Broadcom Inc. (AVGO) experienced a decline in its stock price during Monday's trading session, even as JPMorgan analysts reaffirmed their confidence in the company's artificial intelligence chip partnership with Google. The stock traded at $396.72, down $14.63, as investors weighed the bank's defense against ongoing concerns regarding customer concentration and a broader pullback in high-valuation AI infrastructure equities.
The debate centers on whether Broadcom's AI narrative represents a sustainable long-term contract cycle or merely a series of quarterly chip orders. JPMorgan analysts Harlan Sur and Mayur Ramdhani advised clients to look beyond reports suggesting delays or cancellations in the next-generation TPU v9 2-nanometer program, stating there were "NO delays; NO cancellations," as reported by Benzinga on Monday.
This defense comes after a challenging June for Broadcom, which saw its shares fall 11% during the month. The decline followed guidance that some investors deemed insufficient for a stock priced for near-perfect execution, according to Barron's. Despite this, JPMorgan maintains an Overweight rating with a $580 price target.
Broadcom's recent financial results underscore the intensity of the debate. The company reported fiscal second-quarter AI semiconductor revenue of $10.8 billion, a 143% year-over-year increase, and forecast $16.0 billion in AI semiconductor revenue for the third quarter, representing growth of over 200%. CEO Hock Tan attributed this growth to custom AI accelerators and AI networking solutions.
Beyond Chip Design: The Rack Economics
The company's April SEC filing revealed more than just a commitment to develop and supply Google's future tensor processing units (TPUs). It also disclosed a supply-assurance agreement for networking and other components in Google's next-generation AI racks, extending through 2031. This shifts the focus from chip design to the economics of entire AI racks, where processors are just one part of the cost. The switching fabric, interconnect, power envelope, and delivery schedule are critical factors in turning capital expenditure into usable compute capacity.
Broadcom's June 9 platform with Apollo and Blackstone adds another dimension, with an initial $35 billion tranche for over 1 gigawatt of compute, aiming for more than 20 gigawatts by 2028. Tan described this as a way to meet AI demand with "speed and certainty," while Apollo President Jim Zelter called it a "bold, collaborative model."
Anthropic serves as a concrete example of this mechanism. The AI company announced in April that it had signed agreements with Google and Broadcom for multiple gigawatts of next-generation TPU capacity starting in 2027. CFO Krishna Rao called it Anthropic's "most significant compute commitment to date," noting that the company's run-rate revenue had surpassed $30 billion, up from approximately $9 billion at the end of 2025.
Competitive Pressures and Risks
Broadcom faces competition from Nvidia, which continues to dominate general-purpose AI accelerators, and Marvell Technology, which is also pursuing custom chip contracts. Reuters reported that Broadcom and Marvell help cloud companies like Amazon and Google design custom chips that generate tens of billions in revenue, offering an alternative to Nvidia hardware.
An emerging challenge comes from design-tool startups. Architect Labs recently raised $24 million to use AI to accelerate custom chip design, a process that can take roughly two years and cost hundreds of millions of dollars. While this is unlikely to displace Broadcom in the near term, it signals customer pressure to reduce lead times and supplier margins.
The primary risk for Broadcom remains customer concentration. If Google or other hyperscalers diversify their orders across multiple suppliers, Broadcom's revenue visibility could narrow, even if overall AI demand remains robust. Reuters noted that after Broadcom's June results, Marvell was gaining traction with hyperscale customers, and Broadcom's shares dropped after CEO Hock Tan left the $100 billion 2027 AI chip sales forecast unchanged. Creative Strategies analyst Ben Bajarin commented, "they just didn't raise it."
For now, the market views Broadcom less as a traditional chip supplier and more as an AI capacity broker—part silicon designer, part networking vendor, and part beneficiary of private-credit financing. This is a powerful position if the Google TPU roadmap stays on schedule, but it also creates concentration risk if buyers decide the best way to cut AI costs is to make Broadcom compete harder for every rack.
