Apick — Data-Driven AI Analysis

Published June 1, 2026 • 8 min read • Data from Crunchbase, MIT, Cresset Capital, TechCrunch

AI Startup Valuations Are Crashing — And Nobody Wants to Talk About the $150 Billion Elephant in the Room

The Record That Should Worry Everyone

Let's start with the number that dominates every headline: $150 billion.

That's how much AI startups raised in 2025, according to Crunchbase data. It shattered the previous record of $92 billion set in 2021. The Los Angeles Times called it "a record amount dominated by AI."

But here's what that headline doesn't tell you: $150 billion is not a sign of health. It's a sign of capital burning a hole in investors' pockets while the underlying business models remain unproven.

Venture funding to AI overall hit $211 billion in 2025 — up 85% year-over-year from $114 billion in 2024, per Crunchbase. AI captured roughly 50% of ALL global venture capital in 2025. That's not diversification. That's a bet so large that if it goes wrong, there's no hedge.

And 2026 is accelerating. According to Crunchbase data published April 1, 2026, foundational AI startups raised $178 billion across just 24 deals in Q1 2026 alone — double the $88.9 billion raised across 66 deals in all of 2025. Global VC investment in Q1 2026 hit $297 billion, up 150% year-over-year.

February 2026 alone saw $189 billion raised — the largest single month of startup funding ever recorded.

The Concentration Problem: 4 Companies, 64% of Capital

When you dig into who actually got the money, the picture gets worse.

In Q1 2026, just four companies captured 64% of total funding — roughly $190 billion of $297 billion, according to Crunchbase data compiled by AI Empire Media.

Company	Round Size (Q1 2026)	Valuation	Profit Status
OpenAI	$122 billion	$852 billion	Operating at loss
Anthropic	$30 billion (Series G)	$380 billion	Just hit first profit
xAI	$20 billion (Series E)	Not disclosed	Pre-revenue
Other frontier labs	$8-13 billion combined	Various	Pre-revenue / Pre-profit

Let that sink in. One company — OpenAI — raised $122 billion in a single round. That's 41% of all Q1 venture capital going to a company that, according to TechMarketBriefs, generates $25 billion in annualized revenue but is still projected to lose $14 billion in 2026.

For comparison, OpenAI's $852 billion valuation is higher than the GDP of the Netherlands. And it's still losing money.

This concentration pattern is eerily familiar. During the 1999 dot-com peak, about 55% of capital went to top companies. In Q1 2026, it's 64% to just four AI companies — exceeding dot-com concentration levels.

The $400B Infrastructure Bet With $100B in Returns

Here's the number that keeps institutional investors up at night, per Cresset Capital's December 2025 market analysis:

That's a 4:1 ratio of spending to revenue. In what other industry is that considered sustainable?

Cresset Capital's report describes it as "the most concerning dynamic" in the AI investment landscape. Their analysis notes that hyperscalers committed nearly $400 billion in 2025 capital expenditure while enterprise AI generates approximately $100 billion in actual revenue.

Meanwhile, an MIT study cited by both Cresset Capital and AI Empire Media found that 95% of Generative AI pilot programs fail to achieve business value. Only 5% of enterprises report significant EBIT impact from AI investments, despite 78% of organizations now using AI in some capacity.

"The next 18 months will reveal whether today's infrastructure buildout becomes a platform for lasting innovation, or one of the largest capital misallocations in market history." — Cresset Capital, December 2025

The Human Cost: 3,800 Startups Dead, 113,000 Workers Laid Off

While the headlines focus on OpenAI's $122 billion round and Anthropic's $380 billion valuation, thousands of smaller AI startups are dying quietly.

According to AgentMarketCap's analysis and LinkedIn research on the "AI Graveyard":

• 3,800 AI startups shut down in 2025
• ~1,800 more closed in the first months of 2026
• ~40% failure rate in under 24 months — faster than the typical 50-60% startup failure rate over 2 years
• Over 14,000 new AI startups launched globally in 2024, and nearly half are already gone

The high-profile failures tell a brutal story:

Company	Raised	Peak Value	What Happened
Builder.ai	$445M	$1.5B	Bankruptcy May 2025. AI capabilities overstated; human engineers did the actual work. Microsoft-backed, "AI-washing" scandal.
Humane	$241M	~$850M	AI Pin was "bad at almost everything." Shut down Feb 2025. Sold IP to HP for $116M (51% loss). Customers' devices remotely disabled.
Noogata	$28M	N/A	AI analytics platform. Missed business milestones, couldn't close new funding. Shut down May 2025.
Subtl.ai	Undisclosed	N/A	GenAI knowledge-automation. Strong early interest, but accuracy issues killed retention. Shut down 2025.

And it's not just startups. AI-driven layoffs are surging. According to Founder Reports data: 113,000 tech workers have been laid off in 2026 while AI spending hits $725 billion. Companies cited AI as the primary cause for 54,836 layoffs in 2025. Studies have found no correlation between AI-related layoffs and improved ROI.

The AI Bubble vs Dot-Com Comparison

Every AI bull will tell you "this is different from the dot-com bubble." They're right in some ways — and wrong in the ways that matter most.

Metric	Dot-Com Peak (1999-2000)	AI Peak (2025-2026)
Funding concentration	~55% to top companies	64% to top 4 companies
Revenue reality	Most had zero revenue	95% report zero ROI
Market cap (example)	Cisco: 472x P/E	NVIDIA: 44-47x P/E
Profitability at top	Negative for most	NVIDIA 53% margins (exception)
Infrastructure spend	$500B fiber optic (mostly dark)	$400B AI infrastructure (4:1 spend/revenue)
Capital source	Debt-fueled IPOs	Cash-funded by Big Tech FCF
Retail speculation	Extreme (day trading mania)	Moderate (institutional-heavy)

The key difference, per Cresset Capital's analysis: NVIDIA delivered $99 billion in trailing twelve-month profit with 53% net margins. The Magnificent Seven collectively enjoy 25%+ net margins vs the S&P 500's 13%. That profitability is real. The dot-com peak had nothing comparable.

But the similarities are also undeniable. AI funding concentration exceeds dot-com levels. And at $852 billion, OpenAI's valuation implies investors expect it to be worth more than virtually every public company except Apple, NVIDIA, and Microsoft — despite being unprofitable.

Who Wins, Who Loses

Likely Winners

• NVIDIA — The pick-and-shovel play. $215.9B revenue, real profits, diversified demand. Even if 80% of AI startups fail, they still need GPUs while they're burning cash.
• Hyperscalers (Amazon, Google, Microsoft) — They're spending billions but they're also the ones selling the infrastructure. Azure AI, AWS Bedrock, and Google Cloud are growing.
• Anthropic — First profit milestone (per AI Tools Recap) puts it ahead of OpenAI on unit economics.
• Niche vertical AI — Companies solving specific, measurable problems in healthcare, legal, and fintech with clear ROI.

At Risk

• OpenAI — Losing $14B/year despite $25B ARR. The IPO is being framed as a "trillion-dollar" event but the unit economics don't support it yet.
• Thin wrapper startups — If your "AI" is just a UI on top of GPT-4o or Claude 4, you're already dead. You just don't know it yet.
• AI hardware (wearables, pins, gadgets) — Humane proved that hardware-first AI without a working core product is a $241M bonfire.
• General-purpose chatbots without differentiation — OpenAI and Anthropic are eating this market. There's no room for a "me too."

The Bottom Line: Correction, Not Collapse

The most honest assessment comes from Cresset Capital's December 2025 outlook:

"While the AI sector shows familiar signs of a bubble — lofty valuations, heavy capital inflows, and speculative behavior — its strong profits, consistent revenue growth, and cash-funded infrastructure investments point to a selective correction rather than a systemic collapse."

This is probably right. We're not looking at a 2000-style total wipeout. NVIDIA is a real business. The hyperscalers are generating actual cash. But the gap between the $150 billion raised and the $100 billion in enterprise AI revenue, between OpenAI's $852 billion valuation and its $14 billion in losses, between the hype about "AI transforming everything" and the MIT finding that 95% of pilots fail — that gap is going to close.

When it does, the startups that survive will be the ones that can answer one question without hesitation: "Show me the revenue — and show me the margin."

Everything else is just a press release.