Nvidia seeking a way to complete tricky Arm acquisition

GPU maker's move to buy what amounts to the wellspring of technology on which everyone in the mobile market depends is putting a major scare into its main rivals.
Written by Roger Kay, Contributor

When Nvidia offered $40 billion to pick up British silicon developer Arm a year ago this month, most industry observers knew that the road ahead for the American graphics chip company wouldn't be easy. If you look at a pyramid of cascading intellectual property in the mobile processor industry, Arm sits at the tippy top. Arm provides and updates instruction sets, documentation, rules, and basic designs for high-efficiency processors.

Arm has a variety of customer types. Some, like Apple, Qualcomm, and Nvidia itself, use only Arm's instructions sets, rules, and documentation to design their own processors so that their cores will run software written for Arm. Apple has its designs fabricated by manufacturing specialists such as TSMC and Samsung and puts those processors in its own products. Qualcomm has various fabricators that make chips for system vendors to use.

Others—Vivo, for example—license Arm cores, but they do most of their own development independently.

A third group, mostly the smaller players, license Arm processor cores (and other Arm elements, such as the Mali graphics processing unit, TrustZone security technology, and Arm's artificial intelligence IP) and receive engineering and integration support from Arm engineers.

All these chips go into systems (mostly phones) made, in the case of Apple, in-house or by yet other companies. For example, Xiaomi uses Qualcomm Snapdragon processors in its phones. Nvidia, also an Arm licensee, sits near the top of the pyramid with the likes of Qualcomm and Apple. So its move to buy what amounts to the wellspring of technology on which everyone in the mobile market depends rightly scares its main rivals.

Allies and enemies on both sides of the potential deal

But not all the players are fearful. Nvidia has found allies in its quest to take over Arm, primarily among mid-list suppliers that would tend to benefit from Nvidia's proposed more-democratic scheme for Arm's future IP distribution; that is, the players now in pole positions (Qualcomm, Huawei, Samsung) would tend to lose if the deal goes through. Those further down (Broadcom, Marvell, Mediatek) would tend to win, and those three specifically have endorsed the merger.

Apple is a special case; with its architectural license, it does its own development work and is increasingly moving toward a complete proprietary stack, reducing its dependence on Arm over time. Meanwhile, that relationship is likely to continue as before. So, the interested parties are lined up on both sides, with perhaps one or two neutrals.

Meanwhile, the governments that will have to sign off on the deal (United Kingdom, China, the European Union) are experiencing some of the same trepidation. In particular, the UK government (Arm is headquartered in Cambridge, UK), seems to be unhappy about the prospect of an American company scooping up all that, even if we are two countries separated by a single language and have a supposedly special relationship.

The UK Competition and Markets Authority frames its argument in terms of harm to competitiveness and a loss of innovation. The national security argument is somewhat weakened by the fact that Arm has been foreign-owned since 2016, when the Japanese firm SoftBank bought the company. However, the industry always considered SoftBank an honest broker, since it had no competitive conflicts. That being said, the competitive argument still has some bite.

Nvidia has, of course, tried to be reassuring. In late August, after the British said they were launching a full investigation into the deal, a company spokesman said Nvidia intends to "maintain Arm's open-licensing model, serving customers in any industry all over the world, and it will grow Arm's product portfolio by adding Nvidia's IP, creating a broader offering to all customers." He added this, however: "While open licensing does not mean that all customers receive the exact same products or access, open licensing does mean that Arm supports all customers, providing them the solution they need, as soon as it is ready. Nvidia will continue to provide Arm IP to all interested customers as soon as it is ready." He then noted: "We would also say Nvidia relies on and must work with others in the industry, many of whom are Arm's customers and cannot foreclose the competition without overwhelming retaliation."

Nvidia states its case for the acquisition

All of which makes great sense. Nvidia's best arguments in favor of the deal revolve around alternative scenarios. If the company doesn't buy Arm, Arm may have to turn to the public markets to fund capital expansion, with the risk that it might not be able to attain a level of financing equivalent to what Nvidia is offering. As it is, Arm's resources aren't sufficient to fund its ambitious program to penetrate the server market. Nvidia has the capital to fund Arm's programs. There is always the possibility that a white-knight suitor might arrive, but certainly none has emerged waving a bid higher than $40 billion. Companies that could possibly swing it, like Samsung and Apple, also have conflicts with the rest of the industry. So, Arm can't be Arm 2.0 without the kind of investment Nvidia brings to the table.

Nvidia points out that it has a long history of collaboration with competitors—most notably AMD and Intel. It also asserts that: 

  • jeopardizing Arm's revenue stream wouldn't be in the company's interest; 

  • that long-term contracts in place protect Arm's IP licensees; 

  • that Nvidia needs Arm's current licensees as partners to take on Intel and AMD in the server market; 

  • that it will contribute Nvidia IP to Arm licenses; and 

  • that a group effort will be needed to build a new ecosystem around Arm, including software developers, hardware vendors, and other chip designers. These are good arguments.

But if the deal goes through, Nvidia will periodically face the choice of receiving a small licensing fee from a current Arm customer or selling an entire Arm processor of its own to that customer's customer. To put some numbers on it, that could be a $7 licensing fee vs. $350 for the chip. Which one do you think Nvidia's investors are going to want to see Nvidia choose? They'd like something for their $40 billion.

Nvidia makes the vast majority of its revenue from selling chips and clearly would like to sell a lot more of them in the future. But the Arm acquisition would put licensing on the front burner for Nvidia.

Over the next few months, Nvidia will need to convince the relevant regulatory authorities of its noble intentions. Many of Arm's current licensees are also Nvidia's rivals. If Nvidia were able to reassure them sufficiently, they might decrease their objections to the deal. If they lowered their voices, the authorities could perhaps be persuaded to sign off on it.

That's a lot of ifs.

Nvidia has good arguments, but some of its rivals will be difficult—if not impossible—to win over. On the flip side, perhaps the authorities will be persuaded by Nvidia's assertion that its investments in the Arm ecosystem will create more—rather than fewer—choices for customers.

Editorial standards