Nvidia's newest chip has a secret weapon in the AI race
- Nvidia's Volta GPU was announced to much fanfare.
- The company has since started shipping its Volta chips to data centers.
- The Tensor Core units of the chips provide Nvidia with a distinct advantage over its competition.
When Nvidia announced its Volta graphics processing units, the stock jumped about 17.8% in a single day.
Now, as the first chips are being sold into the data center market, their competitive advantage is becoming clear.
"A key differentiator for Volta is its Tensor cores, which enable it to process matrix multiplication operations in a highly efficient manner - key for neural network training," Mark Lipacis, an analyst at Jefferies, said in a recent note to clients.
The Volta chips are being bought by data centers because the centers are the main training grounds for the growing artificial intelligence trend. GPUs, like Nvidia's Volta series, help speed up the training process of artificial intelligence systems, specifically by allowing for faster matrix multiplication.
GPUs differ from more traditional CPUs by having hundreds or even thousands of smaller "cores" that can perform small operations, compared to the typically much smaller number of more powerful cores (4-8 in many modern machines) found in CPUs.
Nvidia's Tensor cores are specifically designed to be speedy AI system trainers, and are 12 times faster than the company's previous series of chips, according to Nvidia. It's AI prowess could lead to an outsized share in the growing data center and AI markets, said Lipacis.
Lipacis estimates that Nvidia's revenues will grow 30% year over year in 2017, and 15% in 2018. Because of the success of Volta, he raised his price target for Nvidia to $230, about 16.7% higher than the company's current price.
Nvidia is up 97.75% this year.
