critical foothold as hyperscalers seek to decouple from Nvidia's InfiniBand lock-in. The VMware acquisition adds a sticky software layer, with Broadcom now pitching Kubernetes on VMware VCF to bridge enterprise and cloud-native workloads. At $700B market cap, the thesis is that Broadcom becomes the horizontal plumbing layer across custom AI silicon, networking, and infrastructure software — diversified enough to survive any single customer loss, and leveraged enough to ride the AI capex supercycle.

Bear case

Broadcom's custom ASIC business is inherently capped by the design win cycle — each hyperscaler customer (Google, Meta) commissions chips for their own internal use, not for broad market sale. That makes Broadcom's AI revenue a function of its customers' willingness to vertically integrate rather than buy merchant silicon from Nvidia, a calculus that can shift overnight. If Google's TPU roadmap slows or Meta pivots back to Nvidia for next-gen training, Broadcom loses massive revenue with no ready replacement customer. Meanwhile, the networking advantage is under siege: Nvidia's acquisition of Mellanox gives it a captive interconnect, and startups like Enfabrica and Celestial AI are targeting the same AI fabric problem with more specialized silicon. VMware integration adds execution risk, not synergy — Broadcom's history of aggressive post-acquisition cost cutting could alienate the enterprise customer base that makes VCF sticky. The $700B valuation already prices in flawless execution across three distinct businesses, leaving no room for a single design win loss.

Generated by deepseek-v4-flash · June 2, 2026Refreshes quarterly · see disclosures

Broadcom

At a glance

Funding history

Stock performance

How they make money

Investability Index

AI bull / bear

In the news

Apple’s $30B Broadcom Deal Signals Expansions in AI, U.S. Supply Chain

Chip Industry Week In Review

Data Center AI Growth Faces Challenging Bottlenecks

Chip Industry Week In Review

The Wedding of the Year: Why AI Infrastructure Financing Is Becoming a Semiconductor Story

Broadcom and OpenAI unveil custom-built Jalapeño inference processor — OpenAI's first chip is a massive reticle-sized ASIC built in an ultra-fast nine-month development cycle

Watch

Leadership

Products

Custom AI Accelerators (XPUs)

Tomahawk & Jericho Ethernet Switches

VMware Cloud Foundation

Optical & PCIe/Connectivity Components

Most recent patents

Compare Broadcom with