No one in Silicon Valley wants to be called a “chip company” these days. Qualcomm, which makes the processors found in many Android phones and iPhone modems, describes itself as a “platform company”. Intel, once proudly “inside” most of the world’s PCs, now positions itself as a “data company”. Nvidia, whose soaring share price has made it one of the best-performing stocks of any sector in the past two years, describes its graphical processors as “amplifying human intelligence”, thanks to their growing role in deep learning research.
Yet just as quickly as semiconductor companies are trying to rebrand themselves as something else, dealmakers are rushing into the sector. Since 2015, the chip sector has seen more than $150bn-worth of mergers and acquisitions — and that was before Broadcom chief Hock Tan made his bold $130bn move on Qualcomm this month.
Broadcom is the product of several of those deals, not least its $37bn buyout by the smaller Avago in 2015, which retained its target’s name.
Its bid for Qualcomm, whose board has rebuffed its $70-a-share offer as too meagre, is likely to prove a high water mark in the current consolidation frenzy. “In a way, all the M&A has happened,” says James Wang of Ark Invest, an investment adviser. “This is the last one you could even imagine.” A deal could create the world’s third-biggest chipmaker behind Samsung and Intel.
Qualcomm and Broadcom sit at the top of a complex but rapidly shrinking food chain. Before Broadcom swept in, Qualcomm itself was trying to complete the chip sector’s biggest deal to date, buying NXP, the Dutch chipmaker that only two years ago paid $12bn for Freescale. Broadcom and Avago are still integrating with each other after a merger in 2015. It is enough to give even a veteran dealmaker like Mr Tan indigestion.
The last few years’ activity reflect a unique set of circumstances in the tech industry. A decade after the introduction of the iPhone, the smartphone market is vast but mature, offering slower, less profitable growth to companies such as Arm Holdings and Qualcomm.
Executives are convinced that artificial intelligence and augmented reality will soon sweep the globe like the personal computer and the smartphone before them. Softbank’s $32bn buyout last year of Arm, the UK-based chip designer, was a gutsy bet on the internet of things, or IoT, by the Japanese conglomerate’s founder, Masayoshi Son.
However, it is not quite clear when or even which of these revolutions will happen. The result is something unusual in the land of disruption: the status quo is holding. Apple, which is sitting on $269bn in cash and whose latest iPhone X is already proving successful, seems unlikely to fall on harder times like Nokia or BlackBerry anytime soon. The awkward alliance between Samsung and Google based on the latter’s Android operating system is holding up, while the South Korean electronics group’s dominance in chip manufacturing continues to grow. The prospect of a new entrant loosening Lenovo, Dell and HP’s collective hold on more than half of the PC market seems remote too.
The global dominance of a few big tech companies may be causing concern in Washington and Brussels, but it presents an opportunity for the likes of Mr Tan. If consolidation in maturing markets has been one force driving the prolific deal making, the other has been a rush by established powers and newcomers alike to position themselves for the next wave of growth.
“When you have a small number of very powerful customers, you have to have a very small number of powerful suppliers,” says Michael Marks, former chief executive of electronics supplier Flextronics and a former Broadcom director. “Thirty years ago you had 200 computer customers, now you have three. Fifteen years ago you had 20 cellphone customers, now you have two. The hardware sector has consolidated massively over the last 30 years . . . I’m surprised it has taken this long for there to be more consolidation in the semiconductor market.”
At a time when the likes of Intel and Qualcomm have been investing in search of that elusive next big thing, from forays into smartwatches and virtual reality to multibillion-dollar bets on autonomous cars, Broadcom has shown a more narrow focus in the less glamorous parts of the supply chain.
Intel is best known for the central processing units that power Windows PCs and Macs, but its other chips can be found in iPhone modems, data centres, drones and self-driving cars. Unlike many of its rivals, it manufactures chips as well as designs them. Its $17bn acquisition of Altera in 2015 added new capabilities.
Founded in 1993, Nvidia’s strength has always been in graphical processing units, which run alongside a CPU. Nvidia is still popular with gamers, including players of Nintendo’s latest Switch console, but the GPU has taken on a new life in cryptocurrency mining, training artificial intelligence systems and autonomous cars.
San Diego-based Qualcomm’s heritage is in wireless technology, particularly in 3G and 4G cellular networking, where it owns foundational intellectual property. It capitalised on the growth of Android smartphones with its Snapdragon “system on a chip”, while also selling modems to Apple since the first iPhone. It is now investing in the hope of dominating the next generation of 5G wireless networks.
Broadcom supplies components to telecoms, enterprise and industrial companies, which are used in TV set top boxes, smartphones, broadband infrastructure and energy systems. Its origins date back to the 1960s at Hewlett Packard and in AT&T’s Bell Labs. A driver of its $130bn bid for Qualcomm was to expand into cellular wireless chips for mobile devices.
“Broadcom has always been an ingredients supplier, that’s very much their heritage,” says Geoff Blaber, an analyst at CCS Insights.
Mr Tan puts it a little differently. “We have bought five companies over the past five years,” he explains. “The way our business model works is, when we buy a business we look at the strong, the core sustainable product lines . . . We very often invest even more than the original owners have invested in those businesses.”
He is dismissive of diversions into “peripheral” or “adjacent” businesses that the management teams of his targets “can’t help dabbling in”, adding: “In those businesses, we tend to go through an asset rationalisation.”
Contrast that approach with Intel, which went far beyond mere dabbling with this year’s $15bn acquisition of Israel’s Mobileye, a maker of sensors and cameras for computer-assisted and autonomous driving. Self-driving cars are moving from science fiction to reality faster than many expected: Alphabet’s Waymo is already putting truly driverless cars into public service, many of them powered by Intel’s processors.
Even after winning its first slot in the iPhone last year, Intel is searching for new sources of growth beyond the central processing unit, the workhorse chip driving most PCs and servers, after largely missing out on the smartphone revolution. Its attempt to reposition itself as a data company also reflects its strength in data centres, where Intel says it is targeting a $50bn market opportunity for silicon ranging from server processors to networking and AI.
Here, the rise of cloud computing has helped drive faster growth than the PCs that make up its “client” business, which still makes up more than half of Intel’s total revenues. Despite the hype, the “internet of things” accounted for only 5 per cent of Intel’s $45.7bn in revenues for the first nine months of this year. Pressure to diversify also motivated Intel’s $400m acquisition of Nervana, whose technology is more suited to AI systems than the CPU.
In autonomous cars specifically and AI generally, Intel is seen as playing catch-up with Nvidia, whose graphics processing units have been eagerly adopted by researchers working on deep learning. Last year, Nvidia was the best performing stock on the S&P 500, rising by 224 per cent. Its shares have doubled in value again this year.
“Nvidia is at the centre of AI, machine learning and deep learning and where the value really resides, which is in neural networks and training,” says Mr Blaber. “That doesn’t mean it’s not going to face competition from Intel, and Google and Facebook are investing in their own silicon [for AI]. But Nvidia is in a remarkably strong position.”
It has achieved this stronghold without buying its way into the market. Like Intel, Nvidia missed mobile, prompting its 2011 acquisition of Icera, a maker of smartphone modems that competes with Qualcomm. But the modem business was closed last year, just as the value of its original focus on graphics chips was becoming clear in AI.
“We enabled the AI big bang to happen because of this new computing architecture that we invented,” Jensen Huang, Nvidia’s chief executive, told the FT in an interview this year. “Just as software ate the world, AI is going to eat software.”
For now, more than half of Nvidia’s sales still come from gamers wanting the latest graphics cards for their PCs. Though less than a quarter of its most recent quarterly revenues come from AI-centric applications in data centres, it is growing fast, doubling year on year.
“When people think about IoT, that’s an AI problem. When people think about self-driving cars, that’s an AI problem,” Mr Huang said. “I believe there’s going to be a trillion, multiple trillions of devices in the world that are going to have inference capability . . . basically, localised AI capability.”
Once seen as a takeover target, Nvidia’s surging share price has given it a market capitalisation of $126bn — seen by most analysts as too large for potential buyers such as Intel.
Until recently, the same was said for Qualcomm, whose valuation was eclipsed by Nvidia this year. The two companies’ share prices give a clear picture of investors’ enthusiasm for their respective stories: while Qualcomm’s stock would have been broadly flat without the recent rally driven by Broadcom’s bid, Nvidia has increased by more than 600 per cent in the same period. “It comes back to the fact that it’s a very opportunistic move on Broadcom’s part,” says Mr Blaber.
Qualcomm’s stock has been hit hard since January by its antitrust battles with Apple over pricing and patents. Apple has stopped paying Qualcomm, which is trying to block iPhone sales from San Francisco to Beijing. Investors are worried because Qualcomm’s licensing business, QTL, is “just such a disproportionate amount of profit that the Apple weight on its earnings quarter to quarter is substantial”, says Mr Blaber. “Plus [there are] long-term fears about all the other regulatory cases and how that might impact the business model.”
To its proponents, Qualcomm’s biggest strength is its unique combination of chipmaking, with hundreds of millions of modems and “Snapdragon” processors sold to smartphone makers round the world every year, and intellectual property licensing, which feeds off an accumulated $47bn investment in research and development since its foundation in 1985. Critics, including Apple and regulators in several continents, argue that the combination is fleecing customers by allowing Qualcomm to collect more royalties than it deserves for its patents for cellular wireless networking.
If regulators do not intervene first, Broadcom’s proposed takeover of Qualcomm would create even greater supplier concentration for smartphone makers. “With Qualcomm, Broadcom would be able to offer basically every chip in the iPhone besides [Apple’s] A-Series processor,” says Mr Wang.
Yet the presence of the iPhone’s A-Series processor points to what Mr Wang sees as the semiconductor industry’s most successful deal: Apple’s 2008 acquisition of PA Semi, for what now seems like a modest price, $278m. The deal has allowed Apple to take control of the chips at the heart of its devices and with it much of the iPhone’s competitive differentiation.
As the UK-based graphics chipmaker Imagination Technologies found out this year, Apple can quickly go from your largest customer to a brutal competitor. After the iPhone maker said it would bring design of graphics chips in-house, Imagination’s shares fell and it was forced to put itself up for sale.
Hardly anyone wants to be a chip company these days, it seems, except for the world’s most valuable company.
Additional reporting by Richard Waters
Overlaps raise doubts for further consolidation
Regulators have taken a close interest in consolidation among chipmakers, holding up Qualcomm’s $47bn NXP deal and Broadcom’s $6bn bid for Brocade in recent months. Given its size and complexity, analysts expect Broadcom’s $130bn attempt to seize Qualcomm will also face a lengthy review. Qualcomm itself, which is resisting the bid, is using that threat as part of its defence.
Yet chopping up Qualcomm or selling off business units would fit the dealmaking modus operandi of Hock Tan, Broadcom’s consolidator in chief. One idea already pushed by activist investors in Qualcomm is to split its highly profitable intellectual property licensing business from its chipmaking unit. Under pressure from activist investor Jana Partners two years ago, Qualcomm’s board launched a review of the idea but chief executive Steve Mollenkopf ultimately decided there were too many “strategic benefits and synergies” from the combination.
Nonetheless, that unique combination lies at the heart of many of Qualcomm’s current legal and regulatory troubles.
Concerns about these and the NXP deal preoccupied analysts on Qualcomm’s most recent quarterly earnings calls. Instead of discussing its moves into healthcare, robotics or smart cities, investors are worried that its licensing issues with Apple threaten contagion to other customers. Qualcomm has said it is already in dispute with one large Asian smartphone maker, which many analysts believe to be Huawei, although the situation has not yet deteriorated into litigation.
Even before launching his bid for Qualcomm, Mr Tan had been cosying up to the US government. At a press conference alongside President Donald Trump, he promised to relocate Broadcom’s headquarters from Singapore to the US.
Given Broadcom’s considerable overlap with Qualcomm, he may need more than a presidential photo-op to accomplish tech’s biggest ever deal.
Get alerts on Semiconductors when a new story is published