The market for AI products and services is expected to surge, potentially reaching $990 billion by 2027, according to a report from Bain & Company.
As demand for AI-driven hardware and software grows at an annual rate of 40% to 55%, the strain on supply chains is likely to intensify, particularly in semiconductor production.
AI workloads are projected to increase by up to 35% annually, driving the expansion of large data centers, which could cost between $10 billion and $25 billion within five years.
This growth in data centers is expected to boost demand for upstream components like graphics processing units (GPUs) by 30% or more by 2026, raising concerns of a potential chip shortage like the one triggered during the pandemic.
Geopolitical tensions and rising demand could further strain semiconductor supply chains, particularly for next-generation GPUs and chip packaging components.
Beyond GPUs and high-bandwidth memory, other chip types such as power management ICs or networking chips could be affected if growth accelerates in those segments.
Michael Schallehn, partner at Bain & Company, explained enterprises are securing access to GPU chips and GPU computing instances through cloud service providers (AWS, Azure) and GPU-as-a-service specialists including CoreWeave.
“To mitigate potential shortages across other chips, companies should try to understand the type of chips that they are sourcing, and how they are used in areas where AI adoption could accelerate growth such as data centers, AI PCs or smartphones with Gen AI capabilities,” he said.
Secure Supply Chain Resiliency
Schallehn recommended companies monitor these segments—for example AI PCs–for growth inflections while increasing the resiliency of their supply chain by securing access to the chips they need.
“One strategy would be increasing inventories, for example a just-in-case inventory strategy, or long-term supply agreements with chips suppliers,” he said.
He pointed out NVIDIA’s most advanced GPUs are in high demand to train large language models.
“At this point, access to GPUs is impacting the deployment of AI data centers,” he said. “The ecosystem of hyperscalers and colocation providers would be willing to spend more capital expenditure on datacenters if more GPUs would be available.”
The report also projected that as companies increasingly adopt AI technologies, the pressure on data center infrastructure, power and cooling will intensify.
AI and Cloud Continue to Grow
Jason Carolan, chief innovation officer at Flexential, said although AI projects are moving forward, it is important to understand that scaling to production is challenging due to shortages in GPUs.
“Enterprises must be creative when accessing hardware for development, testing and production, by relying on a mix of hyperscalers, tier-2 cloud providers like CoreWeave, SaaS vendors, colocation and system integrators to meet their needs,” he said.
Carolan predicted cloud and AI would continue to have parallel growth tracks for quite a while into the future.
“Both tools meet customer needs and will continue to grow, especially with the upcoming hardware and PC refresh cycle that demands more AI-enhanced capabilities across both CPU and GPU deployments,” he said.
Tim Beerman, CTO at Ensono, also noted some interesting trends and new areas of focus for increasing the processing power needed to handle AI demand.
“Hyperscalers continue to expand their data center footprints to accommodate new forms of data management and cloud services that go hand-in-hand with many AI applications,” he said. “We’ve also seen Microsoft, Google and Amazon recently explore nuclear as a potential solution with new investments in nuclear power production.”
He added all these solutions must be watched closely to ensure environmental impacts and energy usage are considered.
“Regardless, as AI grows, so does the need to power, store and maintain the technology,” Beerman said.