BMO executive tells CNBC to “move away from use cases” as AI pilots lag real deployment targets

AI deployment has become a central test for the next phase of the semiconductor boom, and a senior BMO executive running live AI initiatives says the transition from demonstrations to scaled infrastructure is taking longer than many investors assume.

Speaking on CNBC, the executive argued that the market should “move away from use cases” framing and focus instead on what it takes to build repeatable, scaled systems. She said that three years of AI pilots have not yet become the kind of infrastructure deployments that would change the economics at enterprise scale.

In the remarks, she pointed to 2027 as the year that needs to change. The implication is that the next 12 to 18 months are a proving ground, where pilots either convert into ongoing infrastructure spend or fade into limited experiments.

The timing matters because the comments were made in advance of NVIDIA earnings. NVIDIA is the leading supplier of accelerated computing hardware used in training and inference, and its quarterly results are watched as a barometer for whether AI demand is shifting from early adoption into broader production deployment.

NVIDIA’s business has long been tied to the pace at which customers operationalize AI, including by building data center capacity around its platforms. For the market, the key question is not whether companies have AI pilots, but whether they fund durable infrastructure such as server capacity, networking, and software workflows that support continuous use.

What the public comments did not clarify is which industries or customer segments are most affected, or whether the bottleneck is primarily budget approval, engineering integration, data readiness, or operational constraints. The CNBC segment, as described, focuses on the gap between pilots and scaled infrastructure rather than providing project-level detail.

For NVIDIA and the broader AI hardware supply chain, a sustained “pilot-to-production” gap would typically mean slower conversion of interest into recurring orders. On the other hand, if the 2027 timeline reflects internal enterprise ramp plans, it could also indicate that the market’s expectations for conversion timing may need to reset.

Investors watching the upcoming earnings will likely look for any management commentary that connects demand to deployment stages, such as customer capacity build-outs, new design wins, or evidence that existing pilots are expanding into production environments. The immediate takeaway from BMO’s executive is that proof of scaling, not proof of concept, is the inflection point that the market may still be underestimating.