Back in mid-2013 Nvidia Corp. announced plans to license its Kepler graphics technology to other chip developers in a bid to increase monetization of its intellectual property and to spread its technologies onto markets and devices that the company simply cannot address. Unfortunately, so far Nvidia has not signed a single IP licensing deal.
“[The GPU core licensing] strategy is very important to us and we are making our latest generation GPUs available to [third parties]. Discussions are happening and that will take time,” said Jen-Hsun Huang, chief executive of Nvidia, during a conference call with financial analysts.
Traditionally, Nvidia has sold certain product lines designed for specific applications. However, there are multiple new device categories emerging these days and in many cases fabless companies just cannot design competitive chips for every possible category. In a bid not to stay away from the expanding market, Nvidia decided to adopt the same business model as used by companies like ARM holdings or Imagination Technologies, which license IP to chip designers and they use it to make application processors for billions of purposes.
Kepler is the basis for currently shipping GeForce, Quadro and Tesla GPUs, as well as the new-generation Tegra K1 mobile processor codenamed Logan. Licensees will receive all necessary designs, collateral and support to integrate Nvidia’s powerful graphics cores into their devices. Nvidia naturally does not announce pricing and terms of licensing, so it is impossible to compare them to terms offered by rivals like ARM, ImgTec, Vivante and others. Nonetheless, Nvidia claims that it is negotiating with potential partners.
“This is not something that happens very quickly because it includes people’s methodologies and designing it into the workflow and so on, but we add so much value here. […] I think we have a real great opportunity to find new growth venues for our company,” said Mr. Huang.
It should be noted that integration of any IP takes time, even though companies like ARM do a lot to shrink it. If a developer begins to design an SoC now and only incorporates already developed IP blocks (i.e., ARM Cortex-A57, two Nvidia Kepler graphics processing clusters, a Rambus memory controller with PHY and so on), it will spend several quarters before the processor can be integrated into commercial products. Given that Kepler graphics is a competitive advantage, the developer will not announce the deal [which reveals its advantage to its rivals] before it has its chip at hands.