Computer scientists at the Calif.-based lab plan to begin trying out the $1 million device on Thursday, the Wall Street Journal reports. The neural network-mimicking array consists of 16 computer chips, called IBM TrueNorth, that in total contain a million “programmable neurons,” and 256 times as many electronic equivalents of synaptic connections.

Get Data Sheet, Fortune’s technology newsletter.

“The potential capabilities neuromorphic computing represents and the machine intelligence that these will enable will change how we do science,” said Jim Brase, deputy associate director for data science at the lab, in a statement.

IBM (IBM) has been working to develop these artificial intelligence-friendly microprocessors even as it unceremoniously off-loaded its chip-making division two years ago. The company—along with competitors such as Google (GOOG), Nvidia (NVDA), Qualcomm (QCOM), and Microsoft (MSFT)—aims to power the next generation of machines capable of deep learning, an AI technique that requires complex computer processing.

For more on artificial intelligence, watch:

One appealing aspect of these systems is that they require far less energy to operate than traditional computers. A TrueNorth chip burns a little more than one tenth of a Watt, two orders of magnitude less than a typical computer server chip.

Part of the reason for this is that TrueNorth chips are not always “on.” Instead, they kick into action when needed, best handling tasks such as recognizing images and undertaking “smart” calculations of the sort that allowed Google’s AlphaGo program to declare victory over a champion Go player in a recent tournament of the ancient Chinese board game.

Read more: “The Current State of Artificial Intelligence, According to Nvidia’s CEO”

The lab said it plans to use the machine’s unique potential to explore applications in national security, nuclear defense, and cybersecurity, among other areas.