In its Surviving the global chip shortage report, Forrester recommended that IT leaders moderate their AI ambitions. “Moderation seems like heresy in the current AI-drenched tech environment, but you might have no choice,” the authors of the report warned.
Forrester pointed out that AI requires immense compute capacity, with a big emphasis on graphics processing units (GPUs). The report noted that cloud providers and large system providers have struggled to acquire enough GPUs, which means enterprise IT buyers may have to wait up to a year to get what they need.
“On-premises hardware providers like Dell and Hewlett Packard Enterprise, and cloud providers like Amazon Web Services, Google and Microsoft, scramble to deliver on this demand. The chip shortage is moderating these ambitious roll-outs in datacentres,” the report stated.
While hardware providers have been able to keep up with demand, Forrester pointed out that they are pushing their limits. For enterprise IT leaders considering purchasing AI-optimised hardware, the analysts warned: “A bit more demand will push them over the edge.”
Looking at how IT leaders can work around the chip crisis as they build out their plans for AI, Forrester recommended starting small with tech plans. “The current silicon environment validates that strategy even more. You will make mistakes as you embark on any new journey. Focus on small steps – therefore small mistakes – that will yield big lessons as you scale up later when supply returns,” the analyst advised.
Forrester predicts a shift from the datacentre to edge computing in 2024, with more AI processing distributed to edge compute devices and PCs. For instance, the report pointed to Intel’s Meteor Lake processor for PCs, which incorporates circuitry aimed at AI processing, as an example of AI edge computing. AMD, Nvidia, and various ARM-based platforms are also planning to target this market opportunity, according to the analyst. Forrester expects these edge platforms to be even more transformative than cloud-based offerings, though the two will co-exist.
Beyond AI, Forrester urged IT leaders to use “just in case” inventory planning to ensure they have enough PCs and servers, rather than purchasing as and when they need refreshing using “just in time” replenishment. The analyst firm recommended IT leaders also consider running servers for longer. “Five-year-old technology is usually seen as antiquated, which is why amortisation plans often spread costs over that period and then you send that infrastructure to retirement,” the report stated.
The analyst firm believes many IT leaders will likely need to keep such old systems running for a bit longer. Some might choose used hardware to augment existing systems. While such hardware will not be as efficient as brand new servers, Forrester noted that the hardware already exists in the IT inventory of an organisation, whereas new hardware may need to be purchased. However, the analyst recommended that IT leaders account for dependability risks of older technology if they choose this upgrade strategy.