Big Tech Will Spend $725 Billion on AI Capex This Year, Up 77%, Partly Funded by Layoffs. That's the Solo Consultant's Opening.
Google, Amazon, Microsoft, and Meta plan to spend roughly $725 billion on capital expenditure in 2026, up about 77% from last year's record. Microsoft alone set its calendar-2026 capex at $190 billion. Meta raised its outlook to $145 billion. These are the largest infrastructure commitments in the history of the industry, and a meaningful slice of them is being funded by cutting the same headcount that used to do the work.
That contradiction (spend a fortune on AI capacity, shrink the team that would deploy it) is usually framed as a jobs story. I think the more useful read for a solo operator is a demand story. Companies are buying compute faster than they're staffing to use it, and that gap is where independent builders and consultants get paid right now.
Two things $725 billion does to you
The first is obvious and good: all that compute pushes the price of inference down. The DeepSeek V4-Pro cut, Gemini Flash undercutting last year's frontier at flash prices: these aren't unrelated events. They're what happens when this much capacity comes online and the providers fight to fill it. Your cost of building with AI is on a multi-year downward slope, and the capex boom is the reason. As a builder, that's pure tailwind.
The second is less obvious and more interesting. A company that commits to $190 billion of infrastructure has made a bet it now has to justify internally. Someone has to ship AI features that use that capacity, or the spend looks like waste on the next earnings call. But the same companies are running layoffs. So you get a structural mismatch: enormous pressure to ship AI projects, and fewer people inside to ship them.
That mismatch doesn't resolve itself. It gets outsourced.
Where the displaced demand actually flows
The gap between "we bought GPU capacity" and "we shipped a working AI feature" is wide, and it's full of unglamorous work that big companies are bad at staffing for in a hiring freeze. Wiring a model into an existing internal tool. Building the retrieval layer over a company's messy document store. Standing up an agent that does one specific back-office task reliably. Migrating a workflow off a vendor and onto a cheaper in-house stack now that inference is cheap.
None of that needs a 200-person team. A lot of it needs one competent independent builder who can move fast, doesn't need onboarding, and can ship a working thing in weeks. That's the profile of a solo operator, and it's exactly the kind of scoped, outcome-based work that survives a hiring freeze, because it shows up as a project line item, not a headcount add. A VP who can't get approval to hire two engineers can very often get approval to bring in a contractor to deliver a specific result.
So the practical move, if you do any kind of AI consulting or contract building, is to position around that gap. Not "I do AI." Specifically: "I take the model capacity you've already paid for and turn it into a shipped feature your team doesn't have the bandwidth to build." That's the sentence that maps to a budget line that's actually open right now.
The honest counter-take
I'd be selling you a fantasy if I left it there, so here's the other side. Capex booms have funded an enormous amount of work that never shipped. Half the "AI transformation" budgets announced in the last two years produced a pilot, a slide deck, and nothing in production. A glut of announced spending is not the same as a glut of actual paid projects landing in independent builders' inboxes.
And a layoff-funded boom cuts the other way too. Companies cutting staff to afford capex are companies under cost pressure, and cost pressure is where discretionary spending (including the contractor budget you're trying to win) gets frozen first. The same conditions that create the opening can slam it shut if the macro turns. This window is real, but it is not guaranteed to stay open, and anyone telling you it's a sure thing is selling a course.
What I'd actually do
Don't quit anything on the strength of a capex headline. But if you're already building independently, spend the next quarter getting concrete about the gap you fill. Pick one type of "bought-the-capacity, can't-ship-it" problem (internal retrieval, a single reliable agent, a vendor-to-in-house migration) and become the obvious person to call for it. Specific beats general every time a budget is tight, and budgets are tight precisely because of the $725 billion.
The compute glut is the best tailwind solo builders have had on the cost side in years. The work it creates on the demand side is real but contested, and it rewards the people who name exactly what they ship. Generalists wait for the market to come to them. The window's open for the ones who can finish the sentence "I turn your AI budget into a shipped feature in three weeks."
Author
Lukas
@lukcombinator