Broadcom soars as Wall Street sees the possibilities of custom AI revenue ramp

Broadcom’s (NASDAQ:AVGO) quarterly results may have topped estimates, but its guidance of up to $90B in revenue from custom artificial intelligence chips by 2027 seems to have sent Wall Street into a frenzy.

Shares rose 17.5% in premarket trading, while other semiconductor stocks, like Marvell Technology (MRVL), Nvidia (NVDA) and AMD (AMD) also saw healthy gains.

“AVGO has been communicating that customers are looking to deploy 1M xPU clusters for some time, but the confidence level seems higher on this call,” Jefferies analyst Blayne Curtis wrote in a note to investors. “This should bring back interest in the name, even with current AI trends more C2H25 weighted.”

Curtis reiterated his Buy rating and upped his price target to $225 from $205 after the results and guidance.

On the company’s earnings call, Chief Executive Officer Hock Tan said the serviceable addressable market for custom AI accelerators, or XPUs, will be “in the range of $60 billion to $90 billion in fiscal 2027 alone.”

That $60B to $90B is from the company’s three AI application specific integrated circuit customers — which includes Google (GOOG) (GOOGL) and Meta Platforms (META) — but it could rise further if the company were to win the business of the other two hyperscalers, Microsoft (MSFT) and Amazon (AMZN), Curtis added.

Relief and unbridled enthusiasm

Morgan Stanley analyst Joseph Moore also upped his price target after the outlook was seen as a “relief.”

“In-line quarter and outlook likely is a relief vs. fears of worse around Google transition, while the longer term AI commentary will add to longer term enthusiasm,” Moore wrote in a note. “Expect momentum to build from here over the course of CY25.” Moore reiterated his Overweight rating and raised his price target to $233 from $180.

Evercore ISI analyst Mark Lipacis also expressed enthusiasm for the guidance and said investors are likely “underestimating” Broadcom’s custom AI chip opportunity.

In addition to the $90B figure, Tan said AI-generated synthetic data is likely to be a “multiplier” on top of data generated by humans that will eventually go into large language models, which suggests the tech industry is not even close to scaling the limit of large language models.

“We share Hock’s view on this topic,” Lipacis added. He reiterated his Outperform rating and upped his price target to $250 from $201.