Broadcom’s AI chip business rises 106%, further accelerating strong earnings and sales

Artificial intelligence chipmaker Broadcom Inc. showed little sign of slowing after its latest earnings report beat Wall Street expectations and gave a solid outlook for the current quarter, sending shares up 4% in after-hours trading.

The company reported first-quarter adjusted earnings of $2.05 per share, beating analysts’ expectations of $2.03 per share, and revenue of $19.31 billion, up 29% year over year, beating consensus estimates of $19.18 billion. Net income for the quarter was $7.35 billion, up from $5.5 billion in the year-ago period.

Broadcom Chief Executive Officer Hock Tan (pictured) told analysts on a conference call that “AI revenue from chips alone is expected to exceed $100 billion by 2027.” “We have also secured the necessary supply chain to achieve this.”

Looking to the current quarter, Broadcom expected revenue of $22 billion, significantly beating the Street’s estimate of $20.56 billion. Additionally, adjusted profit margin is expected to be 68%, beating the consensus of 66%. Semiconductor Solutions revenue is expected to exceed $14.8 billion, the company said, slightly ahead of the Street consensus estimate of $13.06 billion.

Broadcom has experienced impressive growth thanks to its key role in helping companies like Google LLC, Amazon Web Services Inc., and Microsoft Corp. design and siliconize their own AI processors. We provide these companies with essential services, backed technology, and intellectual property. Once the design is ready, it is shipped to chip manufacturing plants operated by companies such as Taiwan Semiconductor Manufacturing Company.

Tan said the company’s AI chip-related revenue rose 106% year-on-year to $8.4 billion. This growth was driven by “robust demand for custom AI accelerators and AI networking,” he said. Broadcom also sells the network interconnects, software, and other components needed to link AI processors into large clusters.

Overall, Broadcom’s Semiconductor Solutions segment had revenue of $12.52 billion, beating analysts’ estimates of $12.25 billion. However, its Infrastructure Solutions business failed to meet expectations, with revenue of $6.8 billion, below the Street’s target of $7.02 billion.

Many large companies have come under pressure in recent weeks, with many concerns that AI models could pose a competitive threat to traditional enterprise software companies. The company’s stock price has fallen this year, with the iShares Expand Technology Software Sector Exchange Traded Fund losing 19% of its value so far.

But nothing seems to derail Broadcom, especially after acquiring server virtualization software giant VMware in 2023. Tan told analysts that this quarter was a flash in the pan, and that he expects the infrastructure software business’ revenue to rise 9% to $7.2 billion this quarter, beating the Street consensus. “We would like to emphasize that this growth in our software business reflects our focus and investment in our underlying infrastructure,” he emphasized. “Our infrastructure software will not be destroyed by AI.”

Andrew Rocco, an analyst at Zacks Investment Research, told SiliconANGLE that Broadcom’s acquisition of VMware has been successful so far and is fueled by growth in AI semiconductors. “Broadcom is just entering the early stages of a period of hypergrowth,” he predicted. “We believe investors will be drawn to this stock because regardless of which software company wins the AI ​​race, they are investing in Broadcom’s networking chips and custom accelerators.”

Tan offered a very rosy outlook for the custom AI chip business. He revealed in December that AI model maker Anthropic PBC had placed a $10 billion order for custom accelerators made by Google, and today provided a deeper perspective on the deal. He said he expects Anthropic to buy 1 gigawatt worth of Google tensor processing equipment in 2026 and more than 3 gigawatts in 2027.

Additionally, he said he expects OpenAI Group PBC to deploy more than a gigawatt of first-generation custom processors in 2027. Like hyperscalers, OpenAI is also working with Broadcom to develop its own AI chips. Meanwhile, Tan said he expects the company to get a big boost from Metaplatforms’ next-generation MTIA custom chips, even though some analysts have expressed concerns in recent weeks that development seems to have stalled.

“The MTIA roadmap is still intact,” Tan said, dismissing such concerns. He added that it is currently shipping and that Meta is targeting “multi-gigawatts” of custom accelerator capacity starting in 2027.

Photo: Sarbjeet Johal (Stackpane)/YouTube