In November 2017, I wrote a cover story for The Edge Singapore (Issue 806, Nov 20, 2017), which explained how Penang-born Tan Hock Eng was moving a second-tier semiconductor firm Broadcom’s domicile from Singapore to the US and mounting a US$105.3 billion takeover of rival chip firm Qualcomm to create a company that could dwarf then giant Intel. “Can he pull off the largest tech deal in history?” the story asked, and then he laid out how Tan’s acquisitive ways might change the global semiconductor industry.
The week I wrote the story, Tan was in the White House, praising President Donald Trump, who was only in his ninth month in office, for his sweeping tax cut proposals. At the time, few CEOs were audacious enough to be in the White House lauding the then embattled President. Tan teared up as he recounted his journey as a skinny teenager from Penang who had risen to head a successful global chip player. “My mother could never have imagined that one day her son would be in the Oval Office standing beside the president of the United States,” he said in a voice choked with emotions. “And my mother, too,” Trump said as he put an arm around Tan.
I described Broadcom as “increasingly one of the most-feared technology companies in the world” in 2017. It was, at the time, basically a key supplier to Apple for WiFi and Bluetooth chips, as well as other chips and sensors used in iPhones, and had a market value of about US$110 billion. It was trying to gobble up rival chipmaker Qualcomm, which makes cellular baseband chips, through a hostile takeover.
Yet, less than four months after his Oval Office appearance, White House aides had scuttled Broadcom’s attempt to takeover Qualcomm in part because Broadcom at the time was still legally based in Singapore and also because it was seen as a cost cutter that chopped off research and development expenditure of firms it bought to help quickly payoff for the acquisition. Qualcomm, at the time, was battling to ward off Huawei on 5G technology, and US officials worried that under Broadcom, the cellular baseband giant might become an also-ran.
Seven years on, Broadcom is now the ninth largest company on Earth with a market capitalisation of US$1.07 trillion ($1.4 trillion), ahead of giant chip foundry player Taiwan Semiconductor Manufacturing Co. and Warren Buffett’s investment firm Berkshire Hathaway.
See also: China discusses sale of TikTok US to Musk as one possible option
Just to give you some perspective. Back in 2017, Microprocessor giant Intel, at the time the world’s largest chip firm, had a market value of around US$225 billion or more than twice the size of both Qualcomm and Broadcom. Today, Intel is worth just US$85 billion. And Qualcomm, the chip firm that Broadcom coveted so much back then, is worth US$185 billion. Video gaming chip maker Nvidia, the other major chip player, was worth US$120 billion back in November 2017. Now, as the dominant artificial intelligence, or AI, chip player, Nvidia is worth US$3.4 trillion, just behind Apple, the world’s most valuable firm.
Remaking Broadcom into AI play
So, what’s behind the rise and rise of Broadcom, and could it become a serious challenge to Nvidia? Look further than the ongoing AI boom for the secret to Broadcom’s recent success. Last month, Broadcom reported that revenues for its fiscal fourth quarter ending October 2024 rose 51% year-on-year to US$14.1 billion. Revenues from its semiconductor solutions business grew annualised 12% to US$8.2 billion that quarter, while its infrastructure software tripled to US$5.8 billion, driven by its November 2023 acquisition of virtualisation software firm VMware. In the last fiscal year, Broadcom’s revenues grew 44% to US$51.6 billion, including US$12.2 billion in AI-related revenues, which were up 220% over the previous year.
See also: The future of neurotechnology: merging minds and machines
Broadcom guided AI revenue of US$3.8 billion for its current quarter, up an annualised 65%. Last month, CEO Tan highlighted his firm’s massive opportunity in AI, forecasting that it could have AI semiconductor revenues of between US$60 and US$90 billion by the fiscal year end of October 2027.
Here’s why that’s important: Broadcom’s entire business last year, or total revenues, was US$51 billion. Essentially, what the company was saying was that two segments of their business, one of which was non-existent two years ago, will rake in revenues of US$60 to US$90 billion in just three years’ time. Add to AI revenue surge of over US$25 to US$30 billion, and most analysts believe Broadcom’s infrastructure software business segment is likely to bring in $7 billion, their wireless sales business segment is expected to rake in and at least US$15 billion in revenues from Broadcom’s non-AI server storage and broadband business segments.
Only one company, AI chip giant Nvidia, has had such exponential growth in recent memory. Little wonder, then, that Broadcom shares soared over 24% the day after the early December earnings announcement and a whopping 40% the week after that.
Nvidia’s shares fell in the aftermath of Broadcom’s earnings as investors digested what CEO Tan was saying. Over the past four weeks, both stocks have see-sawed. Both are currently down around 8.5% from their peak last month. Broadcom shares currently trade at 35.8 times this fiscal year’s estimated earnings, while Nvidia stock trades at 34.7 times forward earnings. In comparison, iPhone maker Apple’s shares trade at 32.1 times this year’s forecast earnings, while Microsoft trades at 32.3 times forward earnings. Make no mistake, these are high valuations, but then again, these are global tech juggernauts that have shown they can grow revenues and earnings year after year.
Here is a quick history of how Broadcom became the trillion-dollar giant it is today. US private-equity group Silver Lake Partners acquired the semiconductor division of a Hewlett Packard spinoff called Agilent Technologies, which had an old plant in Singapore to form Avago Technologies in 2005. Silver Lake hired Tan to lead Avago in 2006. Tan, an electrical engineering graduate of MIT who also has an MBA from Harvard Business School, once ran Malaysia construction materials firm Hume Industries before he joined a small private equity Pacven in Singapore. Tan shepherded Avago to its Nasdaq listing in 2009. Avago bought Broadcom in 2015 for US$37 billion, at the time the biggest tech deal ever, and renamed itself after its target but kept its Nasdaq ticker symbol AVGO.
Tan is a serial acquirer. Few CEOs have done as many deals as he has and successfully integrated the target companies. Under Tan, Avago and later Broadcom made eight major acquisitions — five chip firms and three software companies. In April 2013, he bought CyOptics for US$400 million. In December 2013, he purchased LSI Corp for US$6.6 billion. In early 2015, Avago’s merger with Broadcom was consummated. In December of that year, Broadcom bought Brocade.
After Trump scuttled the Qualcomm deal, Tan realised it would be difficult to buy a large chip company even after Broadcom moved its legal headquarters to the US in 2018. So, he twisted his strategy and started buying infrastructure software companies. Months after the Qualcomm deal was blocked, Broadcom swooped in to buy software firm CA Inc. for US$18.95 billion. In 2019, Broadcom acquired cyber security software firm Symantec for US$10.7 billion. Then, in 2022, Broadcom purchased virtualisation software firm VMware for US$61 billion. That merger wasn’t completed until late 2023.
Sink your teeth into in-depth insights from our contributors, and dive into financial and economic trends
Tan’s modus operandi has been simple: buy unloved chip firms, or more recently enterprise software companies, ruthlessly whip them into shape by cutting costs, firing staff, and, as the White House aides pointed out when they helped block its takeover of Qualcomm, even cut research and development expenditure which allows tech firms to keep their edge. Investors love Tan because he has maintained a steady quarterly dividend and aggressively bought back shares. “Broadcom adds value by re-focusing businesses on their core franchise with rigorous resource allocation, discipline and offering an efficient common platform,” notes New Street Research tech hardware analyst Pierre Ferragu.
Broadcom now finds itself at the epicentre of one of the world’s fastest-growing markets — AI chips. Two years ago, graphics chip maker Nvidia, which in 2017 was about the same size as Broadcom and Qualcomm, suddenly catapulted to one of the world’s two largest companies after Open AI introduced the ChatGPT chatbot.
Hyperscalers like Microsoft, Alphabet’s Google, Amazon, as well as Meta Platforms, and Elon Musk’s xAI, rushed to buy bucketloads of AI chips to train their large language models. Nvidia currently has an 85% to 90% share of the AI chip market. It can’t meet the insatiable needs of its customers. In recent months, Nvidia has rolled out its next-generation AI chip, Blackwell, which is only now being ramped up.
If Nvidia has a 90% market share of AI chips and can’t meet even half of its customers’ current needs, why aren’t other chip makers rushing to fill the void? Chipmakers like AMD are trying, but their chips aren’t as good as Nvidia’s. Now Broadcom and, to an extent, Marvell Technology are trying to fill the vacuum. Marvell has a Southeast Asian connection as well. Its Indonesian founder, Sehat Sutardja, who died in September, led the chip firm until 2016, when an accounting scandal forced him and his wife out of the firm.
Broadcom isn’t competing with Nvidia the way AMD was trying to make a GPU or graphic processing unit that was almost as good. It has taken a different route. Hyperscalers like Microsoft, Google, and Meta want more chips, and they want them now. They can’t wait for another year for Nvidia to give them more chips. So, they are working with Broadcom and Marvell to give them some chips that they can use in the interim for their own specific AI-related tasks. Broadcom and Marvell make XPUs that are next-generation custom accelerators and compute ASICs, or application-specific integrated circuits, or chips.
Think of ASICs and accelerators as customised chips made for specific tasks. Let’s say you are Meta’s CEO, Mark Zuckerberg, and you need more AI chips. You wine and dine Jensen Huang, and he says, “Mark, I have already given you all the chips I can, and you need to wait until December for more.” Zuckerberg can’t afford to wait 11 months for Nvidia GPUs. He just needs something. Or else Amazon, Elon Musk, Google or Microsoft will leave Meta far behind.
So he cajoles Tan to make him customised accelerator silicon or tailor-made application-specific integrated circuits to get around his problem. ASIC chips are less flexible than a GPU and certainly a lot less powerful but much more cost-efficient for more narrow tasks. So the hyperscalers are building their own versions of those chips, and Broadcom and Marvell are helping them out. Broadcom’s custom accelerators and Marvell’s ASIC chips are a temporary solution. They are also far cheaper than Nvidia’s top-of-the-line Blackwell chips.
Going forward, Broadcom will benefit from the relationships it forges with companies like Meta, Google, Amazon’s AWS, Apple and others. The AI ecosystem is likely to grow over time, so Broadcom won’t be taking business away from Nvidia but will only be catering to the needs of hyperscalers that Nvidia can’t attend to.
What’s next for Tan? At 72, he is one of the oldest tech executives, 18 years older than Apple’s Tim Cook, and 32 years older than Zuckerberg of Meta Platforms, whose board he joined last year. He puts more hours at work than CEOs 30 years younger. For now, Tan has no plans to retire anytime soon. The board last year extended his contract. He also has no plans to acquire troubled Intel, the former chip industry leader. Ten years ago, he probably would have jumped at the opportunity. As the head of the world’s number two supplier of AI chips, a fading former giant chipmaker just doesn’t interest him.
Assif Shameen is a technology and business writer based in North America
