By Yair Knijn · November 19, 2025
The board asked why cloud cost jumped 40 percent. The CTO had no answer ready.
You can recite the cloud number from memory. The trap opens when you cannot break it down. Reporting total spend in the board pack used to be enough, right up until the number moves the wrong way and the room turns to you.
A 40 percent jump stops being a status line the moment it appears. It is a question now, and the follow-up is the one nobody wants to field cold: why.
Why cloud cost became a board-level metric
Most CFOs now treat cloud spend as a board-level issue rather than an IT budget line, and a large share assume a meaningful fraction is wasted. Azul's 2025 CFO survey put numbers behind both, but you do not need a survey if you have read a recent invoice. AI demand pushed compute past last year's forecast, and finance leaders say that made cost behavior harder to predict.
Once a line item reaches the board agenda, it is a leadership deliverable. The board is not asking you to lower the number. They are testing whether you understand it.
The three questions a board asks about a spike
A board does not interrogate cloud spend the way an engineer does. They want three answers, in order, fast.
- What drove it. Not "usage went up" but the specific service and the specific change behind it.
- Who owns it. A team, a product line, a named decision. "Infrastructure" is not an owner.
- What happens next. A forecast for the coming quarter and the action that bends it, or a flat statement that the spend is intentional and load-bearing.
Miss any one and the answer reads as improvisation. "We are looking into it" admits the spike caught you the same way it caught them.
Attribution: driver, owner, and the line item behind it
Attribution is the work most teams skip until the quarter it bites. A 40 percent jump rarely spreads evenly. It concentrates in one or two line items: a pipeline that started scanning full tables, a forgotten NAT Gateway hauling cross-AZ traffic, an inference workload that fell off spot capacity. The board slide is a sum. The answer lives one level down, in UnblendedCost grouped by service, account, and tag.
That is why untagged spend is a governance failure, not a tidiness problem. If a third of your bill carries no owner tag, then a third of any spike is unattributable on the spot, and on-the-spot is exactly when a routine question turns into a credibility event in front of the people who set your mandate.
Forecast variance the board can actually trust
The follow-up to "why did it jump" is "will it jump again." A forecast built from last quarter times a growth factor will not survive that question, because AI workloads do not grow on a smooth curve. They step. A model ships, a cohort doubles, and the line jumps a tier overnight.
Boards trust variance you predicted. Call the range, land inside it, and the spike is a managed event. Blow past a number you presented with confidence, and the problem is no longer the spend. It is the forecasting, and that is far harder to defend.
Walking in with the answer already prepared
The CTOs who leave these meetings stronger are not the ones who spent less. They walked in already holding the decomposition: this driver, this owner, this line item, this forecast for next quarter. Hand the board that, and the spike becomes evidence that you run the function rather than the other way around.
Cloud Horizons exists to make that decomposition the default instead of a quarterly fire drill. Every workspace keeps spend attributed to a driver and an owner across all your accounts, so the answer to "why did it jump" is assembled before anyone asks. See how the attribution model turns a spike into a sentence you can say out loud.