What the $740B AI Capex Boom Teaches You About ROI
The $740B AI capex surge is the perfect live case for mastering ROI, payback period, and hurdle rate before your McKinsey or BCG interview.
The single most-tested quantitative skill in a McKinsey, BCG, or Bain first round is judging whether a massive investment actually pays back. Bloomberg named $740 billion in big-tech AI capital expenditure as the defining equity theme of 2026, up 69% year over year. That number is not background noise. It is a live, high-stakes case study sitting on your interviewer's mind right now.
This guide breaks down ROI, payback period, and hurdle rate using the AI capex boom as the working example. Read it, and you will know how to structure an investment payback question in under 60 seconds and defend your answer like you have run the analysis before.
Why Big Tech Is Betting $740B and What That Question Means for You
Every dollar of that $740 billion is a bet that future cash flows will exceed the cost of building data centers, buying chips, and wiring the infrastructure together. The interviewer's version of this question is always the same: "Is this a good investment?" Your job is to answer that with a structure, not a shrug.
When a firm like Google or Microsoft commits $100 billion to AI infrastructure in a single year, the board approved it by comparing the expected return against a threshold. That threshold is the hurdle rate, the minimum return the company demands before deploying capital. If the expected return clears the hurdle, you invest. If it does not, you walk.
Think of the hurdle rate like a high-jump bar. The bar is set by the company's cost of capital, typically its weighted average cost of debt and equity. Every proposed investment has to clear it. A $50 billion data center investment that returns 8% annually fails if the company's cost of capital is 10%. The bar wins.
How to Calculate ROI the Way a Consultant Does
ROI is the ratio of net profit from an investment to the cost of that investment, expressed as a percentage. The formula is straightforward: (Net Benefit minus Cost) divided by Cost.
Here is where candidates go wrong. They treat "net benefit" as a single number handed to them. In a real case, you build it. For an AI data center, net benefit includes incremental revenue from new AI products, cost savings from automation, and avoided future infrastructure spend. Each of those requires its own mini-estimate.
In the AI capex context, Microsoft's bet is that every dollar spent on Azure AI infrastructure generates more than a dollar in cloud services revenue over the asset's useful life. Analysts tracking Azure's AI revenue growth in 2025 at roughly 35% year over year suggest the early returns are tracking above plan. That is your real-world benchmark to anchor an estimate in a case.
Payback Period: The Question Behind the Question
Payback period answers a different question than ROI: not how much do you make, but how fast do you get your money back. The formula is simple: Cost divided by Annual Net Cash Flow.
Picture a toll road. You spend $2 billion to build it. Drivers pay $400 million in tolls every year. Your payback period is five years. After year five, every dollar of toll revenue is pure return. Before year five, you are still recovering the construction cost.
AI infrastructure follows the same logic. Nvidia reported that hyperscaler customers are running their H100 clusters at near-100% utilization, which compresses the payback period because the annual cash flow from those assets is maximized. A fully utilized $1 billion cluster generating $300 million in annual margin pays back in roughly 3.3 years. The same cluster at 60% utilization pays back in 5.5 years. Utilization is not an operational detail. It is a core driver of the investment case.
Practice this framework on a real case: the why-enterprise-ai-strategies-fail case on BoardroomIQ puts you in the room.
Apply what you just learned
Browse all 100 real boardroom decision cases.
Hurdle Rate and When to Kill a Good-Looking Investment
A high ROI does not automatically mean a good investment. You have to compare it to the opportunity cost of the capital, and that is the hurdle rate's job.
In 2026, with U.S. 10-year Treasury yields above 4.5%, the risk-free rate is meaningfully higher than it was during the 2020 to 2022 buildout cycle. That raises the hurdle rate for every technology investment. A data center project that looked attractive at a 7% return in 2021 may not clear a 10% hurdle in 2026. Same project, same cash flows, different verdict.
This is exactly the trap interviewers set. They give you a project with a positive ROI and wait to see if you think to ask about the hurdle rate. If you do not ask, you have answered the wrong question.
How to Practice This Framework Before Your Interviews
Build a three-line investment model by hand. Take any public AI capex announcement, pick a company like Meta or Amazon, estimate the annual cash flow the investment must generate to achieve a 10% ROI, and work backwards to a utilization assumption. Write the math in three lines on paper. Doing it with real numbers once is worth reading about it five times.
Practice stating your hurdle rate assumption out loud. In your next mock case, before you calculate ROI or payback period, say: "I am going to assume a hurdle rate of X% because..." This single habit signals to interviewers that you think like an investor, not a calculator.
Time your structure. Set a 60-second timer and practice verbally framing an investment payback question: state the formula, name the key drivers, flag the hurdle rate, and identify the one assumption that swings the answer. Speed and structure together are what interviewers call "executive presence."
The best way to practice this framework is under realistic pressure, with a case that fights back. Open the why-enterprise-ai-strategies-fail case on BoardroomIQ and work through the investment decision before you read the solution.