Welcome to the forefront of conversational AI as we explore the fascinating world of AI chatbots in our dedicated blog series. Discover the latest advancements, applications, and strategies that propel the evolution of chatbot technology. From enhancing customer interactions to streamlining business processes, these articles delve into the innovative ways artificial intelligence is shaping the landscape of automated conversational agents. Whether you’re a business owner, developer, or simply intrigued by the future of interactive technology, join us on this journey to unravel the transformative power and endless possibilities of AI chatbots.
(Sept 3): Broadcom Inc predicted a boom in artificial intelligence (AI) chip sales over the next two years, helping renew optimism that it can challenge Nvidia Corp’s dominance in the lucrative market.
AI chip revenue will double to about US$115 billion (RM464.27 billion) in fiscal 2027 and soar to US$230 billion the following year, chief executive officer Hock Tan said during a conference call. The company also is on track to top earnings of US$30 a share in fiscal 2028, outpacing Wall Street estimates.
The stock rebounded from an earlier decline after Tan gave the guidance. It was little changed as of 6.31pm in New York.
Broadcom’s custom AI semiconductor business is thriving as companies like Alphabet Inc’s Google, OpenAI and Meta Platforms Inc look for ways to supplement their Nvidia-made AI chips with additional supply and variation.
Chatbot makers Anthropic PBC and OpenAI, which are racing to expand their AI infrastructure, have emerged as especially important clients for Broadcom. Anthropic is expected to become the biggest customer for Broadcom’s custom chips in 2027, displacing Google. Tan sees OpenAI becoming the second-biggest client for that business.
“We have six customers — four of them are just simply going to be huge,” Tan said. Demand for chips is higher than what Broadcom can ship, particularly due to supply shortages, he said.
The rapid expansion of data centres also has helped lift sales of Broadcom’s networking products. Capital budgets for the top five hyperscalers — a group that includes the biggest operators of data centres — have climbed about 40% to more than US$700 billion, “strengthening demand visibility for custom silicon and networking”, Bloomberg Intelligence analysts Kunjan Sobhani and Oscar Hernandez Tejada said in a note.
The longer-term outlook pleased investors, who were less impressed with Broadcom’s forecast for the fourth quarter. Revenue will be US$34.8 billion in that period, which runs through October, the company said. Though analysts predicted US$35.1 billion on average, some estimates topped US$36 billion, according to data compiled by Bloomberg.
The stock had been up 6.1% this year through the close, a performance that trailed the rallies enjoyed by many chip peers in 2026.
AI chip revenue alone will generate US$21.7 billion in the fourth quarter, Broadcom said. That narrowly topped the average estimate, with some projections well exceeding US$22 billion.
In the fiscal third quarter, which ended Aug 2, sales rose 86% to US$29.6 billion. Analysts had projected about US$29.5 billion on average. Earnings climbed to US$3.32 a share, excluding some items. That compared with an estimate of US$3.23.
AI semiconductor revenue was US$16.7 billion in the period, compared with an average estimate of US$15.9 billion. That category includes custom-built accelerators — the chips used to develop and run AI models — as well as networking semiconductors.
During the call, Tan said the company is accelerating work with Google, Anthropic and OpenAI. Broadcom will be shipping “multi tens of billions of dollars” worth of custom processors annually over the next several years to Google, Tan said. Measured by energy use — a benchmark of data centre capacity — Broadcom will send five gigawatts’ worth of the chips to Anthropic next year and a further 10 gigawatts the following year.
Broadcom also has “line of sight” to provide OpenAI with more than 5 gigawatts of custom chips in 2028, Tan said. Between now and the end of 2027, Broadcom will deliver three generations of chips to Meta, its fourth major customer.
The Palo Alto, California-based company has also set up substrate fabrication — a key step in chip production — at a plant in Singapore along with a partner, Tan said. Broadcom is investing more broadly in its factories, said Charlie Kawwas, the company’s semiconductor unit president, as it tries to ease some supply shortages.
Tan has positioned his company as a key alternative to the workhorse Nvidia chips that dominate the AI space. He poked at his rival on the call, saying that a new version of the chip it developed with Google is as good if not better than Nvidia’s new Vera Rubin line when running AI models.
Broadcom has been riding high on custom AI chip demand, both by signing deals and leading a charge to find a way to help companies like Anthropic pay for pricey semiconductor purchases. Tan has set up financing vehicles with Apollo Global Management Inc and Blackstone Inc to help Anthropic afford the Google chips that Broadcom helped develop.
The financing partnership, which is meant to back more than 20 gigawatts of computing power, will require hundreds of billions of dollars. That level of capacity would roughly equal the output of 20 nuclear plants.
But there also have been signs of mounting competition. Last month, Marvell Technology Inc announced a deal to help make custom semiconductors for Google. Broadcom and Google have long worked together to develop that company’s tensor processing units, or TPUs, which are chips used in AI data centres.
Uploaded by Chng Shear Lane
Copyright © 1999-2026 The Edge Communications Sdn. Bhd. 199301012242 (266980-X). All rights reserved