Marking a dramatic shift in business strategy, Nvidia announced today that it will begin licensing current and future GPU and modem technology to other silicon vendors. This includes the current Kepler GPU product family, as well as the modem technology acquired from Icera. Admittedly, this opens up Nivida to competition for the company’s Tegra family of processors from silicon vendors using Nvidia technology. However, it will create opportunities for other silicon vendors to use the technology in targeting other markets and applications not targeted with Tegra.
With the addition of Nvidia licensing GPU technology there are now five major GPU IP licensors, including ARM, DMP, Imagination, and Vivante. However, two of those companies also offer CPU cores--ARM and Imagination with the recent acquisition of MIPS. Theoretically, developing both the CPU and GPU technology should provide and edge for ARM and MIPS, but in the case of GPUs, Nvidia has the advantage of developing some of the most advanced graphics technology that power high performance PCs and workstations. It would stand to reason that AMD, the other major PC and workstation GPU vendor, might also benefit from a similar IP licensing model, which it once did before selling the GPU IP group to Qualcomm, which now refers to the product family as Adreno.
So, the real question is why is Nvidia making the change in the company’s business model when its major competitor did the opposite and it may create more competition for the company’s other products? The official statement from Nvidia is to address changes in the market, which include vertical integration and the proliferation of computing technology into other markets and applications, such as embedded and new applications referred to as the internet of things. One likely factor not mentioned is that lack of volume sales of Tegra. Nvidia has raised the competitive bar with both the introduction of Tegra and the move to an annual cadence of new products.
But unfortunately, Tegra has suffered from being a step behind companies like Qualcomm and MediaTek in integrating Wi-Fi and cellular modem technology, which is becoming a critical competitive factor in the mobile segment. In addition, Tegra has not secured the high-volume design wins because of the dominance of a few vendors that produce their own silicon solutions, mainly Apple and Samsung. The outlook does not look much brighter because other emerging OEMs like Huawei and ZTE are developing their own processing solutions in-house or through subsidiaries. In fact, the success of the Apple model has led most mobile OEMs to consider the vertically integrated business model. The market dynamics have also led several semiconductor companies to abandon the mobile segment, including Freescale, TI, and ST-Ericsson.
If you add the rather bleak to modest outlook for the traditional PC market, it is clear that any change in business model that may increase revenues will be welcomed at Nvidia. The change is a better alternative than abandoning future development and the mobile market altogether. The shift by Nvidia also demonstrates the challenges that all high-tech companies face in adapting to a rapidly changing business environment that requires innovation not only in technology, but also in business models.
Jim McGregor is founder and principal analyst at Tirias Research.
You are correct Rick. NVIDIA indicates that a key reason they are adopting this business model is to allow others to pursue these and other custom opportunities because it can be a very difficult business. It also allows NVIDIA to focus on key markets for GPUs and Tegra.
Here's an excerpt from a report by Ross Seymore, analyst for of Deutsche Bank:
NVDA’s attempts to monetize this IP in its own mobile chips (Tegra)... has not delivered a significant return on investment due to hyper competition among tablet/smartphone SoC vendors. The royalty model appears better suited to the tablet/smartphone SoC market.
NVDA’s move is also a tacit admission that generating a meaningful return on investment is very challenging for fabless ARM SoC vendors. Given fabless vendors do not control their own costs and source key IP from others, there are very few barriers to entry or opportunity to differentiate. A direct illustration of this trend can be seen in Chinese vendors such as Rockchip and Allwinner rapidly gaining share and putting downward pressure on ASPs as they operate on much lower margins than US peers.
Interestingly, the next Microsoft, Sony and Nintedo consoles all use competing Radeon graphics chips from AMD.
This news takes some of the heat of the fact Nvidia just lost some big licensing deals for Kepler.
This is a good point, but with a high number of start-ups in emerging regions, particularly china, and a plethora of new applications in the internet of things, there is room for both. It's also important to note that NVIDIA is shipping products for mobile and automotive applications, so they are not giving up the bank or on the future of Tegra. The real questions are: how far with vertical integration go and can 3rd party IP or SoC vendors find a way to capitalize on the trend if it continues?
IP business is difficult, this is repeat of SGI and MIPS. Unless IP is decoupled from the company, people will not use this. This gives more time for NVDIA to get other pieces while generating revenue from IP. I doubt any big company will fall for this unless the IP is seperate entity... I could be worng...
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.