Why Nvidia is spending $40 billion on Arm

It’s official: Nvidia will buy Arm from SoftBank for $40 billion, the companies announced on Sunday. 

The transaction will be the largest semiconductor deal of all time by dollar value if it is completed. But the dollar value may actually understate the importance of the deal.

Arm is at the heart of nearly every smartphone sold today, from Apple iPhones to Samsung Galaxy devices running Google‘s Android software. In addition to smartphone processors, Arm technology is used in lots of smaller chips integrated into all kinds of computer systems and devices. Arm said that more than 180 billion chips with its processor cores and other components have been shipped around the world. 

“They’re in everything, and they power computers of all different types of shapes from the fastest supercomputer in the world, to smart watches and smart thermostats,” Nvidia CEO Jensen Huang told CNBC on Monday. 

But for Nvidia, Arm represents a chance to boost its fast-growing business of selling chipsets and software to data centers, the massive networks of computers that power large companies and cloud computing networks.

In particular, investors believe that Nvidia could closely integrate Arm’s CPU technology with Nvidia’s speciality in graphics processors (GPU) to build a lead on the rest of the semiconductor industry. 

Mark Lipacis, analyst at Jefferies, wrote on Monday that the deal could let Nvidia bolster its data center business by creating an “ecosystem” of server chips and software ranging from Arm-based CPUs to its GPUs. Nvidia could more closely integrate Arm processors into its products, increasing performance without using additional power.

It could also entrench Nvidia and make it harder for competitors to steal business. 

“With Arm, we think Nvidia is poised to also offer general purpose, heterogeneous data center scale solutions and ultimately capture up to 80% of the value of the serial processing part of the data center ecosystem,” Lipacis wrote. “We also think that Arm gives it a larger moat in the data center.”

Nvidia’s shares, already up more than 110% this year, rose more than 5% on the Monday after the deal was announced, suggesting investors have a lot to like from the deal.

Both Arm and Nvidia are silicon-focused companies, but the two focus on different areas of the semiconductor industry. Huang says Arm and Nvidia are “complementary.” 

Arm, based in Cambridge, England, sells the blueprint that other companies need to make low-power chips to power devices like smartphones, tablets, virtual reality headsets, and smart speakers. It’s called an instruction set. Some customers, like Apple, license Arm’s architecture intellectual property and design their own CPU cores. Others, like Qualcomm or Samsung, license entire processor designs from Arm itself. 

Arm makes money by selling intellectual property licenses, software, and tools to the chip designers that want to use its technology, and takes a royalty on every chip that uses its technology. The company generated revenue of more than $1.7 billion in 2019.

Arm has had a long history. It was founded in 1990 as part of a joint venture that included Apple, which wanted to use low-power Arm technology in a portable computer called the Newton.

But Arm ascended to the forefront of the industry over 15 years later with the help of the smartphone boom. Smartphones needed chips that used significantly less power than the x86-based laptop and desktop chips made by Intel — they wouldn’t have as much computing power as a laptop, but at least their battery would last for more than a few hours. The first iPhone had an underpowered Samsung processor with an ARM processing core.

Since then, chips with Arm-based cores have become the standard for both Apple and Android phones, and have become significantly more powerful, even challenging power-hungry desktop or laptop chips in performance for some tasks. Qualcomm’s chips, used in Samsung, Google, and Huawei phones, use Arm cores. 

Arm is also making inroads into other types of computers beyond smartphones. Apple is planning to shift its Mac laptops and desktops to in-house chips using Arm technology. Microsoft has spent a lot of time and money adapting Windows to run on Arm chips, too. And Arm chips are increasingly being used in data centers, where their low power draw can be an advantage.

When Softbank bought Arm in 2016 it also emphasized potential in the “internet of things,” or the trend that an increasing number of things, such as machinery or cars, could be equipped with internet connectivity and a cheap microprocessor — running on Arm technology, of course.

Nvidia, led by CEO and co-founder Jensen Huang, was originally best known for designing chips that would enable central processors to offload 3D graphics work to increase power and efficiency. Its customers were gamers, companies building gaming systems, and people who needed power to build 3D graphics. 

Games are still a big part of Nvidia’s business. But in the past decade, Nvidia has started to expand into a number of different categories.

In 2008, Nvidia started building its own “system-on-a-chip” (SoC) hardware under the Tegra brand name, combining its graphics processors along with Arm processor cores in one product. A Tegra chip is at the heart of the Nintendo Switch console, for example, although it the product line has not gained as much traction among smartphone makers as Qualcomm’s Snapdragon. Nvidia’s self-driving car platform includes hardware chips with Arm cores.  It now builds chips and software for self-driving cars and robots.

But the biggest tailwind at Nvidia’s back in recent years was the rise of artificial intelligence. Researchers found that graphics processors — Nvidia’s speciality — were well adapted to run power-hungry artificial intelligence algorithms more efficiently than older server chips or simple CPUs.

As Nvidia’s ambitions have swelled, so have the company’s fortunes. At the start of trading on Monday it was valued over $322 billion, a startling rise from five years ago when it was worth about $12.3 billion, according to FactSet data. In its fiscal 2020, which ended on January 26, Nvidia reported $10.92 billion in revenue, down 7% from $11.72 billion a year earlier. 

Some analysts see the deal as primarily a boost to Nvidia’s data center business, which is currently the company’s fastest growing unit. In the quarter ending in June, it reported $1.75 billion in data center revenue, or about 45% of the company’s sales, surpassing its gaming unit for the first time. Data center sales were up 167% on a year-over-year basis. 

Here’s how Nvidia CEO Huang described the data center vision on Monday: “If you think about all the chips that have been shipped, 22 billion chips that were shipped last year, we know very well going forward that AI will be infused into them.” 

Nvidia sees an opportunity to sell its AI chips and graphics chips to Arm’s customers, too. “If we could leverage their business model, if we could stand on the network they’ve created and take the technology we’ve invented and put it into that channel and make it available to their thousands of partners, the economics for us would be incredible,” Huang said. 

But many of those Arm customers compete with Nvidia, including Samsung, Qualcomm, Broadcom and Texas Instruments.

These companies may balk at being sold additional technology from a competitor. Nvidia says that it will continue to operate Arm as an independent subsidiary with a “customer neutral” and “open-licensing” model — basically, if companies can pay for it, they can obtain a license, even if they’re direct competitors with Nvidia.

Huang said on Monday that Arm CEO Simon Segars had already spoken to many of its customers to answer their questions about the setup. 

“They want to know they can continue to rely on Arm. The fact that we’re announcing the deal today suggests that the conversations went fine. Although nobody likes change, if they realize we’re committed to the business model, we’re committed to openness and fairness, and more than that, we want to keep everything the same plus we want to add our IP to it, so they can have more to purchase and more value we can deliver to them,” Huang said.