Nvidia and Tesla have worked together since the Model S first went on sale. Nvidia's Tegra 3 powered the Model S's console, and the two companies worked together on self-driving vehicles. Nvidia crowed about Tesla's test-runs last year, declaring that its own Drive PX2 represented a "supercomputer" that would accelerate self-driving car research and deployment. Intel, however, is trying to muscle its way into what Nvidia undoubtedly regards as its turf.
Tesla will switch its driving and infotainment systems over to Intel, rather than relying on Nvidia to provide that technology, Bloomberg reports(Opens in a new window). Earlier this month, there was a rumor that AMD was providing Tesla with a semi-custom chip to handle self-driving tasks. Later reports claimed that this was false and AMD was unwilling to refute or verify these claims. The rumored Tesla partnership would be a huge publicity boost for AMD's nascent AI and machine learning division, but the fact that AMD is so new to this market implies that the rumor isn't true. Nvidia and Tesla have been working together on self-driving cars for years, and Tesla is unlikely to jump ship for a new company unless there's a serious, unresolvable problem with Nvidia or Nvidia's equipment.
While Nvidia is still powering Tesla's self-driving hardware, Intel's entrance into this market is a significant threat. The CPU manufacturer has been relatively quiet about its AI and machine learning research divisions, but its purchases and product launches speak for themselves. In August 2016, Intel bought Nervana, a cloud software developer with a customizable deep learning service. In September 2016, it bought Movidius, a specialized company with low-power chips for computer vision and deep learning. Intel's Xeon Phi family is already capable of running these workloads in high-end HPC systems or servers, and yesterday it announced Loihi, its first custom neuromorphic chip.
Intel's new neuromorphic chips,
Right now, revenue from vehicle partnerships is often small; Nvidia reported $146 million in earnings from its partnerships earlier this year. That's certainly not nothing, but it's not huge, either. But over time, this market will grow. Global car sales in 2016 were 88.1 million, and the overall market grows respectably each year.
The long-term goal is to put processors in the consoles and self-driving capabilities of a huge chunk of the market, and Intel wants to be part of that push. We're not saying Intel is guaranteed to succeed in its drive to displace Nvidia from some of these capabilities long-term, whether in Tesla or some other vehicle--but we wouldn't count the company out, either. Intel has the funds to develop solutions that compete against those from other companies. While its GPU cores are slower and less advanced than its competitors, it also has the funds to develop better solutions than what it currently fields. Loihi is one example of that, and likely not the last.
