① Chen Fuyang stated that the spending frenzy by large technology companies on AI will continue until the end of 2030; ② Last week, Broadcom's total Market Cap surpassed 1 trillion USD, with AI Business annual revenue increasing by 220% year-on-year to 12.2 billion USD; ③ Chen Fuyang previously revealed that Broadcom is developing AI Chips with three large clients.
According to Caixin, on December 20 (Editor Zhao Hao), semiconductor giant Broadcom's President and CEO Chen Fuyang recently stated that the spending frenzy by large technology companies on AI will continue until the end of this decade, i.e., the end of 2030.
Chen Fuyang told the media that his clients in Silicon Valley are "very hurriedly" making a three to five-year investment plan for AI infrastructure, "They are investing with full force, and they will only stop when their money runs out or when Shareholders stop them."
Last Friday, Broadcom's stock price soared 24.43% in a single day, and its total Market Cap historically surpassed the 1 trillion USD mark. The company announced a quarterly revenue of 14.054 billion USD the previous day, an increase of 51% compared to the same period last year, with AI Business annual revenue up by 220% to 12.2 billion USD.
In that day's Earnings Reports conference call, Chen Fuyang stated that Broadcom is currently developing AI Chips with three very large clients, and it is expected that by 2027, each client will deploy 1 million AI Chips in their network clusters.
At that time, Chen Fuyang did not reveal the names of the chip clients, but some Analysts said that Broadcom has collaborated with Google, Meta, and ByteDance to design custom processors, and has also helped Apple develop network technology for Server Chips.
Chen Fuyang
There were earlier speculations that Broadcom might acquire the struggling Intel, but Chen Fuyang stated that he has been "extremely busy" in the field of AI Semiconductors, "This requires me to invest a lot of resources and a lot of energy." He also added that he has "not been asked" to participate in Intel's Business.
Only when feasible will a Trade be reached. This means someone needs to come and ask me. Since Qualcomm, I have learned one thing: do not make hostile Quotes. Broadcom attempted to acquire Qualcomm for 142 billion dollars in 2018, but was blocked by the then president of the USA, Trump.
Chen Fuyang stated that he has been busy completing the acquisition Trade of VMware. Last year, Broadcom spent 69 billion dollars to acquire VMware, developing into a diversified Technology group. Chen Fuyang expressed an 'open attitude towards potential acquisitions in Hardware or Software.'
In September of this year, xAI’s 'Colossus' Datacenter in Memphis was put into Operation, equipped with 0.1 million NVIDIA GPUs, becoming a benchmark in the AI computing power competition. Chen Fuyang believes that by 2027, Broadcom's clients will build clusters with up to 1 million chips.
Although the value that AI can create for businesses is still inconclusive, Chen Fuyang stated that large Technology companies have seen great opportunities to generate more revenue in 'hyperscale Cloud Computing' services.
They need to train AI at an unprecedented scale, which will consume a large amount of silicon (Semiconductors). That's our area of expertise. They have a formula to continue doing this and have not reached the endpoint. All roads lead to: you need more computing Chips.
Chen Fuyang said that Technology companies are 'eager to invest huge amounts of money in the next 3 to 5 years because they believe this technology is within reach. Although these companies are few in number, they could be major customers for AI Chips because the returns are enormous.'
Some Analysts believe that even with 1 million Chips, it is not sufficient to achieve 'Artificial General Intelligence' (AGI). Chen Fuyang commented, 'I don’t think anyone knows, but the opportunity is just too hard to resist, so let’s give it a try.'
Regarding Broadcom entering the 'trillion Market Cap club' in the USA, Chen Fuyang feels this is not 'something new.' 'Value comes from the perspective of some onlookers; we must learn not to get too tangled in this... I think I'm not the only one who believes that generative AI still has a long way to go.'