By Zaheer Kachwala

(Reuters) -Broadcom forecast fourth-quarter revenue above Wall Street estimates on Thursday, banking on strong demand for its custom AI chips as enterprise clients invest heavily in data centers and machine learning applications.

The company has been one of the biggest beneficiaries of the generative artificial intelligence boom, with hyperscale customers, including Google, adopting its custom accelerators due to their ability to process large quantities of data quickly.

Its chips have also emerged as an alternative to Nvidia’s costly processors as companies look for hardware that can be tailored to specific needs as AI becomes increasingly advanced.

“We expect growth in AI semiconductor revenue to accelerate to $6.2 billion in Q4,” said Broadcom CEO Hock Tan. AI revenue grew 63% to $5.2 billion in the third quarter ended August 3.

Investors have placed high expectations on AI-focused chip firms such as Broadcom as they bet on its strong market position and expanding product lineup to fuel rapid growth.

This year the company has launched the Tomahawk Ultra networking chip and a next-generation Jericho networking chip to speed up AI compute, as part of its push to challenge Nvidia’s dominance in the AI semiconductor industry.

“Broadcom’s AI-related semiconductor demand continues to shine. Results were driven by both AI interconnectivity, along with its custom ASIC business,” said Summit Insights analyst Kinngai Chan.

Application-specific integrated circuit, or ASIC, is a chip designed for specific tasks.

Broadcom shares were up 3% in extended trading. They have risen over 30% so far this year, building on a more than threefold surge in the past two years.

The company expects fourth-quarter revenue to be around $17.4 billion, compared with analysts’ average estimate of $17.01 billion, according to data compiled by LSEG.

Its third-quarter revenue of $15.95 billion beat the estimate of $15.83 billion.

(Reporting by Zaheer Kachwala in Bengaluru; Editing by Shilpi Majumdar)