Most cloud cost forecasting guides assume you have a FinOps team. They talk about variance analysis, rolling forecasts, and budget reconciliation. They're written for finance departments, not engineering teams.
But most teams don't have a FinOps person. They have a CTO who needs to know: "Will we stay within budget this month?"
Here's how to forecast cloud and AI spend without a finance team.
Start with Baselines
A baseline is what normal looks like. If you spent $8,000 last month, that's your baseline. If you spent $7,500 the month before, and $8,200 the month before that, your baseline is roughly $8,000.
This isn't sophisticated, but it's better than guessing. Most teams' cloud spend is relatively stable month-to-month. If you're not launching major new features, your baseline is a good predictor.
Account for Trends
Baselines assume the future looks like the past. But if your spend is trending up, your baseline is too low.
Look at your last three months. If spend went $7,500 → $8,000 → $8,500, you have a trend. Your forecast should account for that trend, not just use the average.
A simple trend forecast: take your last month's spend and add half the month-over-month change. If you spent $8,500 last month and $8,000 the month before, add $250 to get $8,750.
Handle Confidence
Forecasts are guesses. Some guesses are better than others.
If your spend has been stable for six months, you can be confident. If it's been volatile, you can't. A good forecast includes a confidence range: "We'll spend between $8,000 and $9,000 this month."
This range accounts for uncertainty. It tells you when to worry (if you're already at $8,500 on day 15, you're trending high) and when to relax (if you're at $7,500 on day 15, you're on track).
AI Spend Is Different
AI API costs are harder to forecast than cloud costs. They're usage-based, not capacity-based. A single feature launch can triple your OpenAI spend overnight.
For AI spend, baselines are less useful. Instead, forecast based on:
- Active features: What's using AI APIs right now?
- Planned launches: What's coming this month?
- Usage trends: Is token usage growing?
If you're launching a new AI feature this month, your forecast should account for that. Don't just extrapolate from last month.
Keep It Simple
You don't need complex models. You need something that:
- Gives you a number (forecasted spend)
- Gives you a range (confidence interval)
- Updates as the month progresses (pace vs. forecast)
Most teams can do this with a spreadsheet or a simple tool. The key is consistency—forecast every month, track accuracy, and adjust.
Automate When Possible
Manual forecasting is tedious. It's also error-prone. If you can automate it, do.
A good forecasting tool:
- Calculates baselines automatically
- Detects trends
- Provides confidence ranges
- Updates daily as the month progresses
This doesn't replace judgment. You still need to account for planned launches, seasonal patterns, and one-time events. But it gives you a starting point.
The Bottom Line
Forecasting doesn't require a FinOps team. It requires:
- A baseline (what normal looks like)
- Trend awareness (is spend growing?)
- Confidence ranges (how uncertain are you?)
- Regular updates (how are you pacing?)
Start simple. Get consistent. Automate when you can. That's enough to avoid most surprises.