Get Data Sheet, Fortune’s technology newsletter.

“The security problem is now an analytics problem,” Fink told Fortune days before taking the stage. He outlined his talk in three sections: a brief history of computer security, the changing role of security command centers, and the company’s bet on a proprietary technology with titillating potential that is called, cryptically, “the Machine.” (Read more about that research project in a Fortune article from last year.)

Fink’s first point is well known among present-day cybersecurity professionals. Perimeter defenses—securing one’s network with the digital equivalent of high walls and blockades—are not enough to prevent determined hackers from breaking into computer systems. Baddies will inevitably gain access. The trick is how to deal with them once they’re inside, an argument that leads directly into part two of Fink’s talk.

For more on cyber threats, watch:

Security operations centers, the IT divisions responsible for raising the alarm whenever a violation is detected, must evolve. The teams tasked with staying vigilant are simply flooded with too many incidents, he said. “It’s no longer reasonable to ask a human to look through 2.5 billion events per day,” Fink continued, tossing out an estimate for the number of alerts that a company might process each day. Even though an overwhelming portion will turn out to be false positives, the remainder is still too large for humans to contend with. “There’s just too much there.”

All of this plays into Fink’s coup de grâce: the aforementioned Machine. This technology, he believes, will serve as an indispensable tool in the information security tool-belt. The idea behind the system was to create a new computer design, one based on memory and data rather than a “processor-centric world,” as Fink put it. The Machine is designed to be bigger, better, and badder than anything cybersecurity pros are used to by storing more data, analyzing it faster, and operating at huge scale for its big business customers.

Fink’s team—he also heads up Hewlett Packard Labs, the company’s R&D division, which built the system—is working now to develop a version of the Machine that will hold a whopping petabyte of storage, he said, the equivalent of four times as much as the Library of Congress holds, compared with 640 terabytes of space with the current model. The Machine, which is designed to serve as a kind of howitzer cannon for analytics, is one way how Fink thinks Hewlett Packard Enterprise can help companies reduce the hundreds of days it takes them to detect network intruders on average.

The technology is still under construction, Fink said. Hewlett Packard Enterprise, which is having a rough go after its recent split with HP, Inc. (HPQ), plans to dribble out bits and piece of the Machine to market, with a completion date slated for 2020. It remains to be seen how the company will sell this innovation-in-progress. “In there we build things and prove they work,” Fink said of the R&D arm. “Then we transfer that to a business unit responsible for turning that into a business model.”