Unlock Editor's Digest free of charge
Roula Khalaf, editor of the FT, picks her favorite stories on this weekly newsletter.
The spending spree by big tech corporations on artificial intelligence will proceed through the tip of the last decade, in accordance with the pinnacle of Broadcom, whose value has risen to a valuation of greater than $1 trillion as investors grow enthusiastic about its AI chip business .
Hock Tan, Broadcom's chief executive, told the Financial Times that his Silicon Valley customers were “in a rush” to develop three- to five-year AI infrastructure investment plans.
“They are investing heavily,” he said. “They will stop once they run out of cash or when shareholders put a stop to it.”
Tan's comments come after Broadcom's share price rose 24 percent in a single day last Friday after it announced its AI revenues rose 220 percent to $12.2 billion in fiscal 2024.
This increased its market capitalization by greater than $200 billion, making the chipmaker's market value exceed $1 trillion for the primary time. Tan told investors last week that Broadcom could generate tens of billions of dollars in additional annual revenue from AI chips by 2027.
Broadcom doesn't name its chip customers, but analysts say the Silicon Valley-based group has worked with Google, Meta and TikTok's parent company ByteDance to develop custom processors that speed up the training and deployment of AI systems.
OpenAI and Apple are also reportedly working with Broadcom to develop their very own AI server chips, as tech corporations search for alternatives to Nvidia, the $3 trillion chipmaker that dominates the marketplace for powerful processors needed to coach large language models grow to be.
Tan's many years of experience within the semiconductor industry and in making serial deals has led to speculation in Silicon Valley that Broadcom might step in to avoid wasting Intel, the troubled U.S. chipmaker whose Chief Executive Pat Gelsinger abruptly departed this month.
However, Tan downplayed the prospect of a Broadcom bid for Intel, saying he has “his hands full” with AI semiconductors. “This takes up numerous my resources and focus,” Tan said, adding that he “wasn’t asked” to get entangled with Intel.
“I can only make a deal if it is possible,” he said. “Practicability implies that someone comes and asks me. One thing I’ve learned since Qualcomm: no hostile offers.”
In 2018, Broadcom's $142 billion hostile takeover bid for chip rival Qualcomm was blocked by then-US President Donald Trump in an unprecedented intervention.
Tan was also busy completing Broadcom's integration of last 12 months's $69 billion acquisition of cloud software company VMware. Still, he said he was “open to potential acquisitions” in each hardware and software: “We're in consideration mode, so to talk.”
This 12 months, Big Tech corporations and AI startups like OpenAI and Elon Musk's xAI have made unprecedented investments in data centers to construct and operate ever-larger AI models.
xAI's “Colossus” facility in Memphis boasted 100,000 Nvidia graphics processors when it went online in September, setting a brand new benchmark within the race for AI computing power.
But by 2027, Broadcom's customers will construct clusters of as much as 1,000,000 AI chips, in accordance with Tan.
Although the worth of generative AI in saving regular corporations money is “not yet decided,” Tan said Big Tech's “hyperscalers” saw huge opportunities to generate more revenue.
“You should train (AI) on a scale the world has never seen before,” he said. “This uses huge amounts of silicon. That’s where we show up.”
Much of the recent progress in generative AI has been driven by the so-called scaling law, which states that combining more data with more computing power creates smarter AI.
“They have a formula to get on with this they usually will not be at the tip of that formula yet,” Tan said. “All roads result in this: you wish more computing chips.”
Tech corporations are “making very big bets over three to 5 years in a rush because they see the technology is inside their reach,” he added. “There are only a couple of of those players doing this, but they’re potentially very large consumers (of AI chips) since the returns are huge.”
Even 1 million chips is probably not enough to attain the last word goal of OpenAI and rival AI startup Anthropic: creating artificial general intelligence, or machines which might be smarter than humans. “I don’t think anyone knows,” Tan said. “But it’s too difficult to withstand this chance and say, let’s try.”
Although Broadcom has joined an elite group of just eight U.S. corporations valued at greater than $1 trillion, Tan said he feels “nothing recent.”
“Value is in the attention of the beholder. “You should learn to not get too caught up in it,” he said. “But it’s an incredible recognition. . . . I believe, and I’m not the just one, that generative AI still has a protracted option to go.”
