Nvidia's Machine Learning Business Seems Very Strong

Trip Chowdhry of Global Equities Research attended the MXNet and TensorFlow Hackathons and shared his findings in a research report.

According to the analyst's observations, Alphabet Inc (NASDAQ: GOOG) (NASDAQ: GOOGL), Apple Inc. (NASDAQ: AAPL), Amazon.com, Inc. (NASDAQ: AMZN) and Microsoft Corporation (NASDAQ: MSFT) are all significantly expanding their demand for NVIDIA Corporation (NASDAQ: NVDA)'s GPU Clusters.

Chowdhry rates Nvidia, one of the best performing stocks of 2016, with an Overweight rating and $105 price target.

Here are the analyst's main findings:

Industry Demand For Nvidia's GPU

According to the analyst, the greater number of GPU Clusters being created results in more data being ingested, which also creates a larger training model. Accordingly, companies will see improvements in prediction accuracy, which will boost Nvidia's revenue base for its GPUs.

Improvements In Prediction Accuracy From 2011 To 2016

Using ML/DL techniques image recognition in 2011 resulted in error rates as high as 26 percent. In 2016, using ML/DL and large datasets and training models resulted in image recognition error rates of just 3 percent — even better than humans at 5 percent.

Chowdhry suggested that Nvidia's GPUs contributed to the improvements over the years.

Cloud Operators Creating More Clusters

ML/DL initiatives at Google include android apps, drug discovery, Gmail, image understanding maps, natural language understanding, photos, robotic research and speech translation, among others.

The analyst stated that Nvidia is the sole enabler of Google's new initiatives and is helping Google's apps become more intelligent — a major trend for 2017 and beyond.

Nvidia's cuDNN 5.0

Chowdhry stated that all three major DL/ML Frameworks — TensorFlow (Google), MxNet (Amazon) and CNTK (Microsoft) — are seeing very strong adoption in creating the next generation of Intelligent Apps and powering this is Nvidia's cuDNN 5.0.

"NVDA cuDNN continues to be a significant differentiator for NVDA, no one else currently even come close to it," the analyst argued.

Image Credit: By Ra Boe / Wikipedia, CC BY-SA 3.0 de, via Wikimedia Commons