Nvidia’s pending $40 billion acquisition of Arm is expected to have a big impact on the chip world, but it will take years before the effects of this deal are fully understood.
More such deals are expected over the next couple of years due to several factors — there is a fresh supply of startups with innovative technology, interest rates are low, and market caps and stock prices of buyers are high, which makes borrowing money or stock deals much easier. In addition, there are a variety of new markets just beginning to heat up, such as 5G, edge computing, AI/ML, and continued development in automotive. Most of these are being held up in the short term by the impact of the coronavirus pandemic, but that won’t last forever. Acquisitions are the fastest way to achieve critical mass and fill out a company’s offerings, and they are a quick way to grow the talent pool inside a company.
The Nvidia-Arm deal fits the bill on all counts. It fuses together the No. 1 GPU vendor with the No. 1 processor core vendor. It extends Nvidia’s reach from the data center, where its GPUs are used for algorithm training, all the way to the edge, where Arm is strongest. And it helps both companies position themselves in the ill-defined but emerging world of edge computing, where highly specialized devices and servers are used to pre-process or fully process the explosive growth in data. In addition, Arm has one of the broadest ecosystems of any company in the semiconductor industry, and strong working relationships in nearly every region and market segment.
“There’s a lot we can do as a combined company,” said Simon Segars, CEO of Arm. “The added investment firepower will allow us to push head-on in the expansion in the data center — and to push AI, which has grown up in the data center, out everywhere through the network and into the end points the edge. Nvidia has a large IP portfolio that they use to build chips and systems and boxes. We license IP to the world semiconductor industry, and we’ve built an ecosystem around that. So we are going to have more IP to license to people.”
Time will tell
Still, acquisitions are best viewed in the rearview mirror, and in the context of other deals. Case in point: Texas Instruments’ $7.6 billion acquisition of Burr-Brown seemed like a blockbuster deal in 2000, but the ramifications of that acquisition proved to be more limited than the initial hype would suggest, according to several industry insiders familiar with that deal. The combination bolstered TI’s analog capabilities and sharpened the company’s business focus, but the impact on the rest of the industry has been less obvious. The same was true of TI’s $6.5 billion acquisition of National Semiconductor in 2011.
In contrast, Analog Devices’ acquisitions of Linear Technology in 2017 for $14.8 billion, and its pending acquisition of Maxim Integrated for $21 billion, would create a powerhouse competitor to TI, which has dominated the analog space for decades. That opens the door to much more price and performance competition, and for more investment in startups in this lucrative market space.
“Intel’s acquisition of Mobileye is potentially significant,” said Segars. “Broadcom and Avago was a significant deal, too. People have talked about semiconductor consolidation for a long time, but it’s really happening now. If you draw a graph of the market caps for these companies, you start to see a long tail, which is really interesting. ADI buying Maxim is another one. This phase of consolidation is a bit slower because of COVID-19. It’s hard for companies to get together that have never met before. With Arm and Nvidia and Softbank, everybody knew each other. And with Maxim and ADI, they had a long-term relationship.”
EDA and IP
EDA has racked up more acquisitions than any other segment of the chip market. Most of them have been relatively small. Nevertheless, the Big Three EDA companies never would have become the Big Three without acquisitions to fuel their growth, and it’s unlikely the chip industry would have developed the way it did. Synopsys alone has done more than 100 acquisitions. Mentor has done more than 70, and Cadence has done more than 50. (The numbers are imprecise because there also are spinouts and acquisitions of assets rather than entire companies.)
“Many fall in the category of incremental, increasing the critical mass technically and also economically,” said Aart de Geus, chairman and co-CEO of Synopsys. “One of the larger ones we did was the acquisition of Avant! Synopsys had built what at that point was called the front end of design — synthesis, simulation, timing, power and so on. The back end at that time was the physical design, which was place-and-route and some of the verification, which was done by different companies. I had become increasingly concerned in the late ’90s that, from a technology point of view, the interdependence between design and back-end design was going to become more important. I saw this coming around 2000, and there was a fear that with a downturn — and it turned out that in 2001 there was a massive downturn — a number of consumers were going to reduce the number of suppliers because they wanted to save. There was only one company, which was Cadence, that had both some front-end and some back-end to tie it together. Mentor was somewhere between all of that, and it was clear we needed to increase our ability to have a complete flow.”
It was Cadence that started the EDA industry on the acquisition track. “The two acquisitions that stand out in EDA are Cadence-Gateway and Cadence-Tangent,” said Wally Rhines, CEO Emeritus at Mentor, a Siemens Business. “Gateway developed Verilog. Gateway got Cadence in the Verilog business just as the world was moving from schematic capture to RTL, which drove Verilog as a standard. The industry consortium was VHDL, but Verilog got the lead. Equally significant was the Cadence acquisition of Tangent. Tangent was a gate-array router company — the best gate-array router company — but they had developed the capability to move to standard cells, as well. Before that, the Mentor IC Station and Cadence’s precursor to Virtuoso, which came out of Solomon Design Automation, were both detail routers. IC Station probably would have been the Virtuoso today, except that internal feuding between the SCS (Silicon Compiler Systems) people and the IC Station people at Mentor put them in a total stall, opening the window for Cadence.”
Arguably the most significant acquisition in EDA, though, involves the 2016 acquisition of Mentor by Siemens AG, a giant conglomerate and the largest industrial manufacturing company in Europe. The $4.5 billion deal provided Siemens with everything from wiring harness design software to a complete semiconductor flow, and it provided Mentor with the deep pockets of a company much larger than all the EDA companies combined. To put this in perspective, in 2019, Siemens’ revenue topped $100 billion. The total value of all EDA companies combined is a fraction of that amount.
IP companies have been active in buying other IP companies, as well. Arm’s growth is at least partly due to acquisitions over the years. And some EDA companies — particularly Cadence and Synopsys — have been buying up small IP companies, as well.
“So far in the IP space, the most significant acquisitions have been Synopsys assembling an almost $1 billion operation that basically focuses on peripheral I/O IP and PHYs,” said K. Charles Janac, chairman and CEO of Arteris IP. “Nvidia’s acquisition of Arm is extremely significant both for the IP industry and the semiconductor industry. Nvidia is trying to become the next generation of computing platforms, which is targeted directly at Intel and AMD. So with Nvidia behind it, maybe the Arm architecture becomes the next brain of the SoC.”
The key is synergy. “I used to work for Joe Costello when he was the CEO of Cadence, and he had a couple of interesting concepts,” said Janac. “The deal had to be a good financial deal. But there also had to be some synergies involved. If you put 2 things together, they should not add up to 2 or even 2.5. They should add up to 3 or 4. If you can acquire products from a small company that is doing well enough in the marketplace and run them through a very large, very capable sales force, then you are going to make money as the acquirer. And that’s exactly what happened. The companies also need to culturally compatible and excited to work together. And the other thing that Costello said was that the best deals are the ones where each side is just a little bit unhappy. If somebody is really ecstatic, then it probably wasn’t a good deal.”
Processors and memory
On the processor side, the Nvidia-Arm acquisition is a big deal in more ways than one. Arm’s ecosystem is so broad that Arm processor cores dominate everything from the application processors in smart phones to a wide swath of portable devices. But it’s certainly not the only important acquisition.
AMD’s $5.4 billion purchase of GPU maker ATI in 2006 helped to establish AMD as a serious rival to Intel in the data center. “AMD bought ATI’s graphics business, which gave them the ability to go head-on with the CPU business, building on Intel-compatible CPUs,” Rhines said. “ATI was a really unique company. If you look at history, a new graphics company came along about ever 10 years as the graphics standards changed. ATI was the only one that lasted through multiple generations and survived.”
In the memory space, a series of acquisitions led to the resurgence of Micron. “Hitachi and NEC combined to become Elpida, which was acquired by Micron,” he said. “That really kept Micron in the DRAM business, and today we have three viable DRAM suppliers. Micron is the smallest, but that deal gave them enough mass and momentum, as well as a Japanese presence, to be significant.”
Shane Rau, research vice president at IDC, agreed. “Micron buying Elpida put them over the threshold,” he said. “Now there are the Big Three memory companies. That threshold was necessary for them to have sufficient supply and demand in DRAM. We expect there will be consolidation in the NAND market, too.”
Government intervention and fallout
Still, not all announced acquisitions come to fruition. Review processes can yield surprising results, and governments have a long history of stepping in wherever they perceive a national threat. In the past, it was largely the U.S. government that was responsible for killing or delaying major deals, but others such as China are beginning to flex their market muscle, tool. China was responsible for derailing the $44 billion acquisition of NXP by Qualcomm.
There is a long history of these kinds of actions. One of the first really significant deals in the chip industry involved the attempted acquisition of Fairchild Semiconductor by Fujitsu in 1987. The deal was killed after the U.S. Dept. of Defense, the Department of Commerce and the CIA jointly asked then-President Ronald Reagan to intervene. The deal put the chip industry, specifically, on the government’s radar for the first time, and led to the creation of another industry powerhouse.
“National Semiconductor acquired Fairchild as a result of the fact that the Fujitsu deal broke down,” said Rhines. “That deal was significant because National had minimized its R&D investment for decades, while Fairchild, under Don Brooks (then president and CEO) had developed wonderful new technology. It gave National new life because a company that had minimized R&D and maximized operational excellence acquired a company that had maximized R&D investments.”
The U.S. government played a significant role in two other deals that indirectly shaped the chip industry, as well. The first involved the 1956 consent decree signed by IBM, which at the time was meant to rein in IBM’s monopolistic pricing for its “market basket” approach of bundling services, software and mainframe computers. IBM was the only company to offer all three. And because its platform was the de facto standard at the time — all software had to be compatible with IBM’s equipment — IBM used its market clout to underprice competitors in whichever of those three segments it was being challenged.
By the time the PC era began in the early 1980s, IBM was still making most of its revenue from mainframes and minicomputers, and the view of its top executives was that the PC was a toy. So rather than throttle the market again and face more enforcement actions from the government, IBM inked deals with Intel and Microsoft rather than trying to own all of the technology.
The U.S. government also broke up the Bell System in 1982, resulting in yet another consent decree. AT&T relinquished control of the Bell Operating Companies, which in turn were split into regional operating companies. Bell Labs, which did a large amount of work in conjunction with the federal government, at the time had one of the premier R&D operations in the world, rivaled only by IBM. Bell Labs created the first transistor in 1947, the Unix operating system that became the basis of Linux, as well as the first optical router.
Bell Labs was sold to Alcatel in 2006 as part of Lucent, spinoff that was part of the breakup, and that was again in 2016 to Nokia. GlobalFoundries, meanwhile, acquired IBM’s microelectronics business in 2015. Both of those acquisitions ended the two largest semiconductor research operations. While IBM still does chip research for AI systems, the heyday of corporate and corporate/government-sponsored research into semiconductors, communication and computing is over in the United States.
Governments have blocked other deals in the chip industry, as well. The Committee on Foreign Investment in the United States (CIFIUS) blocked the $23 billion acquisition of Micron by Tsinghua Unigroup in 2018. The U.S. Dept. of Justice also blocked a $9.3 billion merger between Applied Materials and TEL in 2015.
M&A by the numbers
The Nvidia-Arm deal remains the largest in terms of size for the semiconductor industry, although not by much. Avago bought Broadcom for $37 billion in cash and stock. What’s worth noting here, though, is there appears to be a ceiling for what’s considered acceptable.
“A couple years ago, we determined that semiconductor M&A agreements (excluding those deals for system-level and software businesses) had apparently hit a limit of about $40 billion,” said Rob Lineback, senior market research analyst at IC Insights. “Qualcomm’s failed $44 billion acquisition of NXP was dropped in July 2018 because China kept delaying its approval of the transaction in the midst of a trade war with the U.S. Broadcom’s $121 billion hostile takeover bid of Qualcomm (later reduced to $117 billion) was blocked in March 2018 by President Trump because of concerns about the potential loss of the country’s leadership in cellular technology.”
Despite the availability of capital and the willingness of companies to engage, numbers matter. “Essentially, it appears about $40 billion has become a limit in the size of do-able semiconductor acquisitions because of the high dollar value of mega-deals, rising protectionism among more countries, and growing trade frictions,” Lineback said. “The geopolitical environment and trade wars may continue to limit size of semiconductor M&A. However, Nvidia’s $40 billion agreement bumps up against this hypothetical ceiling. In addition to impacting many areas and major players in the IC business, the Nvidia-ARM deal seems to be a test of the practical chip M&A limits in today’s geopolitical climate.”
Numbers cut a different way in China, as well. Non-indigenous companies need to set up subsidiaries in which Chinese partners own 51% or more of the company. “The deal where Arm spun off 51% of Arm China is significant,” said IDC’s Rau. “That was one of several deals that brought IP to China. MIPS is open to China, and so is RISC-V.”
How new acquisitions will shape the chip industry remains to be seen, but the slowdown in Moore’s Law and the drive toward more heterogeneous design, as well as the need for processing and intelligence everywhere, are changing the dynamics of the chip industry.
“We seem to be entering a new phase in the semiconductor industry,” said Roddy Urquhart, senior marketing director at Codasip. “In the 1970s and 1980s there was often vertical integration with larger semiconductor companies developing their own processor cores, EDA tools and even sometimes processing equipment. By the 1990s, this was breaking down with companies such as Texas Instruments and Siemens Semiconductor (spun off in 1998), abandoning in-house EDA tools in favor of commercial ones. In the same timeframe, IP companies including Arm, MIPS, ARC and Tensilica emerged as providing an alternative to in-house cores. By 2000, most companies around the world were relying on the Big Three EDA vendors and Arm for most of their design tool and IP requirements. In the same timeframe we saw the emergence of pure-play foundries and the growth of fabless semiconductor companies. And with a stable commercial environment, worldwide IC design grew in many geographies, especially in India and China.”
The recent geopolitical tensions have eliminated that sense of stability. “With China unable to access U.S. technologies such as Android, certain EDA tools, and silicon technologies, they will be driven to establish a homegrown ecosystem. Arm had been perceived as a player that was quite independent of their licensees, but this will not apply under Nvidia ownership. In the processor area, RISC-V is the obvious alternative for companies to look at. This is also an opportunity for EDA companies based in neutral locations, such as Germany or Canada.”
The bottom line: More acquisitions are coming as the market continues to restructure around different regions, new technologies, and the need to process much more data in more places. But how quickly and where isn’t clear yet.
—Susan Rambo contributed to this report.
Nvidia To Buy Arm For $40B
Analysis: What’s driving the deal, and what to watch for.
The Next Wave Of Consolidation
Economic conditions are all there. So what’s the holdup?
New Architectures, Much Faster Chips
Massive innovation to drive orders of magnitude improvements in performance.