At the Reuters Global Technology Summit, NVIDIA’s CEO Jen-Hsun Huang as confirmed that his company wants to license its graphics technology (IP) to other mobile chip vendors, thus opening a new business for NVIDIA. This is an idea that we first heard about during the GTC conference in March 2013. Although it has occasionally sold its IP for others to build chips (like the RSX graphics chip for Sony’s PlayStation 3), NVIDIA’s current business model is to sell “chips” rather than IP.
This is a model that has served NVIDIA very well in the past decade. Since PC components are standardized and mostly interchangeable, the company could focus on “performance” or other metrics important to its customers, and when the benchmarks warranted a victory, NVIDIA would get the contract. Little by little, the PC graphics industry went from more than 50 companies down to essentially three on PC and 5-6 if you include mobile graphics companies. During that time, NVIDIA terminated the PC business of companies like Imagination Technologies which has then successfully re-invented itself as a mobile graphics IP provider by scoring contracts with most mobile chip makers.
This is not a “PC World”
The mobile world is different: it has grown around the fact that handset makers have the financial means to build their own system-on-chip (aka SoC) based on IP from ARM (CPU), Imagination Technologies (GPU) and other players. This allows many of them to build processors that fit their specific needs, budget and schedule. Today, the potential to sell chips is lower as Apple and Samsung hold nearly all the market share that NVIDIA is trying to reach (the mid-range and high-end smartphones with good multimedia capabilities), and since they build their own SoC there is little incentive to buy a chip made by someone else, even if there is an occasional disparity in terms of performance and features. Even newcomers like Huawei and ZTE have announced plans to build their own chips instead of trying to save time and resources by going to a 3rd party. Note that ZTE is rumored to build a Tegra 4 phone in addition to using Qualcomm chips.
The LTE modem is one key to “mobile chip paradise”
In reality, that’s not completely true: Qualcomm has done an excellent job at selling its chips and if you look at the current high-end Android smartphones, Snapdragon 600 has made a clean-sweep as it got nearly all the contracts. In some market, the Samsung Galaxy S4 is powered by Samsung’s own Octa 5 which performs really well, but it can’t be used worldwide because of one thing: LTE support and carrier qualification. That’s where the Qualcomm machine comes in with a complete solution, and things like RF360 should further improve its offering.
Because Samsung’s modem has not yet been approved by many carriers, and since the process can be very long and tedious, Samsung and others choose to opt for a Qualcomm platform for which the modem has been already qualified. Since Snapdragon 600 offers a very comparable performance to the Samsung’s Octa 5 processor (which uses a BIG.Little architecture), most users don’t see any difference in user experience. In this instance, handset makers aim for perfect “time to market” (worldwide launch) rather than trying to help their own chip divisions. It’s unclear if Qualcomm can hold this position forever, but for now, it seems to be quite defensible.
The Icera modem: would it turn NVIDIA into the next Qualcomm?
NVIDIA too has an LTE modem since it has acquired ICERA. It’s a great programmable design that is even qualified on AT&T’s LTE network, but at the moment, it has not yet turned into a best-seller and it has not been integrated into a high-end SoC. NVIDIA has already said that this will come with the next-generation product. In the long-run, this may allow NVIDIA to compete with Qualcomm on a more equal footing, but for now it’s a theory that needs to be proven.
In the short-term, licensing its technology would allow NVIDIA to expand its mobile footprint, increase revenues and become a stronger player in the industry. Right now, it looks like only Samsung or Apple could truly provide NVIDIA with what it wants, but players like Huawei or ZTE could be good partners as well
Why would chipmakers license NVIDIA’s IP?
The motivation for customers to license NVIDIA’s IP would come from two things: First, a potential gain/edge in gaming experience/performance. Since Qualcomm’s graphics performance is very high too (look at our Snapdragon 800 benchmarks), this is not yet a point that NVIDIA can use efficiently, but exclusive in-game features and effects could also entice customers to license the NVIDIA graphics core. TegraZone and The SHIELD console is all about providing this for NVIDIA, but this has not yet caused any real griefs to Qualcomm.
Secondly, customers could access next-generation computational photography (CP) hardware and software. Again, Qualcomm has started communicating about CP as well, although at CES 2013 NVIDIA’s has demonstrated new capabilities that have not been matched since, like fast subject tracking, always-on high-dynamic range (HDR) and more. I think that NVIDIA could play a strong hand here, if it can move before others start catching up. Its imaging knowledge could give it a definite advantage.
Conclusion: good move
Since there is no easy way for NVIDIA to “muscle its way in” and to change the status quo, licensing its IP would allow it to quickly expand its footprint to become a larger player in this market. Additionally, NVIDIA could approach Apple and Samsung semiconductor division, as a partner rather than a competitor. In some ways, the strategy is “if you can’t beat them, join them”, and in a market dominated by a few giants, it is a much better strategy than poking them in the eye. Think about it: would you want something like the Samsung Exynos Octa 5 with the gaming and photographic performance of Tegra 4?