Broadcom's AI computing and networking growth path is becoming clearer; J.P. Morgan maintains Overweight

Overview

Based on an investor meeting between J.P. Morgan and Broadcom management, this report focuses on custom AI XPU demand, the expansion of the Apple partnership, AI networking demand, ASIC design capabilities, advanced packaging, the TPU v9 roadmap, OpenAI's next-generation XPU tape-out, and VMware's potential opportunities in on-premises enterprise AI deployment. The report is broadly positive, viewing Broadcom as the world's second-largest AI semiconductor supplier, largest custom chip ASIC supplier, and largest networking semiconductor supplier.

Core views

The core view is that Broadcom's AI computing and networking growth runway is becoming more certain. As inference workloads evolve from one-off inference to inference-based and agentic applications, customers are accelerating XPU tape-outs and increasing design variants. Under Google's five-year TPU agreement, revenue is expected to grow annually and Broadcom is expected to maintain a majority share of shipments throughout the agreement period. The Apple partnership has been extended through 2031 and expanded from RF to ASICs. The OpenAI Jalapeño project validates Broadcom's ASIC execution capabilities, while strong demand for Tomahawk 6 further strengthens its AI networking position.

Analysis framework

The report evaluates the quality of Broadcom's AI semiconductor growth through management interviews and investor meeting notes, examining customer demand, product roadmaps, supply constraints, the competitive landscape, gross-margin resilience, and long-term software opportunities. It combines rating history and disclosed information to form an investment view.

Methodology notes

• Fundamental researchManagement meeting notes and catalyst analysis

  Validating key points of debate through management commentary

  The report addresses investor concerns including whether TPU v9 has been delayed, the scope of the Apple agreement, progress on OpenAI's next-generation chip, the competitive threat from COT, Tomahawk 6 supply and demand, and ASIC gross-margin stability, and distills their implications for revenue growth and competitive barriers.

• Industry chain analysisSupply and demand framework for AI accelerators and networking chips

  Custom XPU adoption, AI networking scarcity, and customer entrenchment

  The report views custom XPUs, SerDes, advanced packaging, HBM integration, and Ethernet switch chips as interconnected capabilities. It believes success in compute chips can drive networking share, while scarce networking supply can also deepen customer relationships.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Broadcom Inc (AVGO / AVGO US)

  Core covered asset; the report maintains an Overweight rating

  Strengths

  Strong custom AI ASIC/XPU design capabilities, with SerDes, advanced packaging, Ethernet switching, and long-term strategic customer relationships; strong Tomahawk 6 demand; TPU v9 roadmap progressing as planned.

  Weaknesses

  The AI supply chain remains affected by input-cost and supply constraints involving wafers, substrates, HBM, and other components; the VMware on-premises AI deployment opportunity remains at an early stage.

  Comparison

  The report states that Broadcom is the world's second-largest AI semiconductor supplier, largest custom chip ASIC supplier, and largest networking semiconductor supplier.

  Risks

  COT insourcing, delays to key customer projects, AI computing demand below expectations, supply constraints, and rising input costs could affect growth or margins.

• Apple

  Broadcom's long-term ASIC and RF collaboration customer

  Strengths

  The partnership has lasted more than 10 years, and the agreement was extended and expanded through 2031, potentially to include AI/data center ASICs.

  Risks

  Changes in Apple's design cadence, product scope, or in-house development strategy could affect related revenue opportunities.

• OpenAI / Jalapeño

  Broadcom's jointly designed first-generation custom XPU accelerator project

  Strengths

  The initial design reached tape-out in just nine months, validating Broadcom's ASIC execution capabilities; the next-generation chip is expected to tape out soon.

  Weaknesses

  Frontier XPU complexity is high, with demanding requirements for bandwidth, SerDes, power consumption, and packaging.

  Risks

  Tape-out, mass production, performance, or customer deployment could fall short of expectations.

• Tomahawk 6

  Broadcom AI networking switch chip product

  Strengths

  Adopted by major frontier model customers, with demand significantly exceeding supply and XPU customers receiving priority allocation.

  Weaknesses

  Supply constraints could limit near-term deliveries.

  Risks

  Supply expansion, customer purchasing cadence, or changes in competing products could affect the timing of revenue.

Key data

• RatingOverweight (OW)J.P. Morgan reiterated its Overweight rating on Broadcom in the report.
• Current price$370.78The report disclosed the AVGO / AVGO US price date as 07 Jul 26.
• Most recent target price$580The rating history shows a target price of $580 on 04-Jun-26.
• Apple partnership termThrough 2031The agreement was extended and the ASIC design relationship expanded, while also including an RF supply agreement.
• Google TPU agreement5 yearsManagement said revenue is expected to increase annually and that Broadcom is expected to maintain a majority share of shipments during the agreement period.
• XPU/GPU mixCould approach 50/50 next yearManagement offered this view regarding leading frontier model providers, reflecting inference-driven custom XPU adoption.
• TPU v9Roadmap progressing as plannedManagement said the next-generation TPU project has not been delayed or canceled, and that 400G-per-lane SerDes has been developed and is operational.
• Tomahawk 6Demand exceeds supply; essentially sold outManagement said major frontier model customers are adopting Tomahawk 6, with XPU customers receiving priority in switch product allocation.

Impact & implications

For investment implications, the report reinforces the certainty of Broadcom's two-pronged AI computing and networking growth story. Custom XPU demand is being driven by the commercialization of inference, long-term relationships with customers such as Apple and Google reduce concerns about future share, the OpenAI project provides additional validation of ASIC execution, and Tomahawk 6 supply scarcity improves networking business visibility. If the non-AI semiconductor business continues to improve and VMware revenue is further released, revenue and earnings growth could receive additional support.

Risks

• Mature in-house COT tools at customers could pressure Broadcom's XPU/TPU share, although management believes this concern is overstated.
• Rising input costs and supply constraints for wafers, substrates, HBM, and other components could affect delivery schedules or gross margins.
• Delays in the tape-out or mass production of TPU v9, OpenAI's next-generation XPU, or other customer projects would reduce growth visibility.
• An oversupply of AI computing capacity or a slowdown in customer capital expenditures in 2027 or 2028 could affect order strength.
• The high concentration of large customers such as Apple and Google means that changes in agreement scope, share, or design cadence could affect revenue expectations.

What to watch

• Subsequent milestones and customer validation progress for TPU v9 ahead of the CY28 ramp.
• Whether OpenAI's next-generation XPU tapes out soon as management indicated, and the subsequent mass-production cadence.
• Supply improvements and customer allocation for Tomahawk 6, Jericho, and other AI networking products.
• Whether Broadcom's annual revenue growth and majority share under Google's five-year TPU agreement are realized.
• Whether the ASIC design scope under the Apple agreement through 2031 becomes clearer, particularly whether it includes AI/data center ASICs.
• Whether ASIC gross margins can remain stable amid rising HBM, substrate, and wafer costs.
• Commercialization progress for VMware in on-premises enterprise AI deployment.