Growing AI Adoption Fuels Infrastructure Upgrades

A majority of companies participating in an annual industry survey report they are either using or “experimenting” with machine learning technology for enterprise applications ranging from cloud-based AI services to online shopping.

The AI adoption survey released Tuesday by 451 Research also found that more than half of respondents said their current IT infrastructure must be upgraded in order to scale AI workloads. That finding portends a renewed wave of enterprise migration to public cloud services, the survey concludes.

Text and natural language processing along with computer vision are seen by early adopters as general-purpose AI technologies. In the running also despite growing privacy concerns is facial recognition technology.

Compared with last year, the percentage of corporate AI workloads and pilot projects jumped this year by roughly one-third, 451 Research reported. Twenty-nine percent of respondents moved machine learning applications to production while 28 percent reached the proof-of-concept stage.

“AI, though still a young technology in terms of enterprise adoption, is maturing rapidly and becoming an integral part of everyday experiences,” noted the adoption survey. AI “is already being implemented to improve and automate many business processes from data security to customer experience, and with millions of people now working remote, AI is being recognized by enterprises as a capable tool for workforce optimization and management.”

Machine learning adoption strategies vary. Nearly half of those companies polled favored application development using cloud-based AI services while 25 percent said they purchased applications with machine learning built in.

Indeed, software services and hosted applications were the most frequently cited environments supporting AI and machine workloads, followed by on-premise and hosted private clouds. Those trends are expected to continue over the next two years as prototype efforts move to production.

About one-third of respondents expect “moderate increases” in data volumes used for AI training and inference, with volumes for training more complex models running in the range 50 to 499 terabytes. (A small percentage said data volumes for training and inference exceeded 1 exabyte.)

Respondents were divided on the pandemic’s impact on their AI initiatives: Some said it would slow or halt development while a larger percentage said it would accelerate efforts to automate business processes or add intelligence to enterprise applications.

The financial services, energy, manufacturing and retail sectors (in that order) are expected to record the highest AI technology adoption rates in 2021. The survey found that 82 percent of financial services firms expect to increase new AI initiatives in response to the pandemic.

The survey also poses an intriguing question: Given growing enterprise adoption, why has no pure-play AI company emerged? The market analyst notes that AI services, applications and IT infrastructure generates substantial revenues for public cloud vendors, “but AI is nowhere near being the largest contributor to revenues in any of those divisions.”

Instead, the first pure-play AI enterprises might be Chinese. For example, 451 Research notes that Chinese computer vision specialist SenseTime has so far raised at least $1.6 billion in venture funding. Add to that China’s long-term investment strategy designed to lead the world in AI technology development by 2030.