feature article
Subscribe Now

ARM’s Sleight of Hand

Nvidia Acquisition Raises as Many Questions as it Answers

When you want to unload some rusty old lawn furniture, you have a lot of choices. You can list them on eBay or Craigslist. You can put an ad in your local newspaper or Nextdoor.com, or post a few pictures on Instagram, Facebook, and/or Twitter to see if you get any nibbles. Or just leave it out by the curb and wait for some bozo to come along and take it off your hands. 

But, if you’re selling an original Rembrandt oil painting, your choices are somewhat limited. 

That was the problem facing SoftBank, the owner of ARM. SoftBank wanted to consciously uncouple itself from its once-beloved acquisition, but the astronomical asking price wasn’t attracting too many takers. There are only so many places you can unload an 11-figure white elephant. 

And the buyer, as we know, is Nvidia. The Santa Clara–based chipmaker has decided it really, really, really wants to be in the CPU-licensing business. The company emptied its coffers, traded billions of dollars’ worth of its own stock, printed some more stock, and generally leveraged every resource it had in order to close one of the biggest tech deals in recorded history. 

If $40 billion sounds like a lot, it is. You could buy your own private island, a pair of Gulfstream G550s to get yourself there, furnish it with 365 Rolls-Royces (one for each day of the year, natch), and decorate it with the world’s most expensive oil paintings – and it would all still be just a rounding error. Yes, $40 billion is a lot of money. 

Point being, Nvidia was one of the few companies on the planet with the fiscal throw weight to play in this arena. Maybe not the only potential suitor, but one of the few. And possibly the only one that could pass geopolitical muster. But more on that later. 

The deal will make Nvidia the sole owner of ARM and make SoftBank the largest shareholder in Nvidia. ARM itself will remain domiciled in Cambridge, UK, even though its new owner is based on California. No big deal; ARM’s previous owner SoftBank was based in Japan but that didn’t seem to affect ARM’s business much. More on that later, too. 

SoftBank acquired ARM in 2016 for $32 billion and is now selling it four years later for $40 billion. That’s a pretty good return on investment (ROI), even though SoftBank didn’t buy ARM with the intention of flipping it for a quick profit. At the time, SoftBank’s CEO Masayoshi Son said he was investing in ARM for the long term because it was foundational technology that would play a key role in the rise of the Internet of Things (IoT). ARM was to be the centerpiece of a whole IoT strategy within SoftBank, including other companies and other technologies.  

That was a good, solid strategy. And it even worked for a while. Trouble was, SoftBank also made some bad investments along with its good ones, and the Japanese firm found itself embarrassingly short of cash. Bad investments like WeWork, Uber, and T-Mobile forced Son-san to sell the family silver, so to speak. Even though ARM was doing well and was part of a larger strategy, it had to go. Now, where’s the category on eBay for secondhand IP companies? 

Although SoftBank’s 25% ROI seems like a great deal, it wasn’t, really. The overall stock market grew more vigorously than that over those four years, and tech stocks typically doubled. Back in 2016, Nvidia and ARM had about the same stock market valuation, but since that time Nvidia’s value has skyrocketed (the company is currently worth 50% more than Intel), while ARM’s growth has been more modest. In hindsight, SoftBank would’ve been much better off just buying Nvidia shares back in 2016 instead of acquiring ARM and then selling it to Nvidia. Then again, retrospective armchair investments are always more lucrative than the real thing. Let’s just say that SoftBank didn’t lose a ton of money on the deal, but it didn’t it make much, either.  

Acquiring ARM should be “fiscally accretive,” meaning Nvidia will simply add ARM’s revenue to its own balance sheet. There’s no overlap in the two companies’ operations, so no layoffs or redundancies are planned, and there’s no cannibalism when one product line overlaps another.

Although this may seem like a done deal, it’s not. This is just the beginning. All three companies – ARM, SoftBank, and Nvidia – are big enough that this will need governmental and regulatory approval from just about every major country involved. That includes the United States, United Kingdom, Japan, and China. Why China? Because ARM’s joint venture for handling Chinese licensing is based there, which means the Chinese government gets a say in whether this deal goes down or not. More on that later. 

Regulatory approval is not a slam dunk, and the ARM sale could realistically fall apart before it ever comes together. Smaller mergers and acquisitions than this one have been scuttled because of antitrust, monopolistic, or political reasons. When Qualcomm wanted to acquire NXP, the deal never got regulatory approval, and it was abandoned. 

There are three ways this could play out: The deal gets scuttled for legal or political reasons. Or the deal goes through but it’s a disaster for ARM users and customers. Or the deal goes through and Nvidia is a fantastic new owner for ARM. Let’s look at the last one first. 

Everything’s Rosy

If all goes well, Nvidia takes over control of ARM sometime in early 2022, after the deal has passed all the necessary regulatory hurdles. It continues to license its world-famous CPU cores, along with new IP from Nvidia. ARM continues to develop new Cortex-A, -R, and -M cores, it’s business as usual, and the world doesn’t fall off its axis. That’s the happy ending that the board members of ARM, SoftBank, and Nvidia are all hoping for. 

It’s certainly the picture they painted in the analyst call. The heads of ARM and Nvidia said that ARM would continue to be based in Cambridge, headcount will increase, R&D will expand, and a new supercomputer based on ARM and Nvidia technology would be built in the UK. So, no reason to panic. 

Everything’s Not Rosy

The second alternative is that the deal goes through but it’s a disaster – or at least, a midsized train wreck – for ARM’s licensees. Will ARM really remain independent? For almost all of its history, ARM has been a standalone company, owned by and reporting to nobody. That was part of its charm. It was as neutral as a Swiss palace guard. You could license processors from ARM knowing that the company was fair and even-handed about offering the same technology (if not always the same prices) to everyone. Sure, ARM would turn around and sell the same processors to your competitors in a heartbeat, but that’s just business. Mercenaries gonna mercenary. At least ARM itself wasn’t your enemy. 

When SoftBank bought the company in 2016, everyone worried… for about five minutes. SoftBank also wasn’t anybody’s enemy because nobody could figure out what the company actually did. It definitely didn’t make chips or computers or smartphones, so it must be okay. SoftBank was a benign and benevolent caretaker for ARM, and things continued to run smoothly. 

Not so with Nvidia. The Santa Clara company most definitely competes with many current ARM licensees. Take AMD, for example. It’s Nvidia’s main rival in graphics chips, yet it’s also an ARM licensee. AMD won’t like funneling royalties to its GPU competitor, and it really won’t like negotiating across the table with Nvidia for the next generation of ARM cores. Revealing a new upcoming processor requires an NDA; would you divulge secret technical details to your archrival, even if they pinky swear not to tell their coworkers?

Qualcomm is in the same boat. The company makes chips for smartphones, where Nvidia tried, and failed, to compete. But Qualcomm and Nvidia both make chips for self-driving cars, as do other ARM licensees. That’s going to make negotiations awkward when it comes time to license the next Cortex-A99 or whatever. Will Nvidia deal fairly with Qualcomm, Samsung, and its other competitors in that space? Even if it does, will they trust Nvidia?  How about Ampere and its ARM-based server chips? The list goes on.

Given that nearly every chipmaker in the world uses ARM processors, who isn’t a potential competitor with Nvidia? Whether you make graphics hardware, vehicle systems, AI accelerators, game boxes, or TV appliances, you’re both competing with and beholden to Nvidia and its ARM monopoly. That’s a recipe for friction. 

Intel, AMD, Qualcomm, Samsung, and Apple are just a few of the big ARM licensees that could be forced to become Nvidia licensees. Some of them do not get along with Nvidia. At all. Apple support for Nvidia graphics hardware is noticeably (and deliberately) lacking. Same goes for Linux support on Nvidia silicon. Linus Torvalds himself called Nvidia “the single worst company we’ve ever dealt with,” punctuating his statement with an emphatic gesture

Nvidia might also use its ARM licensing network to strongarm its GeForce GPU technology into places where it’s not welcome. ARM already has Mali, its own in-house GPU technology. Nvidia has hinted that it would like to license its GPU technology alongside ARM’s microprocessor IP, which raises questions about Mali’s future. Nvidia could make licensing GeForce very attractive, or it could make licensing Cortex processors very unattractive unless you also adopt GeForce as your GPU technology. Companies have been known to use their dominance in one product category to force the adoption of another. Just ask any PC motherboard maker. 

One thing stuck out to me during the official analyst call announcing the deal. ARM’s CEO Simon Segars said, “We will maintain our neutral business model and will keep a level of independence.” Sounds about right. But that phrase, “a level of independence” has me worried. It’s vague and ambiguous. What level of independence? A high level, a low level, a legally defensible level? He may have meant nothing by it, but the comment made my Spidey-sense tingle. Segars is a native English speaker, and he’s well versed in the analytical hair-splitting that goes on during investor conference calls, so his ambiguity doesn’t feel accidental. How independent will ARM really be under its new owners? 

Everything Falls Apart

The third option is that the whole deal falls apart and all this speculation is moot. Nvidia’s promise to keep ARM based in the UK, and to build a new supercomputer there, is likely a sop to the UK government, which has come under fire lately for allowing UK technology firms to move overseas. That’s a nice promise, but there are no teeth in it. Nvidia could close or move offices any time it likes, though to be fair, there’s little reason for the company to do so. The Cambridge area (and the UK as a whole) has a terrific talent pool, so my hunch is that everything stays put. 

The bigger issue is China. The US government doesn’t have much trouble dealing with British technology firms, nor with Japanese firms. Shifting ARM’s ownership from Britain to Japan to America isn’t a big deal – if you’re American. It’s much less palatable if you’re Chinese. The current regime in Washington has been noisily antagonistic toward China, to the point of banning products from Huawei and forcing the partial sale of TikTok. The antagonism is mutual, so Chinese approval of the deal is not assured. 

Nvidia may be forced to restructure ARM’s Chinese joint venture or risk losing Chinese customers who don’t want to (or aren’t allowed to) deal with an American-owned business, or even abandon the China market completely, which could be disastrous. China could make the approval process as painful as it wants to. 

It also doesn’t help that Nvidia’s CEO, Jen-Hsun Huang, was born in Taiwan, a country that China does not officially recognize. Any mention of the island nation, to say nothing of making deals with a successful Taiwanese-American businessman, does not go over well in Beijing.  

ARM’s Length

In the best-case scenario, the deal goes through and… nothing happens. We all continue to license, use, and program ARM processors just like before. Apart from some occasional headlines in the financial news, you’d never know anything had happened. That’s swell. 

More likely, the deal churns through the relevant government bureaus for a year or two, then gets approved sometime in 2022. In the meantime, ARM users start to drop off, one by one. Whether the reality supports their fears or not, the perception of conflict will drive some customers away from ARM/Nvidia and toward… what? 

RISC-V is the obvious alternative on almost everyone’s lips. It’s roughly as capable as a low-end or midrange ARM processor. It’s got decent software support and an ecosystem that’s growing nicely. And it’s free. More importantly, it’s not owned by anyone so it can’t be sold by/to anyone. It certainly feels like a good second choice. 

RISC-V International’s CEO, Calista Redmond, wasted no time in fanning the flames, obliquely referring to ARM as a “legacy architecture for certain product lines” while positioning RISC-V for “next generation applications” in the “rapidly evolving frontier of computing.” In short, if you’re disenchanted with ARM’s embrace, we’re here for you.

Had this all happened a few years ago, MIPS might still have been a viable alternative, too. But MIPS has slipped into oblivion. Or China. Either way, its days in the limelight are behind it, just another near miss for the once-great CPU architecture.

Changing processors is hard. It’s almost like a religious conversion – or maybe a weekly support group that meets in the church basement. Nobody changes processor families unless they have to, and it’s not clear that any perceived problems brewing inside Nvidia will be enough to force that change. Still, it’s probably a good time to be a RISC-V vendor. Or an Nvidia shareholder.


Leave a Reply

featured blogs
Dec 5, 2023
Introduction PCIe (Peripheral Component Interconnect Express) is a high-speed serial interconnect that is widely used in consumer and server applications. Over generations, PCIe has undergone diversified changes, spread across transaction, data link and physical layers. The l...
Nov 27, 2023
See how we're harnessing generative AI throughout our suite of EDA tools with Synopsys.AI Copilot, the world's first GenAI capability for chip design.The post Meet Synopsys.ai Copilot, Industry's First GenAI Capability for Chip Design appeared first on Chip Design....
Nov 6, 2023
Suffice it to say that everyone and everything in these images was shot in-camera underwater, and that the results truly are haunting....

featured video

Dramatically Improve PPA and Productivity with Generative AI

Sponsored by Cadence Design Systems

Discover how you can quickly optimize flows for many blocks concurrently and use that knowledge for your next design. The Cadence Cerebrus Intelligent Chip Explorer is a revolutionary, AI-driven, automated approach to chip design flow optimization. Block engineers specify the design goals, and generative AI features within Cadence Cerebrus Explorer will intelligently optimize the design to meet the power, performance, and area (PPA) goals in a completely automated way.

Click here for more information

featured paper

Power and Performance Analysis of FIR Filters and FFTs on Intel Agilex® 7 FPGAs

Sponsored by Intel

Learn about the Future of Intel Programmable Solutions Group at intel.com/leap. The power and performance efficiency of digital signal processing (DSP) workloads play a significant role in the evolution of modern-day technology. Compare benchmarks of finite impulse response (FIR) filters and fast Fourier transform (FFT) designs on Intel Agilex® 7 FPGAs to publicly available results from AMD’s Versal* FPGAs and artificial intelligence engines.

Read more

featured chalk talk

Automotive/Industrial PSoC™ High Voltage (HV) Overview
Sponsored by Mouser Electronics and Infineon
In this episode of Chalk Talk, Amelia Dalton and Marcelo Williams Silva from Infineon explore the multitude of benefits of Infineon’s PSoC 4 microcontroller family. They examine how the high precision analog blocks, high voltage subsystem, and integrated communication interfaces of these solutions can make a big difference when it comes to the footprint size, bill of materials and functional safety of your next automotive design.
Sep 12, 2023