In January, AI was blamed for roughly 7% of announced US job cuts.
By May, about 40%.
The technology did not get 5 times better in four months. You already know that. What you may not know is where the money actually went.
It went to GPUs. And the people cut were the financing.
And when a stated reason moves that fast while the underlying thing barely moves at all, you are not looking at a technology story. You are looking at a narrative that pays.
I want to walk you through what is actually happening in these decisions, because it is a finance story, not a technology one. It is being decided in capital allocation meetings by people who do your job.
And the version being told publicly is not the version in the model.
Let’s dive in.
Sam Altman calls it AI washing
Challenger, Gray & Christmas has now tracked AI as the single most-cited reason for US job cuts for four straight months in 2026. In April alone, employers attributed 21,490 cuts to AI, roughly a quarter of that month’s total.
Some of these cuts are genuine automation. Support, QA, content production, and routine coding. That work is being absorbed, and the people doing it are losing jobs.
None of what follows is a denial of that.
But look at who is questioning the story.
Cognizant’s Chief AI Officer Babak Hodjat, whose entire job is selling AI transformation, told Nikkei Asia that
“Sometimes AI becomes the scapegoat from a financial perspective, like when a company hired too many, or they want to resize, and it gets blamed on AI.”
Sam Altman calls it AI washing.
A layoff framed around AI reads to investors as technological ambition.
The same layoff framed as over-hiring or soft demand reads as failure.
Executives cite AI when the real driver is that they lack the cash flow to fund AI investment without freeing up capital somewhere else.
It is not a technology claim at all.
It is a funding claim.
Layoffs are not the savings. Layoffs are the financing.
Here is the arithmetic that reframes the entire story, and it comes from the cleanest case study available.
Meta’s total human compensation, every salary, every benefit, and every stock grant comes to roughly $27 billion a year.
If Meta fired every single employee tomorrow, the entire payroll saving would cover less than a fifth of the infrastructure bill.
The capex line is four to five times the whole payroll line. So when 8,000 people were cut in May, that was never a cost-reduction program in any meaningful sense. The savings are a rounding error against the spend.
Meta’s own CFO said so on the Q1 earnings call. Susan Li’s words:
A leaner operating model “will allow us to move more quickly while also helping to offset the substantial investments we are making.”
Offset the investments. That is a CFO describing a financing decision in public, correctly, using the right word.
So when a company says it is reducing headcount because of AI, there are two possible meanings, and they are not the same thing at all.
One is that software now does the work.
The other is that the company wants the payroll dollars for compute.
Only the first is a story about your team’s productivity. The second is a story about your balance sheet, and it is the one running at scale right now.
I have not seen this written anywhere, and this is the reason this matters more than a media-criticism argument.
Capital expenditure does not hit the P&L when you spend it.
It arrives later, as depreciation, spread across the asset’s useful life.
Microsoft faces comparable arithmetic. Oracle’s is worse because its buildout is heavily debt-funded, so it carries interest alongside the depreciation.
Now, let’s put the two facts together.
These companies are cutting operating expense today, in advance, so that the depreciation wave arriving in 2027 and 2028 does not destroy their margins when it lands.
That is not AI replacing workers.
That is a CFO looking at a five-year depreciation schedule and pre-emptively taking cost out of the operating line to protect a future EBITDA number.
When free cash flow falls 80%, something has to give.
What gave was headcount, and the press release said AI.
Both statements are technically true.
Only one of them is the reason.
The Bottom Line
I am not arguing that AI destroys no jobs.
It clearly does, in specific functions, and the people affected are not comforted by the distinction I am drawing.
I am arguing something narrower and more useful to you.
The largest wave of AI-attributed layoffs in history is substantially a capital allocation event. Payroll dollars are being converted into compute dollars, on the judgement that GPUs will generate more return than the people they displaced.
That is a defensible position for a board to take. It is a normal finance decision, made by people in your seat, using your tools.
It is simply not the decision that is being described in public.
Which brings this to your desk, because the same logic is coming for you at your own scale.
You will not be spending $145 billion.
But you will be asked to fund an AI project, and the question of where that money comes from has exactly one honest answer. Something else stops being funded. Usually it is headcount, because headcount is the largest and most flexible line you control.
When that conversation happens, three things are worth holding onto.
Know which decision you are making. Cutting ten roles because software genuinely does that work is one decision. Cutting ten roles to free up budget for an AI investment is a completely different one, with a different risk profile and a different set of things that can go wrong. Both can be right. They should not be confused, least of all in your own model.
Say it accurately internally, whatever goes in the press release. If the board paper says AI efficiency and the model says funding reallocation, you have created a gap between the story and the numbers. That gap will be found. It is usually found by an auditor, a journalist, or an employment lawyer, and none of those are good days.
Watch your own depreciation schedule. If you are capitalizing AI infrastructure, that spend is coming back as depreciation for years. Model it now. The hyperscalers are cutting today for a bill that lands in 2027. If you are building anything meaningful, you have the same shape of problem in miniature and far less room to absorb it.
The headline says AI is taking the jobs.
The cash flow statement says something more specific.
Capital moved, and payroll paid for it.
Every CFO must be able to tell the difference.
And that’s all for today.
See you on Thursday.
Whenever you’re ready, there are 2 ways I can help you:
