This week NVIDIA (NASDAQ:NVDA) unveiled its new DGX-1 deep learning supercomputer at the GPU Technology Conference in San Jose. The small box (that's a compliment) sports 7 terabytes of storage, eight new Tesla P100 graphics processors, two Intel Xenon processors, and packs the same processing power of about 250 servers -- all for just $129,000.
The company is targeting university research groups and has already signed up NYU, MIT, Stanford, and others to use the DGX-1 (textbook markups at the campus bookstores should easily cover the computer's cost).
NVIDIA is staking lot on its new supercomputer, and believes it could help discover "better cancer drugs" and "cleaner, more efficient fuel."
But how exactly will it do that?
What the DGX-1 is built for
NVIDIA says the DGX-1 has 170 teraflops of processing power and is 12 times faster than the company's own deep learning hardware and software of last year.
Deep learning refers to the way some computers process information. The new DGX-1 uses neural network training through the company's Pascal architecture, which allows the supercomputer to process information in similar ways that human brains do.
Deep learning is being used by Alphabet (NASDAQ:GOOG) (NASDAQ:GOOGL) and a host of other tech companies to improve everything from image recognition to self-driving cars. In fact, Google's Raja Monga was at the DGX-1 announcement to talk about how the computer uses Google's open-source machine learning framework, TensorFlow.
TensorFlow is the system that helps power learning-based artificial intelligence (AI) in Google Search, Google Photos, Gmail, and other Google products (you can find more about it here).
But DGX-1 isn't made to sift through emails. NVIDIA is pairing the supercomputer with its new Drive PX 2 autonomous car computer to help power self-driving cars.
Speaking about the company's autonomous car know-how, NVIDIA CEO Jen-Hsun Huang said, "Our technology can track 15,000 important points per second, per camera. It can collect 1.8 million points a second. We load this into sky, register it, and calibrate it using DGX-1." Essentially, NVIDIA will use its on-board Drive PX 2 computer to process information in self-driving cars but also pair it with DGX-1 computers in the cloud.
NVIDIA's deep learning potential
On the stage, Jen-Hsun said that some estimate that the artificial intelligence industry will be worth $40 billion just four years from now and that the broader cognitive computing market could become a $2 trillion industry.
And there's plenty of reason to believe deep learning holds massive revenue opportunity for NVIDIA. A recent report from Tractica says that while some of deep learning's capabilities are overblown, it "ultimately will have a profound impact on how data is utilized across a wide range of industries, disrupting traditional business models while also facilitating new ones." Tractica expects the deep learning computing market to be worth more than $10 billion by 2024.
NVIDA doesn't have a monopoly on deep learning, of course, but with DGX-1's power, relatively inexpensive price tag (compared to servers), and the company's current plans to implement the supercomputer into the ever-expanding world of self-driving autos, I believe NVIDIA is poised to take a dominant position in the deep learning space. And with shipments of the DGX-1 starting in the second quarter, it appears NVIDIA will being building its position sooner rather than later.
Suzanne Frey, an executive at Alphabet, is a member of The Motley Fool's board of directors. Chris Neiger has no position in any stocks mentioned. The Motley Fool owns shares of and recommends Alphabet (A shares) and Alphabet (C shares). The Motley Fool recommends Intel and Nvidia. Try any of our Foolish newsletter services free for 30 days. We Fools may not all hold the same opinions, but we all believe that considering a diverse range of insights makes us better investors. The Motley Fool has a disclosure policy.