Broadcom closed out its fiscal year by handing investors exactly the kind of headline they wanted: a beat on the quarter and a bullish, AI-fueled forecast. The chip-and-software giant reported $18.02 billion in fourth-quarter revenue—comfortably above Street estimates—and told Wall Street that specialized AI semiconductor revenue is set to surge to $8.2 billion in the fiscal first quarter.
Hock Tan, Broadcom’s CEO, framed the number as confirmation that the company’s push into custom accelerators and high-speed networking silicon is paying off. On the call he said Broadcom added a fifth customer for its custom AI chips; the company also confirmed that Anthropic was the previously mysterious fourth customer placing large orders tied to Google’s latest TPU platform.
The quick math
- Adjusted EPS: $1.95 (vs. $1.86 expected)
- Revenue: $18.02B (vs. $17.49B expected)
- Net income: $8.51B, up roughly 97% year over year
- Fiscal Q1 revenue guide: about $19.1B (well above the Street average)
Broadcom said semiconductor revenue tied to AI climbed 74% year over year during the quarter—part of the company’s Semiconductor Solutions segment, which reported $11.07 billion in sales. Its Infrastructure Software business, which includes VMWare, also kept up strong growth, at $6.94 billion.
The color behind the numbers matters. Broadcom isn’t just selling off‑the‑shelf chips: it’s designing custom ASICs with hyperscale customers and shipping networking gear—Tomahawk and Jericho families—that move massive amounts of data inside AI data centers. That mix explains the outsized margins and why the company expects AI-related semiconductor sales to roughly double year over year.
There are still interesting notes in the fine print. Management disclosed a $10 billion order tied to Anthropic and Google’s TPU deployments from earlier in the year, and said a newly added fifth customer has placed roughly a $1 billion order slated for delivery in late 2026. Investors cheered the numbers initially in after‑hours trading before the stock cooled off—an indication that even bullish results are being parsed for sustainability and timing.
Why this matters beyond Broadcom: the industry is remaking its buying patterns. Cloud providers and AI builders are increasingly comfortable commissioning custom accelerators and alternative designs to Nvidia GPUs. Broadcom’s networking portfolio also benefits from the demand to shift data across ever-larger clusters. The company’s recent collaboration with OpenAI and repeated confirmations of big hyperscaler engagements underline how the AI boom is reshaping supplier relationships.
Companies plotting exotic expansions—everything from densifying on‑prem racks to speculative concepts like orbital data centers—are part of the broader backdrop for rising infrastructure demand. That long‑term buildout is one reason why developments such as Google’s Project Suncatcher or advances in large‑scale model deployments like Gemini’s Deep Research feed into the same narrative: models and the places that run them both need specialized silicon and high‑speed networking.
Broadcom also sweetened the shareholder pot with a dividend bump to $0.65 per share and reiterated that it sees multi‑year opportunities across custom chips, data‑center networking and infrastructure software. Still, questions remain about product cycles, competition from GPUs and other custom accelerator suppliers, and how much of the current spending is front‑loaded by hyperscalers.
For now the company sits in a privileged position: a deep engineering bench, sticky hyperscaler deals and a business mix that spans both silicone for compute and the software that manages infrastructure. The next few quarters will tell whether Broadcom can convert guidance into sustained growth rather than a one‑time rush tied to the first waves of AI deployment.
Broadcom investor relations has the full release and filings for readers who want the granular numbers and management commentary.