Broadcom Earnings on the Horizon: Three Key Things to Watch
Claire Weston
Broadcom reports Q2 fiscal 2026 earnings after the close on June 3. The market is focused on three things: custom chip client expansion, the pace toward a $100 billion AI-chip revenue target, and AI networking's rising share — each one reshapes the AI chip competitive landscape.
Custom chips — what are Broadcom's biggest clients doing?
Broadcom's AI revenue depends heavily on six core clients — Google, Meta, OpenAI, Anthropic, and others. This means → this is not a broad-market chip seller; any move by a top client directly lifts or drags the numbers.
The Google partnership is nearly a decade old, centered on co-developing TPUs — tensor processing units, chips purpose-built for AI training and inference. Google recently launched TPU 8t (training) and 8i (inference) and began selling access to select customers.
In April, Anthropic announced an exclusive deal with Google and Broadcom for multi-GB-scale compute. The same month, Broadcom extended its partnership with Meta to co-develop MTIA — Meta's training and inference accelerator.
In plain terms = Broadcom doesn't sell off-the-shelf chips. It custom-builds AI silicon for its biggest clients. The key question for investors: is that custom circle still growing?
The $100 billion target — is Hock Tan's ambition credible?
CEO Hock Tan predicted on the Q1 earnings call that Broadcom would hit $100 billion in AI-chip revenue by fiscal 2027.
Current pace: Q1 semiconductor revenue was $8.4 billion; Q2 guidance is $10.7 billion. This means → quarterly revenue needs to leap from the low tens of billions to a hundred-billion annual run rate. Growth must keep accelerating.
Tan also argued that AI data centers are too dependent on GPUs — graphics processors, the current workhorse for AI training — and predicted Broadcom's custom chips would gradually surpass GPU designs.
This reflects a "de-Nvidia" narrative Broadcom is building. The market will use this earnings call to test how far that narrative has actually come — and how much real share Nvidia stands to lose.
AI networking — the underestimated second growth engine?
In Q2 guidance, AI networking revenue is expected to reach 40% of total AI revenue — the top end of a 30%–40% range. Based on $10.7 billion in quarterly guidance, AI chips contribute roughly $6.4 billion and AI networking about $4.3 billion.
Broadcom's core networking products are the Tomahawk switch and Jericho router — the hardware that moves data inside large-scale AI clusters. The next-generation Tomahawk 7, expected in 2027, will double switching bandwidth versus the current generation.
In plain terms = AI doesn't just need chips that compute fast — it needs networks that move data fast. Broadcom holds significant share on that track, but Nvidia's NVLink and NVLink Switch are expanding aggressively. The head-to-head competition is intensifying.
Content is for reference only, not financial advice.