PALO ALTO, California, April 20, 2026, 07:38 PDT
Broadcom’s race toward the $2 trillion mark hit a snag Monday. Reuters reported that Alphabet’s Google is negotiating with Marvell Technology on two fresh AI chips, hinting that one of Broadcom’s most valuable custom-silicon ties could be shifting. Broadcom slipped roughly 1.7% early in U.S. trading; Marvell jumped 4.3%.
Supportive takes are piling up just as rivals begin to surface. On Monday, a Forbes column pointed to Broadcom’s software roster as a key driver for its AI ambitions. Last week, Morningstar’s William Kerwin put the stock 24% below his fair value estimate. Over the weekend, Yahoo Finance pegged Broadcom as one of a few long-term standouts.
Some of that bullishness is grounded in actual contracts, not just upbeat talk. Broadcom announced April 6 it has secured a long-term deal with Google to develop and deliver custom AI chips through 2031. Anthropic, for its part, said it’s expanding its relationship with both Google and Broadcom, locking in multiple gigawatts of next-gen TPU capacity starting in 2027.
Meta locked in its Broadcom alliance through 2029, putting up an initial pledge topping one gigawatt in computing power. Mark Zuckerberg framed the deal as key to creating the “massive computing foundation” for “personal superintelligence to billions of people.” Hock Tan, for his part, described the launch as the first step in a “sustained, multi-generation roadmap.” Broadcom Inc.
But Broadcom’s focus isn’t just on chips. Last week, the company rolled out VMware Tanzu Platform agent foundations—software designed to help organizations push AI agents out of the lab and into controlled production within VMware Cloud Foundation. “A quick start” for bringing agentic concepts into real-world use on private cloud, is how Purnima Padmanabhan, who leads Broadcom’s Tanzu division, described the new offering. Broadcom News and Stories
That blend is key to Broadcom’s draw. For the first quarter, the company reported AI revenue up 106% to $8.4 billion, and it’s forecasting $10.7 billion this quarter. CEO Hock Tan told analysts Broadcom can already see a path to more than $100 billion in AI chip revenue by 2027. D.A. Davidson’s Gil Luria called that kind of visibility a sign of “significant growth in demand.” Reuters
A big chunk of the fight centers on ASICs—those application-specific integrated circuits built for specific tasks, unlike Nvidia’s broader AI chips. Google relies on Broadcom’s TPUs already, both for training its models and powering inference, that step where the AI spits out answers. But on Monday, Reuters reported Google has been in talks with Marvell about an inference-oriented TPU, plus a new memory processing unit to go with it.
Broadcom’s outlook isn’t as rosy as some of the hype suggests. Marvell keeps gaining ground as a legitimate custom-chip contender, Nvidia maintains its grip on AI accelerators, and AMD is pushing for the same enterprise contracts. Back in March, Reuters noted that Broadcom’s stated customer volumes had started to rival the size of AI deals chalked up by Nvidia and AMD.
Broadcom’s real challenge isn’t an imminent Google exit. The larger issue: major cloud players are getting better at splitting chip design work among different vendors, squeezing margins and ramping up pressure to deliver. Russ Mould, investment director at AJ Bell, pointed out that Nvidia’s competitors are bound to target this market, adding, “It also makes sense for customers to diversify their sources of supply, if they can, so they can spread technological and supply chain risk.” Reuters
Investors haven’t pulled back from Broadcom just yet. Late last week, the company’s own investor site pegged its market cap at roughly $1.88 trillion. Yahoo Finance, on Friday, had that number closer to $1.92 trillion—near enough to $2 trillion that even minor moves could tip Broadcom over or under the mark. Shareholders are scheduled to gather later Monday in Palo Alto.
