Morningstar | Nvidia Returns to Gaming Roots at CES 2019; New RTX 2060 Well Positioned for Mainstream Gamers
At Nvidia’s CES 2019 press conference, CEO Jensen Huang primarily addressed Nvidia’s bread and butter business: gaming. Most notably, he announced the Jan. 15 launch of its mainstream gaming graphics processing unit on the latest Turing architecture: the RTX 2060 for $349. While this price is above prior mainstream iterations ($299 Founder’s Edition GTX 1060), the performance of the RTX 2060 is substantially above that of the GTX 1060 and even GTX 1070 FE (which launched at $449), showcasing the superior value proposition Nvidia is bringing to its core contingent of gamers. In fact, the firm noted a third of its GTX installed base is on the GTX 960, 970, or 1060, which bodes well for the prospects of the RTX 2060. We are maintaining our $120 fair value estimate for narrow-moat Nvidia. Despite the substantial sell-off in recent months, we continue to view the shares as modestly overvalued.
In our meeting with CFO Colette Kress, she said recent average selling price increases in Nvidia’s gaming segment have been driven by upward mobility along the value stack, rather than straight price increases. Nonetheless, on the November earnings call, Kress cited elevated inventories caused by a cryptocurrency-related hangover as a leading driver for recent gaming GPU weakness, while postulating that the channel would remain bloated for one to two quarters. We think the firm priced its RTX 2060 such that original-equipment manufacturers would consequently cut prices of older graphics cards to accelerate the reversion of channel inventories to a normalized level. Concerning China, Asia-Pacific represents about a third of Nvidia’s gaming business with China serving as a meaningful portion of that, per Kress. However, she expects the upcoming Chinese New Year to buoy demand for the firm’s latest GPUs.
During our meeting, we discussed the data center opportunity for GPUs extensively. When presented with Advanced Micro Devices' launch of its 7-nanometer-based GPUs for gaming and the data center, Kress quickly shot down the competitive threat by claiming the node transition (to 7 nm) in isolation is insignificant, as there are other software architecture-centric aspects more important to the overall performance of the GPU. These include the CUDA platform on which deep-learning frameworks sit and Tensor cores that are programmable for AI-specific acceleration. Benchmark-wise, Kress claimed AMD’s 7 nm data center offering had a fourth of the performance relative to Nvidia’s 12 nm V100. We agree process node is not the sole parameter to focus on when comparing processors, a fact we believe investors should keep in mind when comparing AMD and Intel CPUs.
At the firm’s CES booth, automotive dominated the scene, with Nvidia announcing its DRIVE Autopilot system for "the world’s first commercially available" Level 2+ autonomous driving, with production starting in 2020 with major suppliers such as Continental and ZF. The platform utilizes Nvidia’s Xavier system on chip announced in January 2017. Volvo and Daimler were two key OEMs that are expected to leverage these platforms as well. We saw multiple Nvidia autonomous driving boards at ZF’s booth named ZF ProAI and spanning from L2+ through L4/L5 robotaxi vehicles. ZF is also integrating Xilinx’s multiprocessor system on chip for aggregation, preprocessing, and distribution for sensor and autonomous driving data while providing low latency and high efficiency for accelerating AI performance. This corroborates with our view that a multitude of firms beyond Nvidia, including Xilinx and Intel (via Altera and Mobileye), will play a major role for enabling self-driving, especially for redundancy reasons.
For further insight into our views on the AI and self-driving prospects for chipmakers, please see our technology Observers "Accelerator: Rise of the Machine (and Deep) Learning Phenomenon" and "Automotive Chipmakers in the Fast Lane--Destination: High-Speed Growth."