Broadcom FY26 AI Revenue Expected to Exceed $56 Billion; Google TPU Roadmap and Long-Term Agreement Remain Intact
AI summary card
Broadcom FY26 AI Revenue Expected to Exceed $56 Billion; Google TPU Roadmap and Long-Term Agreement Remain Intact
J.P. Morgan believes Broadcom's FY26 AI revenue could exceed $56 billion, up more than 180% year over year, followed by 2.0x to 2.5x growth in FY27. The report states that design wins for Google's next four generations of TPUs and supply and revenue commitments through 2031 remain unaffected by competitive rumors, and reiterates its OW rating.
- FY26 AI revenue is expected to exceed $56 billion, up more than 180% year over year
- FY27 AI revenue is expected to grow 2.0x to 2.5x
- Design wins for Google TPU v8 through v11 are locked in by a five-year agreement
- The 3nm TPU v8i is ramping, while the 2nm TPU v9 is expected to ramp in early 2028
- Demand for Tomahawk 5 and Tomahawk 6 exceeds supply capacity both this year and next year
- J.P. Morgan believes the involvement of Alchip, GUC, AMD, and MRVL does not affect Broadcom's core TPU opportunity
Report interpretation
Overview
The report examines Broadcom's recent business trends and concerns about competition in Google TPUs. J.P. Morgan concludes that the company's FY26-FY27 AI growth trajectory, Google's roadmap for the next four generations of TPU designs, other customer XPU programs, and demand for AI networking chips all remain on track. The market is underestimating its long-term agreements, technological leadership, and execution track record, and the firm therefore reiterates its OW rating.
Core views
First, J.P. Morgan believes Broadcom remains on track to generate more than $56 billion in AI revenue in FY26, representing growth of more than 180% year over year, and to deliver 2.0x to 2.5x AI revenue growth in FY27. Growth is not limited to Google TPUs but also includes ASIC XPU programs for customers such as Meta, OpenAI, Anthropic, ByteDance, Alibaba, SoftBank/ARM, and SambaNova. Networking chips are another important driver: demand for Tomahawk 5 remains strong, while the next-generation Tomahawk 6 is sold out through next year. Demand for both products exceeds Broadcom's supply capacity this year and next year. Second, the report argues that the market has not adequately appreciated the five-year agreement signed by Google and Broadcom in April 2026. The agreement runs through 2031 and includes Google's commitment to increase its TPU-related revenue contribution to Broadcom each year, corresponding supply commitments from Broadcom, and a locked-in design-win roadmap covering four generations of products: TPU v8, v9, v10, and v11. Based on this agreement, J.P. Morgan believes Broadcom's medium- to long-term revenue visibility in the Google TPU business has not changed because of recent competitive developments. Third, the specific product roadmap continues to progress as planned. Broadcom is ramping Google's next-generation 3nm TPU v8i and expects to ramp the subsequent 2nm TPU v9 in early 2028. The report states that v9 uses 4 compute dies, 16 HBM stacks, and 400Gbps SerDes. Broadcom began designing its 2nm IP in the first half of the prior year and moved into SoC design in the second half; the program remains one of the company's highest-priority projects. The team has also begun early IP design for the subsequent TPU v10. J.P. Morgan explicitly rejects claims of CoWoS cuts or program delays or cancellations, maintaining that v9 remains on its original roadmap. Fourth, the report distinguishes between core TPU design and peripheral support projects. J.P. Morgan believes Alchip and GUC were primarily brought in to assist Google's underperforming in-house COT team, with their first project being TPU v8t. AMD and MRVL, meanwhile, are being evaluated to design attach support chips deployed alongside the TPU. These assignments differ from the core TPU ASIC design controlled by Broadcom, so the report concludes that these competitors' involvement will not erode Broadcom's core design wins or revenue opportunity. Fifth, Broadcom's XPU programs outside Google are also at various stages of design, ramp-up, or full production. The report cites Meta's 5 MTIA XPU ASIC programs, OpenAI's first- and second-generation Jalapeno XPUs, Anthropic's TPU v7/v8/v9, and related ASIC programs for ByteDance and Alibaba, SoftBank/ARM, Apple, and SambaNova, concluding that these programs broadly remain on track. The report also states that the recent Samsung supply agreement implies that Broadcom's cumulative AI revenue could exceed $1 trillion over the next five years. Finally, J.P. Morgan bases its assessment on Broadcom's long-standing ASIC expertise. The report states that the company has led the ASIC market for more than 30 years, with advantages extending beyond SerDes to expertise in compute, networking, memory, and storage architectures, a broad IP portfolio, and advanced chip and packaging design capabilities. J.P. Morgan estimates that Broadcom's lead exceeds 18 months and emphasizes that the company has helped Google launch 14 state-of-the-art ASIC chip designs over the past 12 years. The report believes this technology, design cadence, and execution track record constitute a durable competitive advantage that remains underappreciated by the market, and therefore reiterates its OW rating on AVGO. The rating history table lists the latest price target at $580, versus a share price of $362.48 on August 19, 2026.
Analysis framework
The report first uses recent primary research to verify the FY26 and FY27 AI revenue trajectory, then examines Google's long-term agreement, the progress of successive TPU generations, other customers' XPU programs, and Tomahawk supply and demand conditions item by item. It then distinguishes competitors' COT assistance and TPU peripheral-chip assignments from Broadcom's core ASIC design wins, and assesses whether competitive barriers have changed based on the company's track record in architecture, IP, advanced packaging, and historical delivery.
Methodology notes
Primary Research and Project Roadmap Verification
J.P. Morgan uses recent research findings to cross-check revenue targets, chip design stages, ramp timelines, and customer program status to determine whether external claims of delays, cancellations, and intensifying competition are valid.
AI Networking Chip Supply-Demand Verification
The report compares demand for Tomahawk 5 and Tomahawk 6 with Broadcom's supply capacity. Demand exceeding supply for two consecutive years and Tomahawk 6 being sold out through next year are used to support the AI revenue growth outlook.
Segmentation of Roles in the TPU Industry Chain
The report distinguishes core TPU ASIC design, support for Google's in-house COT efforts, and TPU-attach peripheral chips to assess whether the involvement of Alchip, GUC, AMD, and MRVL will affect Broadcom's core design wins and revenue.
ASIC Design and Execution Barriers
The report illustrates Broadcom's competitive advantage through its lead of more than 18 months, broad IP portfolio, chip and advanced packaging capabilities, more than 30 years of ASIC experience, and track record of helping Google launch 14 advanced ASICs over 12 years.
Long-Term Supply and Revenue Agreement Analysis
The report treats the five-year agreement signed by Google and Broadcom as a key event and assesses the company's business visibility through 2031 based on its annual revenue growth provisions, supply commitments, and locked-in TPU v8 through v11 designs.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- Broadcom Inc (AVGO.US)The report's core covered company, which is expected to benefit from Google's long-term TPU agreement, multi-customer XPU programs, and demand for AI networking chips.
- Strengths
- Leading-edge ASIC design, advanced chip and packaging capabilities, a broad IP portfolio, a lead of more than 18 months, and a long-term delivery track record.
- Comparison
- The report believes Alchip and GUC primarily assist Google's COT team, while AMD and MRVL mainly participate in TPU-attach support chips; none affect Broadcom's core TPU design wins.
Key data
- FY26 AI Revenue$56B+Expected to grow more than 180% year over year
- FY27 AI Revenue Growth2.0x-2.5xThe report expects another multiple-level increase in FY27
- Google Agreement Term5 years, through 2031Includes annual increases in TPU revenue and Broadcom supply commitments
- TPU Generations Locked In by the AgreementTPU v8, v9, v10, v11Covers the design roadmap for the next four generations of TPUs
- TPU v9 Process and Ramp Timeline2nm, expected in early 2028The report believes there is no delay or cancellation
- TPU v9 Configuration4 compute dies, 16 HBM stacks, 400Gbps SerDesNext-generation XPU ASIC design specifications listed in the report
- Meta MTIA Programs5 XPU ASIC programsAt various stages of design, ramp-up, or production
- Broadcom's Relative LeadMore than 18 monthsThe report's assessment of Broadcom's advantage in leading-edge ASIC design
- Google ASIC Execution Track Record14 advanced chip designs over 12 yearsUsed to support the assessment of Broadcom's long-term execution capabilities
- AI Revenue Implied by Samsung AgreementMore than $1T cumulatively over the next five yearsThe report's inference based on the recent supply agreement
- Current Share Price$362.48AVGO price on August 19, 2026
- Latest Listed Price Target$580OW price target dated June 4, 2026, in the rating history table
Impact & implications
The report believes Google's long-term agreement, locked-in designs for four generations of TPUs, other customer XPU programs, and supply-constrained Tomahawk products collectively enhance the revenue visibility of Broadcom's AI business through FY27 and even 2031. Competitors are currently focused mainly on supporting in-house teams or developing TPU peripheral chips and have not changed Broadcom's core design position. Combined with a technology lead of more than 18 months and a long-term execution track record, J.P. Morgan believes the market undervalues the durability of its AI business and therefore reiterates its OW rating.
What to watch
- Whether FY26 AI revenue can reach more than $56 billion and grow more than 180% year over year.
- Whether FY27 AI revenue can achieve the 2.0x to 2.5x growth projected in the report.
- The current ramp progress of the 3nm TPU v8i and whether the 2nm TPU v9 can ramp as planned in early 2028.
- Google's annual TPU revenue commitments through 2031 and execution of the TPU v8 through v11 design roadmap.
- Supply capacity and delivery progress for Tomahawk 5 and Tomahawk 6 this year and next year.
- The progress of XPU ASIC programs for Meta, OpenAI, Anthropic, and other customers from design to ramp-up and full production.