Just this week Hewlett Packard Enterprise (HPE) completed its purchase of Cray, the tech company that built the very first supercomputer in 1976. HPE acquired Cray for $35 per share as announced in May. The transaction value of the deal went up to $1.4B because Cray had less cash on its balance sheet at the time of close.

Image: The State of Enterprise AI Adoption 2019, Cray

HPE’s purchase reflects industry momentum behind artificial intelligence (AI) including predictions that the global high-performance computing market growing from $31 billion in 2017 to $50 billion by 2023. To crunch the massive amounts of data that powers AI solutions, corporations, universities and governments will need the power of the super computers that Cray builds.

The Department of Energy has selected Cray’s Shasta supercomputing systems, and Slingshot interconnect for the first three exascale systems in the United States.

HPE says that high performance computing is key to the company’s vision and growth strategy as it provides the backbone to process, analyze, and extract insights from massive amounts of data. In a press release, Phil Davis, president of Hybrid IT at HPE said that by combining the deep expertise and R&D engines of HPE and Cray, the teams are better positioned to help customers solve the most data-intensive challenges today and into the future.”

IT professionals also are preparing for the AI transformation that is coming. Cray recently surveyed IT professionals to measure the impact of artificial intelligence on the workplace. “The State of Enterprise AI Adoption 2019” survey showed that all business units are launching AI projects: IT support, research and development, operations, customer service support, marketing, and sales. AI is “crossing the chasm from theory to adoption,” the report said.

Sixty-seven percent of respondents predict that AI will improve operational efficiency, and 25% say that the technology is already critical to business operations.

According to the survey of IT professionals, the top barriers to implementation are:

  • The cost of infrastructure, talent acquisition, and application development
  • A lack of technical expertise
  • The time to value

The survey showed that data challenges such as availability, access, quality, and cleanliness were no longer a barrier to implementation. However, expertise barriers still exist.

SEE: Deep learning: An insider’s guide (free PDF) (TechRepublic)

IT professionals are working to address the technical expertise barrier with 72% saying they have spent time learning about the technology. These educational activities include:

  • Attending conferences or training – 48%
  • Attending vendor webinars – 41%
  • Taking self-study courses – 40%
  • Downloading and reading reports – 35%

Survey respondents are confident that they will rise to the challenge of AI: Only 32% were afraid they would not be able to keep up with the pace of change.

The survey also shows a mix of on-site and cloud-based installations for AI projects. Fifty-three percent of respondents use the cloud for some of their AI workloads. However, 36% want to transfer cloud-based projects to on-site installations. Performance was the deciding factor at 39% when making an infrastructure choice for AI projects.

About 35% of the 319 survey respondents work for companies with 501-2,500 employees and about a third of the respondents were in healthcare, financial services and manufacturing.

Earlier this year, the Department of Energy announced that the first exascale computer will be built at DOE’s Argonne National Laboratory in Chicago. “Aurora” will be capable of an exaFLOP of performance – a “quintillion” floating point computations per second. Cray will build Aurora with technology and architecture from Intel.

Cray built the first supercomputer in 1976 and smashed all previous performance records thanks to a fast processor clock, successful vector implementation and a large 1 Mword memory.

Image: The State of Enterprise AI Adoption 2019, Cray