“Our AI will run in the public cloud”

Chief Data and AI Officer · Cloud infrastructure

For many organisations the first AI projects do start in public cloud, and that's reasonable. But AI runs where the data can go. Ask where their training data lives today, what it costs to move it, and who has to approve that before it leaves the building.

What the buyer says

Our AI projects will all run in the public cloud. That's where the GPUs are.

How to handle it

Listen

Hear the logic: GPUs are scarce, public cloud has them, and the AI team wants to move quickly. Listen for which projects they mean: experiments, or systems the business will run on every day.

Acknowledge

Agree that public cloud is often the fastest way to start, and that access to GPU capacity is a real constraint.

Explore

Ask where the data for these projects lives, how much of it there is, and what moving it costs in time and money. Ask about governance: customer data, where data is allowed to be held, who signs off. Ask what happens to costs when a project moves from experiment to everyday use.

Respond

Position your offer for the stage where it fits: running AI close to where the data lives, or production workloads with steady demand. Don't try to stop the experiments. Plan for what comes after them.

Say this
“That makes sense. It's often the fastest way to get started.”

Say this
“Where does the data for those projects live today, and what would it take to move it?”

Say this
“When a project goes from experiment to everyday use, where do you expect it to run?”

Variations

CTO: “We'll rent GPUs as we need them. Buying them makes no sense.”
Head of Data Science: “Our models are trained in the cloud. That's not changing.”

Questions people ask

How do you respond when a buyer says their AI will run in the public cloud?

Agree that it's often the fastest way to start, then ask where the training data lives, what moving it costs and who has to approve it. Position your offer for production workloads, or for running AI close to the data, rather than arguing against early experiments.

What should you ask a data leader about AI infrastructure?

Where the data lives, how much of it there is, what rules apply to moving it, and how costs change when projects move from experiment to everyday use.