You open LinkedIn on a Tuesday morning. Every other post says AI is transforming business. Your CEO gave a keynote about "AI-first strategy." Your department bought three AI tools nobody uses. Everything is changing. Allegedly.

Then you look at the actual numbers. And the numbers tell a very different story.

Three headlines. Same punchline.

Headline one: a San Francisco AI lab called Humans& — three months old, no product — raised a $480 million seed round at a $4.48 billion valuation in late March 2026. The founders used to work at OpenAI, Anthropic, Google, xAI, and Meta. That's the entire pitch. "We once worked at places that built things."

Headline two: a Duke University/Federal Reserve survey of 750 CFOs found that 502,000 jobs will be cut due to AI this year — nine times more than 2025's 55,000. But here's the twist that makes it art: 60% of executives admitted they fired people in anticipation of AI replacing them. Only 2% said AI was actually doing the work.

Headline three: Microsoft is down 21% year-to-date, spending $120 billion on AI infrastructure. Their flagship AI product, Microsoft 365 Copilot — an AI assistant baked into Word, Excel, and Outlook — has 15 million paid users. Sounds big until you do the math: that's 3.3% of their 450 million subscribers. 96.7% of customers said "no thanks."

Three stories. Same punchline: the AI industry is drowning in money and starving for results.

The money firehose

The funding numbers in early 2026 are genuinely unhinged. OpenAI raised $110 billion in February — the largest private fundraising round in human history — at an $840 billion valuation. A valuation is what investors think the company is worth, and in this case, they think it's worth more than most countries' GDP. Anthropic closed $30 billion at $380 billion. According to Crunchbase, February alone saw $189 billion in total startup funding, vaporizing every previous record.

Forty new unicorns — startups valued over $1 billion — minted in 2026 so far. AI companies are hitting unicorn status at seed stage, which is like getting a Michelin star before you've opened the restaurant. A $480 million seed round for a company that's existed for 90 days isn't investing. It's a religious offering. You're not buying equity — you're buying a prayer that these ex-BigTech researchers build AGI before they burn through half a billion.

The jobs paradox

The Fortune/Duke CFO survey is the most honest document in tech this quarter. The headline again: 502,000 AI-related job cuts expected in 2026. About half from white-collar positions — the office workers, analysts, and mid-level managers who were told AI would be their assistant, not their replacement.

But only 44% of CFOs plan any AI-related cuts at all. And of the executives who did cut, 60% did it anticipating AI gains that haven't materialized. Only 2% had large reductions because AI was actually performing the work.

Read that slowly: companies are firing humans to make room for AI that doesn't work yet. Pre-firing. Anticipatory layoffs. "AI might do your job someday, so pack your desk now."

Study co-author John Graham invoked Solow's paradox — an economist's 1987 observation that "you can see the computer age everywhere but in the productivity statistics." Same observation, 39 years later, shinier hardware.

The real winners

Here's what's actually happening beneath the hype. The infrastructure providers — NVIDIA with 75% gross margins, memory manufacturers sold out through year-end, utility companies powering data centers — are printing money. They sell shovels during the gold rush. The AI profits flow to hardware and electricity, not software.

The software giants? Bleeding. Microsoft's per-seat licensing model — charging per employee per month — faces an existential question: if an AI agent does the work of ten people, you need one-tenth the seats. Salesforce and Adobe are seeing double-digit pullbacks. Open-source models — free AI models anyone can use and modify — from Meta and DeepSeek keep narrowing the gap between what you can get for free and what you're paying for.

The uncomfortable truth

As of March 29, 2026, we're not in an AI revolution. We're in an AI anticipation economy. Companies aren't deploying AI — they're deploying the idea of AI as cover for decisions they wanted to make anyway. Cut costs? Blame AI. Raise prices? AI infrastructure needs. Fire people? AI will handle their jobs. Eventually. Maybe.

The 2% figure should be framed and hung in every boardroom in America. Two percent of companies had AI actually replacing work. The other 58% of executives who cut headcount were running on vibes and LinkedIn thought leadership.

The funding numbers have 1999 energy — the dot-com bubble era — but with better GPU specs. The AI revolution is real. The technology works and keeps improving. But the AI economy — the part where money goes in and results come out — is a mess. And 502,000 people just lost their jobs because a CFO wanted to sound "AI-forward" on the quarterly earnings call.

The revolution will be real. The bill is already here. The receipts are still missing.

ai-bubble, ai-jobs, microsoft, startup-funding, ai-economy