JPMorgan believes market concerns over the delay of Google's TPU v9 are unfounded, as the project remains on track with Broadcom and is scheduled for mass production in 2028. The five-year agreement signed in March this year has secured Broadcom's leading role across the next four generations of TPUs, extending revenue visibility through 2031. Google's in-house team still lags Broadcom by more than 18 months in technological capability and poses no near-term threat. As AI companies accelerate their shift toward custom chips, Broadcom's AI-related revenue is expected to grow 2 to 2.5 times by 2027.
In response to recent market rumors $Alphabet-C (GOOG.US)$ that the next-generation TPU v9 project has been delayed or even canceled, JPMorgan’s latest report explicitly refutes these claims, stating that $Broadcom (AVGO.US)$ the project, which is being developed by the responsible team, remains on track for mass production in 2028, with no unexpected changes as feared by the market. The report also notes that Google’s internal chip design team still faces a technology gap of at least 18 months before it could pose a real challenge to Broadcom.
More importantly, JPMorgan believes the market has overlooked a critical development: the five-year strategic agreement signed by both parties in March this year. This agreement not only secures Broadcom’s lead design role across the next four generations of TPUs (v8 through v11) but also includes escalating annual purchase commitments, extending revenue visibility for Broadcom’s AI business all the way to 2031.
According to JPMorgan, investors have overly focused on the potential competitive threat posed by Google’s in-house ASIC development while underestimating Broadcom’s substantial moat in advanced chip design, packaging technology, and intellectual property (IP) accumulation. As global large-model developers accelerate their shift toward custom AI chips, Broadcom is well-positioned to become one of the primary beneficiaries of this infrastructure wave. The report forecasts that Broadcom’s AI-related revenue will grow 2x to 2.5x year-over-year in 2027 and double again in 2028.
The TPU v9 project is 'neither delayed nor canceled.'
JPMorgan’s latest supply chain research indicates that Google’s next-generation AI training chip, the TPU v9, shows no signs of disruption. Broadcom remains responsible for its development, with mass production still slated for 2028—contradicting external rumors of delays or cancellation. Built on a 2nm process node, the chip integrates four compute dies, 16 stacks of HBM memory, and 400Gbps SerDes interfaces, forming a critical component of Google’s AI infrastructure.
In fact, Broadcom initiated IP design for the project as early as the first half of 2025 and transitioned to SoC development in the second half. The project has consistently held the highest internal priority at the company. Meanwhile, preliminary R&D for the next-generation TPU v10 has already begun. Judging from product planning and development cadence, the partnership between Google and Broadcom remains solid, and market fears of a rift are unfounded.
Five-Year Agreement Locks in Next Four TPU Generations
The report highlights that the market has currently overlooked a more crucial detail: in March this year, Google and Broadcom signed a new five-year cooperation agreement that locks in the roadmap for the next four TPU generations—TPU v8, v9, v10, and v11. The agreement includes long-term supply arrangements and, more significantly, revenue commitments from Google, providing exceptional certainty for Broadcom’s AI business.
JPMorgan estimates this implies Broadcom’s TPU-related revenue will continue growing year-over-year through 2031. Within today’s AI supply chain, very few companies have secured long-term purchase commitments from major large-model developers, and Google’s sustained investment undoubtedly opens a clearer growth runway for Broadcom over the coming years.
Google’s In-House Chips Unlikely to Challenge Broadcom in the Near Term
Over the past year, Google has continued to strengthen its Custom Silicon Team (COT) and brought in additional suppliers to participate in TPU development—a move that many investors view as Broadcom’s biggest future risk. However, JPMorgan believes this concern is significantly overstated.
The report notes that Google’s internal team is still optimizing its existing TPU v8t (Zebrafish) project, while Broadcom’s TPU v8i (Sunfish) completed certification in mid-2025 and is set to enter mass production soon. The two sides remain at least 18 months apart in terms of advanced AI chip design capabilities.
The report emphasizes that competition in ASICs is far from a contest based on a single technical metric; rather, it is a multidimensional battle encompassing computing architecture, network design, memory systems, advanced packaging, and IP portfolio—all areas where Broadcom has built core competitive moats through decades of sustained investment. The report specifically highlights that over the past 12 years, Broadcom has helped Google complete the design of 14 of its most advanced chips, and its experience and execution capabilities cannot be replicated in the short term.
Significant AI order backlog advantage
Broadcom’s strong client positioning in the AI ASIC market is also noteworthy.
Broadcom currently holds the world’s largest pipeline of AI ASIC projects, serving leading AI companies including Google, Anthropic, OpenAI, and Meta. It is also involved in SoftBank/Arm’s XPU project, SambaNova’s RDU project, and Apple’s AI CPU initiative.
As an increasing number of large AI model developers seek to reduce reliance on general-purpose GPUs and shift toward custom ASICs, Broadcom is emerging as one of the primary beneficiaries of this trend. Based on this, JPMorgan forecasts that Broadcom’s AI-related revenue will grow 2x to 2.5x year-over-year by 2027 and double again in 2028.
Advantage potentially undervalued: Advanced packaging and mass production barriers
Current market discussions have overly focused on whether Google will bolster its internal chip team, overlooking Broadcom’s true competitive advantages.
Whether in advanced chip design capabilities, packaging technology, IP portfolio, or large-scale manufacturing execution experience, Broadcom continues to maintain a clear lead. With the ramp-up of its advanced substrate facility in Singapore and the planned release of advanced packaging capacity by 2028, the company’s leadership position in the AI ASIC market could further widen in the coming years.
In JPMorgan's view, the market sees Google attempting to reduce its reliance on Broadcom, while overlooking the fact that Google’s next four generations of TPUs are already deeply tied to Broadcom. In the race for custom AI chips in the AI era, Broadcom remains one of the most difficult core players to replace.
Want to select stocks or get a stock diagnosis? Curious about the opportunities and risks in your portfolio? For all your investment questions,Just ask Futubull AI!
Editor/joryn