How to predict and control your AI SaaS cloud spend with Ed Barrow @Cloud Capital

saas.unbound is a podcast for and about founders who are working on scaling inspiring products that people love, brought to you by https://saas.group/, a serial acquirer of B2B SaaS companies.

In episode #4 of season 6, Anna nadeina talks with Edward Barrow, Co-Founder & CEO @ Cloud Capital, a Cloud Cost Management platform helping finance (CFOs) and engineering teams control, forecast, and optimize cloud spending, especially on AWS, by bridging financial and technical visibility.

Cloud costs have gone from predictable line item to one of the biggest financial headaches for founders, CFOs, and engineering teams—especially for AI-first businesses. You can save money by rearchitecting, turning things off, or holding less data. But the fastest way to lower run-rate without draining engineering time is understanding, forecasting, and then using financial instruments to capture committed pricing safely.

Why cloud is a control problem, not just a technical one

Most teams treat cloud as a purely engineering problem. In reality the bigger challenge is control and predictability. By the time the bill arrives, mistakes are already locked in. That makes cloud spend the second largest and least understood cost after headcount for many companies.

Benchmark to check against

• Typical companies spend 10–20% of revenue on cloud infrastructure.
• AI-native businesses often spend significantly more—training and inference are compute intensive, data volumes rise, and unpredictability increases.

The commitment trade-off: discounts versus risk

Cloud providers offer steep discounts if you move from pay-as-you-go to committed pricing (one to three years). That is the single biggest lever to reduce cost without heavy engineering work. But those commitments are financial liabilities: if your architecture or usage changes, you can be stuck paying for capacity you no longer need.

Key tension

• Pay-as-you-go = flexibility, higher unit cost.
• Commitments = lower unit cost, higher financial risk.

As one founder put it:

“Making a commitment for three years feels like making a commitment for 30 years in the startup world.”

How AI changes cloud economics

AI has altered the cloud landscape in three major ways:

1. Higher spend per customer: Inference, model hosting, fine-tuning and data processing multiply compute and storage needs.
2. Greater volatility: New features or a viral product can change usage patterns overnight—what you commit to today may be irrelevant tomorrow.
3. Provider dynamics: Cloud vendors have invested billions in specialized hardware. They want committed revenue to justify that capex, and they will push enterprises and startups toward long-term contracts.

A practical model: forecast, commit, transfer risk

There are three practical steps every founder or CFO should consider:

1. Get a proper forecast

Combine three signals: current spend, growth projections, and engineering roadmap. Model scenarios (conservative, base, upside) and stress test for volatility—what happens if usage suddenly doubles or a major feature gets pulled?

2. Capture committed pricing where it makes sense

Commit to capacity only when your forecast shows stability and the unit savings outweigh the downside risk. Prioritize commitments for steady workloads (databases, core services) rather than experimental or quickly evolving parts of the stack.

3. Use financial partners to transfer downside

If your team lacks the appetite or capital to take on long-term commitments, consider financial instruments that do this for you. These solutions make the commitment on your behalf and assume the downside, so you keep the discount without the tail risk. Think of it as unlocking committed pricing while keeping operational flexibility.

Who benefits most

• Bootstrap and cash-conscious startups that must protect runway without diverting engineering to optimization projects.
• Growth-stage SaaS that want better gross margins and predictable costs for PE or board reporting.
• AI-native teams where cloud can become a cash sink early and unpredictably.

Partnerships as a growth and cost lever

Partnerships are powerful but often misunderstood. Rather than building them top-down as marquee announcements, design partnerships from the bottom up—make sure you help the partner’s sales reps close deals faster and with less friction. When the partner sees concrete value (reduced friction, better win rates, or direct revenue), the partnership becomes self-sustaining.

That same thought applies to cloud vendors: help their sales teams close long-term deals by providing the forecasting, risk tolerances, and instruments that early-stage customers usually cannot provide themselves.

Quick, practical hacks you can apply today

• Turn off idle environments—automate non-production shutdowns at nights and weekends.
• Use committed or reserved pricing for predictable services, but only after forecasting usage.
• Implement cost observability that ties spend to product features, experiments, and customers, not just AWS accounts.
• Model volatility scenarios—what happens if a feature goes viral or you rearchitect mid-term?
• Consider a financial partner to take commitments on your behalf if your team can’t or won’t assume the downside.

How to use AI without creating financial chaos

AI should be an amplifier, not a wildcard. Use AI to improve forecasting, automate routine cost optimizations, and scale processes. Pair domain expertise with AI tools: experts plus AI will outperform AI alone in decision-making and risk control.

Don’t hand AI the steering wheel if you don’t have the underlying knowledge to validate its recommendations. AI speeds execution, so mistakes compound faster.

Founder-level takeaways

• Think of cloud spend as a financial instrument as well as an engineering problem.
• Invest in forecasting early—uncertainty is the enemy of cost-effective commitments.
• Use commitments selectively and consider third-party solutions to transfer downside risk.
• Build partnerships that help individual sellers win, not just CEO-level announcements.
• Use AI to scale talent and workflows, but keep humans in the loop for strategic decisions.