AI PCs may finally make sense for enterprise buyers

AI PCs are starting to look less like a niche experiment and more like a practical way for companies to control AI spending. New Gartner research says buying more capable PCs could help enterprises avoid some of the unpredictable costs tied to cloud-based AI, especially as data center buildouts fall behind demand.

According to the report, pressure on supply chains and growing opposition from local communities to new data center projects are creating constraints that could push metered compute costs higher than some businesses expect. In that environment, shifting part of an AI workload onto employees' machines offers a simpler financial model: a one-time PC purchase instead of ongoing cloud token fees.

Gartner argues this makes AI PCs a good fit for a hybrid compute model. Rather than replacing the cloud, they could handle lighter tasks locally while more demanding jobs still run in hyperscaler data centers. The firm says workloads such as speech and chat, text generation, and image and audio generation could increasingly move onto workers' PCs.

The shift also lines up with broader model trends. Smaller language and reasoning models, including versions trained for specific business use cases, require far less compute than frontier models and are therefore better suited to local execution.

Recommended reading

Rosatom to build CNC motors in Russia by 2027

Gartner expects that by 2029, around 30% of enterprises could use AI PCs to cut cloud AI token costs. By 2030, 70% of corporate PCs could be able to run some GenAI tasks locally.

Omdia has also reported rising use of small and medium models, especially for domain-specific tasks that do not need maximum-scale compute. As Alexander Harrowell, Senior Principal Analyst for Advanced Computing, put it:

“Older GPUs are retaining value and remaining in service, as they continue to offer a cost-effective option for small and midsized model inference and disaggregation,”

Via The Register