A strong answer. “My starting assumption is that engineering is probably right that it's necessary, and also that nobody knows which 40% is the necessary part. Those are compatible, and that's the useful frame — it turns an argument into a shared question.

The first thing I'd want is spend per unit of business, not spend in rupees. If revenue grew 50% and infrastructure grew 40%, that's operating leverage improving and I should say so plainly rather than flagging it as a problem. If revenue grew 5%, that's a different conversation entirely. Absolute cloud growth is almost never the right metric and finance teams flag it constantly.

Then attribution, which is where this usually falls apart. Most companies can't tell you cost per customer or cost per product line because nothing is tagged. So I'd work with engineering on tagging — not to police them but because they usually want it too; they just have never had a reason strong enough to prioritise it.

Then I'd look at the shape. Typically some is genuine growth, some is a specific product that is much more expensive than we priced it, and some is waste — idle non-production, over-provisioned instances, old snapshots. The third is real and it is rarely 40%.

The finding that changes decisions is usually the second one: a product line whose gross margin is far worse than the blended number suggests. That's a pricing conversation, not a cost conversation, and it's worth more than any efficiency work.

And I'd give engineering the visibility rather than a target. Teams that can see their own spend reduce it without being asked; teams handed a target cut whatever is easiest to cut.”