Designs custom AI accelerators (TPUs for Google, others) and high-speed networking chips critical for connecting AI data centers. $50B+ annual revenue diversified semiconductor giant with AI-related products now driving ~40% of revenue. Focused on custom silicon ASICs for hyperscalers and the networking infrastructure that connects GPU clusters in massive training facilities.
Company Profile
The behind-the-scenes giant — builds custom AI chips for Google and connects GPU clusters together.
Key Products & Platforms
Google TPU (custom)
Custom AI ASICDesigns Google's Tensor Processing Units
Meta MTIA (custom)
Custom AI ASICCo-develops Meta's custom AI silicon
Tomahawk / Memory
Networking SwitchHigh-speed data center switching
Jericho3-AI
AI Fabric ChipConnects thousands of GPUs in AI clusters
VMware
SoftwareCloud infrastructure software platform
Key Customers
Competitive Position
Market Share
~5% AI accelerators (via custom ASICs), #1 in networking silicon
Competitive Moat
Deep co-design partnerships with hyperscalers; irreplaceable networking IP
Key Risk
Custom ASIC revenue depends on a few large customers; high concentration risk
Why This Company Matters
Broadcom is the secret infrastructure play. While NVIDIA gets the headlines, Broadcom builds the custom chips Google uses and the networking fabric that connects massive GPU clusters. No Broadcom, no large-scale AI training.
Key Milestones
Original Broadcom Corp. founded August by Henry Samueli (UCLA professor) and Henry Nicholas in Irvine, CA; focused on broadband communications ASICs for cable modems, set-top boxes and Ethernet switching.
Original Broadcom Corp. went public on NASDAQ at $24/share; became one of the highest-flying networking-chip stocks of the dot-com era at peaks of $250+ in 2000.
Avago Technologies (predecessor of today's Broadcom Inc.) IPO'd on NASDAQ at $15/share after spinning out of HP/Agilent in 2005; this Avago entity later acquired the original Broadcom.
Avago acquired CyOptics for $400M, signaling a shift toward optical/datacom components and away from a pure-RF/analog focus.
Avago closed .6B acquisition of LSI Corp August; deal made Avago a top-10 chip supplier with Symbios Logic storage controllers and PCIe switching IP.
Avago acquired LSI Corp for $6.6B, gaining storage and networking ICs; the deal foreshadowed Hock Tan's M&A roll-up strategy that would consolidate the analog/connectivity industry.
Avago announced $37B acquisition of Broadcom Corp; the combined entity took the Broadcom name with Hock Tan as CEO, creating one of the largest semiconductor companies by revenue.
Closed Avago-Broadcom merger February 1; combined company immediately ranked top-5 globally in semiconductor revenue at ~$15B/year and pivoted to growth-by-acquisition.
Tomahawk 3 (BCM56980) Ethernet switch chip launched at 12.8 Tbps; the silicon became the de-facto leaf-and-spine standard in hyperscaler fabrics, including the AI back-end network fabric.
Launched $130B hostile takeover of Qualcomm, the largest tech-M&A bid ever proposed; Trump administration blocked the deal in March 2018 over national-security concerns related to 5G.
Began Google TPU v2 partnership announced via custom ASIC engagements; Broadcom became Google’s exclusive TPU implementation partner for v2 through v6.
Closed .9B acquisition of CA Technologies, Hock Tan’s first software acquisition; pivoted Broadcom from a pure chip company to a semi+software hybrid.
Announced $61B agreement to acquire VMware May 26, the largest enterprise-software deal ever; required 18 months of regulatory review across US, EU, UK and China.
Closed $61B VMware acquisition November 22 after 18 months of regulatory review; expanded into infrastructure software and immediately raised VMware enterprise prices ~3x, sparking customer revolt.
Disclosed AI revenue run-rate above $12B for FY2024, anchored by Google TPU v5/v6 and Meta MTIA wins; AI segment doubled YoY to 24%+ of total semiconductor revenue.
Crossed $1T market cap December 13 on AI revenue surging 220% YoY to $12.2B, driven by Google TPU and hyperscaler ASICs; only the eighth US company to reach the milestone.
AI revenue topped $4.4B in Q2 FY2025, up 46% YoY on accelerating Google and Meta custom-silicon shipments; AI segment now 30%+ of total semi solutions revenue.
Reported $10B custom AI silicon order, widely attributed to OpenAI for inference ASIC starting in 2026; the deal vaulted Broadcom into a structural top-3 AI chip vendor.
OpenAI and Broadcom announced October 13 a 10GW custom AI accelerator collaboration to deploy through 2029; deal formalized the $10B order disclosed in September earnings.
Debuted TPU v7 (Ironwood) co-developed with Google on TSMC 3nm, 192GB HBM3e and 9.6 Tbps ICI bandwidth; Google deployed Ironwood for Gemini 3 inference at scale.
Meta announced co-development partnership with Broadcom for next-generation MTIA accelerators; deal extends Broadcom's hyperscaler ASIC franchise alongside Google TPU and OpenAI.