On January 16, 2026, PitchBook and the NVCA released the 2025 year-end data. The headline looks like a victory lap: U.S. VC deal value hit $339.4 billion, nearly matching the 2021 mania.
But the aggregate number is a lie.
Here is the line that actually matters: 50% of that capital was invested in just 0.05% of deals.
In a pool of ~35,000 transactions, a handful of giants (OpenAI, Anthropic, xAI, Databricks) hoovered up half the planet's venture dollars. This isn't a recovery; it's a two-tier system where "winners" get sovereign-scale checks and everyone else gets a liquidity desert.
The 59-sec takeaway:
Venture capital has permanently split.
Market A (The AI Elite): 0.05% of deals. 50% of capital. Infinite LP demand.
Market B (Everyone Else): 99.95% of deals. 50% of capital. Collapsing deal counts. Zero institutional interest.
If you aren't in the "Mega-round 68," you aren't in a recovering market. You are in a survival game.
The 5 Signals Informing my take
Signal #1: The 2,000x Capital Gap
U.S. VC deal value hit $339.4B, but half of it went to 0.05% of deals (mostly OpenAI’s $40B round and Databricks' $4B Series L).
The Reality: If you have 10,000 deals, 5 deals split $169.7B ($34B each). The other 9,995 split the remaining $169.7B ($17M each).
The Take: Remove the top 5 AI companies, and 2025 looks exactly like the brutal "nuclear winter" of 2023.
Signal #2: AI is a Capital Black Hole
In 2016, the top 10 AI companies took 1% of AI investment. By 2024, they took 51%. Today, 25% of ALL venture capital flows to just 10 AI companies.
The Take: If you’re not building a foundational model or a "Project Prometheus"-scale infra play, you aren't just competing with other startups, you're competing with the gravitational pull of $84B flowing into 5 firms.
Signal #3: The Death of the "Bridge"
Deal count fell 60% year-over-year for non-AI sectors. Seed funding dropped 30% in 2025.
The Reality: VCs are writing far fewer checks, but the ones they do write are massive.
The Take: The Series A bridge has collapsed. If you raised seed in 2024, your 2026 Series A odds are <10% unless you are AI-native with proprietary data.
Signal #4: The "Tournament" Economy
More than 33% of total global funding went to just 68 companies that raised rounds of $500M or more.
The Reality: Close to 60% of all invested capital went to just 629 companies.
The Take: VC is no longer an "innovation engine", it's a tournament where the top 0.2% of players get 250x more capital.
Signal #5: Institutional "Zero Interest"
Robot Ventures partner Anirudh Pai notes that institutional LPs now show "Zero interest" in non-AI deals.
The Reality: VCs are publicly admitting what founders already know: if you aren't AI, you're "statistically un-fundable."
The Take: Even Thomas Krane (Insight Partners) admits that while non-AI assets may be "attractive," they lack the defensibility required for the 2026 venture model.
Where I could be wrong
Blind Spot #1: This is just a standard "Infrastructure Phase"
The Counter-Argument: Every tech cycle begins with massive concentration. In the late 90s, capital pooled into Cisco and Sun Microsystems to build the pipes. In 2010, it pooled into AWS and iOS. The argument is that once OpenAI and Anthropic finish the "base layer," capital will inevitably diffuse into thousands of application-layer startups.
Why I’m Skeptical: In previous cycles, infrastructure players (like Cisco) didn't compete with their customers. OpenAI does. With the release of Sora and their agentic "operator" tools, the foundational labs are vertically integrating at light speed. They aren't just building the pipes; they’re building the water, the sink, and the faucet.
When the 5 biggest labs raise $84 billion in a single year, they don't just buy GPUs—they buy market share in every vertical from legal to coding.
Blind Spot #2: Non-AI companies can still build "Good Businesses"
The Counter-Argument: Thomas Krane (Insight Partners) recently noted that non-AI companies can still be "very attractive assets."
There is a world of profitable, growing SaaS companies that don't need a GPU cluster to survive. Bootstrapping and niche market leadership are still valid paths to success.Why I’m Skeptical: This is a newsletter about Venture Capital, not small business ownership. VC is a game of 100x power laws. If institutional LPs show "zero interest" in non-AI funds, then the bridge to Series B and beyond for non-AI founders isn't just "higher", it’s gone.
You can build a great $20M revenue business, but in the 2026 climate, if you aren't "Agentic-native," you aren't a venture-scale asset. You’re a "Zombie" in the eyes of the cap table.
Blind Spot #3: The "Series A Crunch" is just a flight to quality
The Counter-Argument: Some analysts argue the 60% drop in deal count is a healthy "culling of the herd." They point to the fact that top-tier founders are still raising $20M Series A rounds. The bar is simply back to historical norms: $2M–$5M ARR and 100% growth.
Why I’m Skeptical: The 2025 data shows down-rounds hit a decade high of 16%. That isn't a "higher bar", that’s a structural breakdown. Seed funding fell 30% last year. If the "on-ramp" is shrinking while the "Series A requirements" are doubling, we aren't seeing a flight to quality; we’re seeing a bottleneck that will starve out an entire generation of innovation that doesn't fit the "Megaround" mold.
Blind Spot #4: Capital concentration is actually "Efficient"
The Counter-Argument: Maybe US should be pouring 25% of all capital into 10 companies. AI is an arms race. If the U.S. doesn't concentrate capital into OpenAI or xAI, they lose the sovereign tech race. This concentration accelerates the arrival of AGI, which will eventually lower the "cost of intelligence" for everyone else.
Why I’m Skeptical: Concentration of capital leads to concentration of power. When 0.05% of deals take 50% of the money, the "innovation ecosystem" becomes a "monopoly ecosystem." We are effectively letting 10 CEOs and 5 VC firms decide what the next decade of technology looks like. That’s just a planned economy.
The 2026 Forecast: What to Watch
Q1 2026 Concentration: If the Top 10 deals capture >30% of capital, the bifurcation is accelerating.
Down-Round Peak: 16% of 2025 deals were down-rounds. If this hits 22% in Q1, expect mass liquidations.
The "Zombie" Reckoning: 2021 vintage companies that can't raise a megaround will be forced into M&A at 2-3x revenue multiples by Q3.
Now What?
For Founders: If you aren't AI-native, stop chasing traditional VC. Pivot to bootstrapping, alternative capital (Family Offices), or pre-negotiate an M&A exit while you still have leverage.
For VCs: Stop pitching "broad recovery." Narrow your focus to vertical specialists or secondary funds. The middle is dead.
For LPs: Demand a "Concentration Report." If your GP isn't in the top 0.05% of deals, they are likely sitting on a portfolio of 2021 "zombies."
Your Turn:
Are you seeing the bifurcation in your own fundraising? Reply with your take.
This week’s sources:
→ PitchBook-NVCA Venture Monitor
→ JP Morgan - AI leads venture activity, IPOs building momentum
→ Economic Times - OpenAI close to finalising its $40 billion SoftBank-led funding
→ Databricks Announces Over $4B in Funding at $134B Valuation to Accelerate Enterprise AI Platform
→ Private Equity / Venture Capital Deal volumes at 10-yr high in 2025
→ AI startups captured over 50% of venture funding in 2025: Report
Talk next week,
Pavan
P.S. Kyle Stanford (PitchBook): "If you look beyond AI, dealmaking isn't hitting new highs." Translation: The "rebound" is a myth for 99.95% of you. Adapt accordingly.