$189 billion in a single month. In February 2026, global venture investment hit an all-time record. But 83% of that money went to just 3 companies — OpenAI, Anthropic, and Waymo. Is this an AI gold rush or a bubble? We followed the money.
What Is This?
According to Crunchbase, global venture investment in February 2026 hit $189B. That's more than double the previous record of $78B set during the Web3 boom in November 2021. A 780% increase vs the same month last year ($21.5B). The numbers look absolutely wild.
But dig into the reality behind that number and the story changes. Of that $189B, $156B — 83% — went to just 3 companies.
Looking at full-year 2025, $202.3B was invested in AI, accounting for roughly 50% of all global funding. That's 75% growth from 2024's $114B. AI is effectively monopolizing the venture market.
And it's not just "AI is hot." US startups captured 92% ($174B) of global funding in February. Up sharply from 59% the previous year. Geographic concentration of AI investment is becoming extreme too.
Where Money Is Flowing vs Fleeing
So within AI, where exactly is the money going? We can break it into three main streams.
| Sector | 2025 Investment | Trend | Key Companies |
|---|---|---|---|
| Foundation models | $80B (40% of AI funding) | Mega-round concentration | OpenAI, Anthropic, xAI |
| AI infrastructure | $18B (2x YoY growth) | Rapid growth | Cerebras, CoreWeave, Lambda |
| Agentic AI | $6.7B (estimated) | Fast growth, 10% share | Cursor, Devin, Salesforce Agentforce |
| Vertical AI | $15B+ | Healthcare, legal & finance focus | Writer, Glean, healthcare startups |
| Robotics | $22.2B (69% YoY) | Rising fast | Figure, Waymo, 1X |
Foundation models and infrastructure are attracting explosive investment. Mega-rounds ($500M+) account for 58% of AI funding. Meanwhile, seed funding actually declined 11% to just $2.6B. Big money is betting only on proven players, while early-stage startups find it harder to raise.
Agentic AI is also worth watching. $2.8B was invested in H1 2025 alone, and Prosus CEO declared "we've passed the tipping point for AI agent adoption." Coding agents like Cursor have become the fastest-growing software company in history.
What Makes It Different?
The biggest question is this — Is this a bubble or a real platform shift?
| It's a bubble | It's a platform shift | |
|---|---|---|
| Investment concentration | 83% to 3 companies → extreme skew | Early internet also concentrated on Amazon & Google |
| CapEx vs revenue | Hyperscaler CapEx $400B, AI revenue ~$100B4 | OpenAI ARR $20B (3x up), Anthropic $9B (9x up) |
| Enterprise adoption | MIT: 95% of GenAI pilots fail4 | Enterprise AI revenue $37B, 3x YoY growth3 |
| Seed ecosystem | Seed funding down 11% → ecosystem health concern2 | Stronger early-stage selection → quality improvement |
| Historical parallels | Possible repeat of 1990s telecom bubble | Internet era overinvestment ultimately became infrastructure foundation |
Sendbird CEO John Kim's warning is sharp. "For this investment to make economic sense, multiple trillion-dollar companies need to emerge within 5 years. That has never happened in history." Hyperscalers issued $108B in debt in 2025 alone to build AI infrastructure, and for that money to be recouped, companies need to actually generate revenue from AI.
Goldman Sachs projects AI CapEx will exceed $500B in 2026. Meanwhile, actual enterprise AI revenue is around $37B. Whether this gap narrows is the key thing to watch over the next 12–24 months.
"When the investment time horizon shrinks from long-term value creation to the next fundraising cycle, it's no longer value investing."
— John Kim, Sendbird CEO
Both sides have merit
The bubble argument is right about: Investment concentration is extreme, the CapEx-to-revenue gap is large, and the seed ecosystem is shrinking. The platform shift argument is right about: OpenAI and Anthropic revenue is growing 3–9x YoY, enterprise AI revenue has actually reached $37B, and historically, infrastructure overinvestment ultimately became the foundation for the next era.
The Essentials: How to Read This Trend
Whether you're an investor or a founder, positioning in this market requires understanding a few things.
- Don't be dazzled by mega-rounds
83% of $189B went to 3 companies. The overall market isn't booming — it's extremely top-heavy. The seed and Series A market is actually getting tighter. - Watch the application layer, not infrastructure
Money floods into infrastructure (GPUs, data centers), but actual revenue comes from the application layer. Of $37B in enterprise AI revenue, $19B comes from user-facing products. What you build on top of infrastructure is where the game is won. - Agentic AI and vertical AI are the next opportunity
Foundation model winners are largely decided. But agentic AI ($6.7B) and vertical AI ($15B+) are still early. Startups that deeply embed in specific industry workflows have the opportunity. - "Revenue velocity" is the new benchmark
In 2026, investors look at revenue growth speed, not usage. OpenAI proved "AI can generate revenue" by hitting $20B ARR, and now every AI startup is held to the same standard. - The next 12–24 months are decisive
Goldman Sachs expects the current AI CapEx cycle's validity to be determined by mid-2027. That's when enterprise AI pilots either scale up or get killed. Until then is the positioning window.
Caution: Circular funding structures
A significant portion of AI investment has a circular structure. Cloud companies invest in AI startups, and those startups spend the money back on cloud credits. This isn't actual market demand — it's money going around in circles. It needs to be distinguished from real revenue.
