The AI Bill Just Came Due, and Finance Noticed

Something shifted in the last two weeks of earnings, and it will land on your desk soon. The cloud giants are still growing fast. But the checks they're writing for AI have gotten big enough that Wall Street stopped clapping. When the companies with the deepest pockets get grilled on return, the pressure rolls downhill to you.

Let me catch you up on what changed and what to bring to your next budget review.

The growth is real, and so is the bill

Azure grew 43% last quarter and crossed $100 billion in annual revenue for the first time. That is not a company in trouble. Amazon's AWS grew 28%, its fastest in nearly four years. Google Cloud jumped 82%. Demand for this stuff is not made up.

But look at the cost side. Microsoft's capital spending hit a record $41 billion in one quarter, and free cash flow dropped 23% even as profits climbed. The company expects to spend roughly $190 billion this calendar year. About two-thirds of that goes to GPUs and other hardware that wears out fast.

So both things are true at once. The revenue is arriving, and the spending is arriving faster.

When the leader posts negative cash flow

Google is the sharpest example. Alphabet made nearly $120 billion in revenue last quarter and still reported negative free cash flow of $5.9 billion, the first time that number went below zero since the company went public. It spent $44.9 billion on infrastructure and only generated $39.1 billion from operations.

Then Google raised its 2026 spending forecast to as much as $205 billion, well above what analysts expected. Its CFO warned free cash flow would stay under pressure and that spending climbs again in 2027. The stock fell 7% the next day.

Read that again. A company beat expectations and got punished anyway. The market decided the spending is running ahead of the payoff.

Finance stopped taking "trust us" for an answer

The old pattern was simple: spend big, promise AI returns, get a pass. That pass is gone. Elizabeth Lopatto put the mood plainly when she wrote that AI is finally expensive enough to make Wall Street nervous, pointing out that Google essentially admitted it can't forecast its own costs.

The skeptics have numbers now. Moody's said the six largest cloud and AI platforms will spend about $785 billion this year and close to $1 trillion in 2027, then added the line that matters: "the ultimate return on investment is unclear." Even Amazon, where Andy Jassy insisted the company isn't spending $200 billion "on a hunch", watched its free cash flow fall to $1.2 billion from $25.9 billion a year earlier.

Your CFO reads these headlines. The ROI question they ask about your AI line is the same one Wall Street is now asking the giants.

Seats sold is not value delivered

Here is the gap that should worry you most, because it mirrors your own rollout. Microsoft 365 Copilot passed 30 million paid seats, up from 20 million last quarter. Sounds great. But that's still under 7% of the roughly 450 million commercial Microsoft 365 seats. Investors have poked at that gap all year.

Buying seats is easy. Getting people to use them and produce something worth the money is the hard part. Microsoft knows it, which is why it launched a $2.5 billion effort to put 6,000 engineers inside customer organizations to help them actually deploy the tools. Amazon and Google made similar moves.

If the vendors are spending billions just to drive adoption of what they already sold, assume your own seat count is telling you less than you think.

The deep cut

Watch where the investment gains came from, because it's the tell finance will use on you. Microsoft's blowout quarter included a $3.2 billion gain from its Anthropic stake, nearly matching the full-year gain on OpenAI in a single quarter. Google's record net income leaned on roughly $98 billion in mostly unrealized gains from its investment portfolio, not its core business.

So when you build your AI case, separate the two things finance will separate. One column is the paper value of bets and partnerships. The other is money the work actually made or saved. Come to the review with the second column filled in, tied to a specific team, a specific task, and a real before-and-after number. If you only have seat counts and vendor promises, you're walking in with the same weak hand that just got Google's stock knocked down 7%.

Three questions for your team

1. For each AI tool we pay for, can we name one workflow where usage is high and output improved, with a number, not a vibe? If not, why are we renewing?

2. What share of our AI budget is a bet on future capability versus spend tied to work happening today? Can we defend that split to finance next quarter?

3. If our CFO cut the AI line by a third tomorrow, which purchases survive on proven return, and which ones are we keeping on faith?