Broadcom Expands Custom AI Chip Plans Beyond Google; Nike Plunges and Oil Falls in Multi-Catalyst Friday

The Quick Take

• Broadcom (AVGO) is reportedly accelerating its artificial intelligence semiconductor roadmap, with plans to develop custom AI training chips (XPUs) for multiple hyperscaler clients beyond its existing Google TPU partnership.
• Nike (NKE) shares plunged on earnings disappointment, with North American wholesale weakness and slowing China recovery driving a miss against analyst estimates.
• Crude oil prices fell on the combination of the G7 strategic reserve release announcement and weaker-than-expected U.S. jobs data reducing demand growth assumptions.
• Friday's combination of macro catalysts — jobs miss, oil decline, Nike earnings, and Broadcom AI news — created a day of significant cross-asset repricing with broadly positive implications for technology equities.

Broadcom's reported acceleration of its custom AI silicon roadmap represents the most structurally significant technology business development in Friday's news flow. The company's existing relationship with Google — manufacturing the TPU (Tensor Processing Unit) custom AI training chips that power Google Cloud's AI infrastructure — has been among the highest-margin revenue streams in Broadcom's semiconductor portfolio. Reports suggesting Broadcom is extending this custom XPU (custom AI accelerator) model to additional hyperscaler clients would materially expand the total addressable market for its AI silicon business, potentially creating a competitive moat in custom AI chip manufacturing that complements NVIDIA's dominant position in off-the-shelf GPU supply.

The competitive dynamics of the AI silicon market are evolving rapidly in Broadcom's favor. NVIDIA's H100 and H200 GPUs command 6-12 month lead times due to manufacturing constraints at TSMC and CoWoS advanced packaging capacity, creating a sustained demand environment for alternative AI compute solutions. Hyperscalers that have experienced supply constraints purchasing NVIDIA GPUs — particularly Google, Meta, Amazon, and Microsoft — are accelerating their custom silicon programs precisely to reduce dependence on any single supplier. Broadcom's application-specific integrated circuit (ASIC) expertise, combined with its advanced networking silicon (Tomahawk, Jericho) that connects AI accelerators in large cluster configurations, positions the company as the vertically integrated custom AI infrastructure vendor that hyperscalers need.

The near-term risk for Broadcom is that custom silicon development timelines are long (2-3 years from design to volume production) and hyperscaler commitments are not guaranteed until silicon tape-outs proceed. Any report of a hyperscaler reducing its custom chip program — as has happened with some smaller cloud players — would weaken the Broadcom AI pipeline narrative. The longer-term opportunity, however, is compelling: if even two additional major hyperscalers formally commit to custom XPU programs with Broadcom, the company's AI revenue run rate could exceed $10 billion annually within three years, justifying substantial valuation multiple expansion from current levels. Monitor Broadcom's next earnings call for updated AI revenue guidance and any public customer acknowledgment of new XPU programs.

Synthesized from 1 source.