NVIDIA Acquires Arm: Implications for the Market

On September 13, 2020, NVIDIA announced it had signed a definitive agreement to acquire Arm Holdings from SoftBank for $40 billion in a combination of NVIDIA stock and cash. SoftBank will retain a 10% interest in Arm. While there are regulatory issues to be resolved in the UK, the US and China, NVIDIA anticipates this process will take 12 to 18 months. The process of integrating the two companies will not start until the deal is consummated.
 
The Basics of the Acquisition

• NVIDIA will retain Arm’s company headquarters in Cambridge, UK, and will expand Arm’s presence in Cambridge by establishing an AI research and education center in the UK.
• NVIDIA will retain Arm’s current business model for licensing Semiconductor Intellectual Property (SIP).
  • NVIDIA intends to continue to invest in Arm’s Mali Graphics GPU technology, as well as all the other Arm SIP product types.
  • NVIDIA plans to use Arm’s experience to license NVIDIA’s GPU designs as SIP.
  • NVIDIA intends to continue to use RISC-V CPU SIP internally.
• NVIDIA will continue to honor the agreement between SoftBank and the British government to keep Arm’s headquarters in the UK and to increase headcount in their UK locations.

Industry Thoughts Regarding the Acquisition
 
There has been much speculation about the possible sale of Arm ever since SoftBank signaled last year their intention to either IPO Arm or sell it outright. Many articles have been written that were both for and against this transaction. Companies rumored to be interested included Intel, AMD, Qualcomm, Samsung, Amazon, Google, Microsoft, Facebook and even Netflix were possible suspects for such an acquisition. The reasoning was that their purchase of ARM would lead to a ‘custom’ architecture that would give them a competitive lead in the data center and the emerging AI market. Apple was also mentioned as a potential buyer given that they are already replacing Intel in their desktop and laptop PC products with their own custom architecture. Why not purchase Arm and start to enter the data center as a new market for Apple?
 
As it turns out, none of these opinions focused on what the real motives for this acquisition were: Arm is the pre-eminent SIP company, and any analysis about the acquisition needed to take this into account as a very important point and not an afterthought!
 
Arguments Against the Acquisition
 
For the past 30 years, Arm has been licensing its CPU architecture to all semiconductor and system designers to incorporate into their own designs. Needless to say, Arm dominates the mobile market, and some would say Arm enabled the broad adoption of mobile communications. NVIDIA’s purchase of Arm essentially means semiconductor designers will lose the leading ‘independent’ semiconductor IP provider. Even if NVIDIA carefully maintains a separation of Arm’s business operations, the perception will still create some suspicions regarding favorable pricing, early market access, or special service.    
 
At one time, a purchase price of $55B was being thrown about as the amount SoftBank would require to sell Arm. This is a very big number, and even $40B is somewhat mind-boggling on its face. There is the question of how long it will take to make back this investment? When Softbank purchased Arm in 2016 for $32 billion it was questionable how Softbank would capitalize on such a major purchase. Not all purchases can be viewed for their short-term revenue returns.  This acquisition will most likely be justified for its strategic vision.
 
The Motives Behind the Acquisition
 
NVIDIA has a vision of making their GPU technology available to customers beyond those that only purchase their system’s products. To accomplish this, NVIDIA needs to turn their technology into SIP that is licensable. The problem with this vision is that NVIDIA has no experience in licensing their SIP to the market. In addition, NVIDIA sees an emerging world where AI is pervasive and is in literally every SoC and system on the planet. In acquiring Arm, NVIDIA fulfills three important requirements to make this vision happen:

• Arm excels in creating high-performance, energy-efficient SIP.
• NVIDIA gets access to Arm’s 30 years of experience in licensing SIP and access to their SIP customer base.
• Arm offers the most extensive product development and support ecosystem centered around semiconductor design IP.

Jensen Huang made an especially important point on the conference call that discussed the details of the acquisition. In Huang’s view, the data center is undergoing a transition from being a solitary point for the collection, processing and dissemination of data to becoming fractionalized around different workloads and servicing different data types. This is a pretty powerful observation about the market and about the role AI is going to play in it.
 
There will no longer be a ‘one size fits all’ version of the data center. This is already starting to happen, and the trend will accelerate. The data center will be resized and distributed into the infrastructure closer to where the data lies and will transform into servicing more specialized functions such as dealing with only IoT data, automotive data or HPC data. This fractionalization places the compute resources closer to where the data is created and reduces latency and maximizes bandwidth efficiency.
 
So, what type of silicon architecture do you need to realize this vision? A silicon architecture that uses a pervasive SIP type, is economical on power, combines the best-in-class SIP for AI and GPU cores and has a powerful interconnect structure. Semico’s own forecast for SoCs shows unit volumes approaching 44 billion units by 2025. While this forecast includes SoCs that are AI-focused, it does not include the synergy that the pairing of NVIDIA and Arm are going to provide to the market.
 
The fractionalization of the data center is going to present many more opportunities for system, silicon, software and SIP revenues both in the servers in these data centers and also in all the devices that connect to them. This represents at least a $200B - $300B opportunity and is the holy grail NVIDIA is seeking. After all, Arm is the puzzle piece that unites all these different elements into a cohesive strategy necessary to attack such a large opportunity.
 
At the end of the day, Semico believes the acquisition of Arm by NVIDIA creates the potential for important and powerful advances in computing for the AI, IoT, networking, automotive and 5G markets. This is an exciting development in the electronics and semiconductor industries and promises many great things to come.
 
Put in these terms, this deal makes sense and all that was needed to put the picture in focus was the right vision to unite the different elements each company possesses into a vehicle big enough to address the larger concept – AI always and everywhere – the key to a brighter future for all!