There are some semiconductor companies that I feel are more susceptible to disruption. NVIDIA isn’t one of them. There are numerous reasons why I say this, but the main one is focus and R&D. NVIDIA has been the use case I use when talking with the industry when I give a counterexample to first party semiconductor initiatives of companies like Google and Amazon. While I fully believe it makes sense for some companies to make some of their semiconductor/data center hardware there are many cases where this isn’t a good idea.
NVIDIA and specifically their efforts and investment focus have generally been my counterpoint to the trend to verticalize. The reality is, Google and Amazon, as examples, will never beat NVIDIA in specific semiconductor applications and supporting software. NVIDIA invests massive amounts R&D on hardware and software innovation focused on AI. Companies looking to build some of their own data center hardware focused on AI/deep learning are only investing a fraction of the R&D NVIDIA is which begs the question of whether those companies can do as good of a job in these core applications. More specifically, is it worth the many millions of dollars in investment–which may or may not pay off–when it may be easier and wiser to buy the superior tech from NVIDIA and call it a day.
This is going to be an interesting story to watch when it comes to the data center specifically. Both Intel and NVIDIA are market share leaders in the data center but the biggest cloud platform providers like Amazon, Microsoft, and Google, want to create some of their own silicon to differentiate their platforms and create some lock-in for customers. But, I maintain, NVIDIA may still out-innovate them in core applications, and if those big firms who want to verticalize to a degree can’t offer competing solutions with a competitor who is all in with NVIDIA, then they risk losing sales. It is a fascinating dynamic and one that will be interesting to watch.
AI and Autonomy
Besides the companies I mentioned above, the other example I think is interesting is Tesla. Perhaps using Tesla as an example here will bring the points above I made into more clarity. Tesla was using NVIDIA technology in their vehicles for a variety of things including many of the autonomous features. Tesla has since decided they want to start making their own silicon for autonomous solutions. While I remain extremely skeptical about Tesla’s solution is better than NVIDIA’s, from an end-to-end standpoint, the risk for Tesla is if they can’t compete with other car companies who do go all in with NVIDIA.
Case in point, at their annual GTC conference, NVIDIA announced a partnership with Toyota Research to accelerate the use of self-driving cars. As NVIDIA gets closer with automotive companies, it becomes more likely those companies will use core NVIDIA IP. Moreover, I still have not seen a more complete self-driving solution than the NVIDIA Drive AGX. NVIDIA’s continued investments in total end-to-end self-driving systems continue to put them ahead of others and make me think NVIDIA is well positioned to capture share in the ~1,000 dollars of estimated semiconductor content in each autonomous vehicle. Another way to look at the economic upside of autonomy are the forecasts which estimate the semiconductor opportunity for autonomous vehicles to be ~$60 billion in 2025.
From talks I have had with both automotive industry insiders and component providers, it seems all the major automotive companies are making decisions on where to place some bets and what to own vs. what to buy. Their concern is as the car becomes a computer they risk losing the control of the customer experience they feel they need. This is the one area where I feel their approach the future like old companies not like new ones and a primary reason Tesla could be so disruptive.
When it comes to autonomous solutions, my hope is the car companies make wise decisions and partner with the tech companies who are making massive investments to turn cars into computing platforms. I have little doubt these companies can do it themselves, and the safety concerns alone, which NVIDIA is also addressing, are areas where a wrong decision or the wrong bet on technology can mean life or death for a car company.
There is more to say on this, and I have a number of private research reports on the future of autonomous cars that I’ll summarize in a sector report for subscribers in the coming months.