Intel has purchased AI (artificial intelligence) startup Nervana Systems, in a move aimed at shoring up its position in a market generally seen as poised for massive growth. Terms of the deal weren't disclosed but one report placed the price tag at about $408 million.

Nervana Systems' deep learning technology spans both software and hardware. It has developed a deep-learning software framework called Neon, which is an open source, Python-based language and and large assembly of libraries. It's also working on the Nervana Engine, an ASIC (application specific integrated circuit) with a unique architectural approach, as the company's website notes

Training a deep neural network involves many compute-intensive operations, including matrix multiplication of tensors and convolution. Graphics processing units (GPUs) are more well-suited to these operations than CPUs since GPUs were originally designed for video games in which the movement of on-screen objects is governed by vectors and linear algebra. As a result, GPUs have become the go-to computing platform for deep learning. But there is much room for improvement — because the numeric precision, control logic, caches, and other architectural elements of GPUs were optimized for video games, not deep learning.

As authors of the world’s fastest GPU kernels for deep learning, Nervana understands these limitations better than anyone, and knows how to address them most effectively. When designing the Nervana Engine, we threw out the GPU paradigm and started fresh.

The Nervana Engine is scheduled for release in 2017. It provides 32GB of on-chip storage and can access up to 8TB of memory per second. 

It remains to be seen how successful the acquisition is for Intel, but on paper, it makes sense.

"One of the four core requirements for an AI platform is computing power," says Constellation Research VP and principal analyst Alan Lepofsky. "Speed is of the utmost importance as vendors strive to differentiate their offerings. The next generation of AI-enabled software requires a backend infrastructure that can run the computations necessary to surface the trends and make the connections between huge sets of data.”

Nervana's focus area within AI is also of importance, notes Constellation Research VP and principal analyst Doug Henschen. "Deep learning is an advanced form of machine learning that’s rapidly advancing computer-based natural language processing, time-series analysis, and speech, image and video recognition," Henschen says. "These capabilities are being applied in areas such as healthcare diagnostics, agricultural research, financial fraud and risk analysis, oil and gas exploration and production, and autonomous driving."
 
Meanwhile, getting computers to make sense of unstructured data "is an incredibly compute intensive challenge," he adds. While Nervana's library of deep-learning models are valuable its proprietary chip technology, which promises a 10x speed advantage, is crucial.

The Bottom Line
"By buying Nevana, Intel is advancing its own chip technology, keeping these advances out of the hands of competitors and complementing its AI ambitions with Nervana’s libraries and cloud-delivery expertise," Henschen says. "If Intel can bake AI capabilities into its chips, developers and analysts can focus on the data science rather than getting sidetracked with setting up frameworks and libraries."

24/7 Access to Constellation Insights
If you’d like unrestricted access to Constellation Insights, consider joining the Constellation Executive Network for analyst advice and analyses that you can use.