The merger of NVIDIA and ARM announced this week will change both the CPU and GPU landscape dramatically and have a material impact on AI capabilities by the end of 2021. It will change the competitive landscape for PCs and servers while creating some new opportunities and risks in the smartphone space. It can also impact everything from robotics and autonomous vehicles to how OEMs buy core technology.
Let’s explore these points this week.
NVIDIA + ARM: Regulatory Hurdles
The U.S. should approve this merger with few reservations, given that it's a U.S. company acquiring a firm that was based overseas. However, the EU and China may have issues. The EU may consider ARM a European asset and want conditions that assure that jobs and headquarters remain in Europe for the acquired firm. NVIDIA has already indicated this is in the plan, but the EU may require that NVIDIA formally commit to this before approval. I don’t see a problem with that request given this is the stated plan, so the EU’s approval is all but inevitable.
China, which is in a trade war with the U.S., may feel that this merger puts them at a disadvantage and have their unique conditions about control and ownership. Getting China’s approval, therefore, will be more difficult because they’ll want assurances that the U.S. Government doesn’t somehow gain inordinate influence over the technology. NVIDIA has no control over the government and must comply with U.S. rules and regulations. This concern will be harder to overcome because of the lack of trust between China and the U.S. and will make coming up with enforceable conditions problematic. And this is an election year in the U.S., meaning that getting anything done will be problematic, mainly if it means leaving the ARM jobs in Europe. I don’t expect this problem to be insurmountable, but NVIDIA’s merger team is likely to find most of their effort tied to getting China’s eventual approval.
Changed Competitive Dynamics
Up until this merger, there were three comprehensive core technology component vendors for personal computers and smartphones. They were Qualcomm (much heavier in smartphones than PCs), AMD (PCs and game systems), and Intel (PCs only). With this merger, NVIDIA joins this group, and all of these competitors have efforts that span PCs and servers, particularly for AI. Most of them have interests in robotics, autonomous cars, and IoT platforms as well.
The interesting dynamic will be between Qualcomm and NVIDIA given that Qualcomm is a licensing entity -- and NVIDIA, with ARM, becomes one as well. NVIDIA has indicated they will pivot to more of the licensing model over time for a broader cross-section of offerings, and Qualcomm is a significant ARM licensee. Except for their AI work, NVIDIA and Qualcomm don’t compete with each other that much and will effectively be partners, under the ARM licenses, once this merger is complete.
Now the two firms will have an exciting choice: strengthen the partnership, make it strategic and leave things as they are, or Qualcomm abandons ARM and creates the CPU technology that they license. The first option may have some antitrust exposure but could result in some rather interesting joint opportunities. The second option, which is also the most likely, would require NVIDIA not to abuse their power and Qualcomm to be judicious concerning how they share their technology with ARM (this would be similar to how car companies often get engines from competitors).
The final option would be wicked expensive for both firms and likely weaken Qualcomm significantly before, and if, their new CPU technology got to critical mass in the market. Since Qualcomm is a licensing entity and knows how to manage these relationships, I expect they are comfortable continuing to license from the merged NVIDIA/ARM entity. Given Qualcomm’s recent challenge to their licensing model by the FTC, having another large company using a similar licensing model emerge should reduce Qualcomm’s exposure in this area long term. Qualcomm’s recent appellate court win also did much the same thing so the benefit to them isn’t as great as it would have been pre-appellate judgment.
Intel and AMD
Now, this merger potentially places ARM and X86 against each other broadly while transitioning NVIDIA into a licensing powerhouse. AMD's custom business and lack of a fab arguably makes them a licensing-like entity and places Intel as the odd man out in terms of business model. If Intel allows broad licensing of X86, it will reduce the justification for Intel to own its FABs and force one of the most significant changes to its business model in history. If they hold, they will increasingly be standing alone, and while, at their size, they can certainly do that, with their general support for open technologies, resisting the move to more of a licensing model will be difficult.
If Intel does more aggressively, license firms will be given a choice between licensing ARM and licensing X86. If the PC and server OEMs then choose to license, Intel will either have to open their FABs to this business, or they likely won’t sustain volume at a high enough level to support those FABs. As a result, this merger may impact Intel the most. But it could merely accelerate Intel down a path they were already following.
AMD has been an ARM licensee in the past and is currently the largest X86 licensee. As noted, their custom business does already put them on an exciting path, but they don’t own and can’t license X86 to others since they are under license from Intel restricting that. However, they could license their additive X86 technology to the OEMs to enhance the offerings the OEMs produce, as these OEMs compete with their peers. This change would open up new licensing opportunities for AMD but, like Intel, reduce the market opportunity for AMD's parts.
Wrapping Up: The Move To Licensing
Overall I think this merger will force a broad move from producing core parts to licensing the technology to produce their own. We already have cloud vendors working to produce their core components, often licensed from ARM. If this merger forces Intel and AMD to more aggressive licensing, it could both accelerate this trend and offset these efforts that currently don’t use X86 technology, making it more competitive as a licensed technology.
One thing is certain, though; this merger is going to change the market for core technology products across several segments, including cloud, PC, smartphone, AI, IoT, robotics, autonomous vehicles, and security. I doubt there will be a single tech company that isn’t significantly impacted by the resulting changes.
Rob Enderle has been a TechnologyAdvice columnist since 2003. His areas of interest include AI, autonomous driving, drones, personal technology, emerging technology, regulation, litigation, M&E, and technology in politics. He has an AS, BS, and MBA in merchandising, human resources, marketing, and computer science. Enderle is currently president and principal analyst of the Enderle Group, a consultancy that serves the technology industry. He formerly worked at IBM and served as a senior research fellow at Giga Information Group and Forrester.