The latest SVB State of the Markets report just came out, summarizing all their venture and startup activity through 12/31/25 or so. This is their 30th edition, and buried in these 34 pages are some hard truths that every founder, operator, and investor in B2B needs to internalize.
I’ve pulled out the 10 most critical learnings. Some will validate what you’re feeling. Others might change how you think about your next round, your burn, or your exit timeline.
Let’s dig in.
#1. The “VC is Back!” Headlines Are Misleading. Strip Out Mega-Deals and Activity is Flat.
US VC hit $340B in 2025, approaching all-time highs.
But nearly two-thirds of those dollars are going to deals over $500M. At the 2021 peak, mega-deals were just 18% of total investment. Venture is far, far more concentrated than it was at the peak in 2021. We’ve seen this in other datasets, and SVB confirms it,
When you look at deals under $100M—you know, where most of us actually live—activity is essentially flat at pre-pandemic levels. About 1,150 deals per month, down from 1,650 at peak.
The takeaway? If you’re not raising a $500M+ round backed by OpenAI-level hype, the market is the same one you’ve been grinding in since 2023. Don’t let the headlines make you think capital is suddenly easy again.
#2. AI Gets 58 Cents of Every VC Dollar. But AI Companies Are Actually Less Efficient.
This one surprised me, and it’s somewhat inconsistent with data from Inconiq and others. We hear constantly about AI enabling lean teams and efficient growth. “Seed strapping.” “Vibe coding.” One-person billion-dollar companies.
The SVB data though tells a different story entirely.
AI companies have:
- Lower revenue per employee ($60K at Series B for AI vs. $89K for non-AI)
- Worse profit margins (-172% at Series B for AI vs. -131% for non-AI)
- Higher burn multiples (5.0x at Series A for AI vs. 3.6x for non-AI)
The median Series A AI company is burning $5 to gain $1 of new revenue. That’s $1.40 more than non-AI companies at the same stage.
Why? Because when capital is cheap and markets are new, efficiency isn’t the priority—winning is. As Greycroft’s Ian Sigalow put it: “The markets are brand new. You want to win them and create a monopoly business.”
But if you’re a founder thinking AI automatically equals capital efficiency, think again. The data says the opposite.
#3. Graduation Rates Have Collapsed. It Now Takes 10 Years to Go From Seed to Series D.
This is the stat that should keep every early-stage founder up at night.
Half as many startups are graduating to the next round within three years compared to 2020. In 2020, 22% of seed companies made it to Series A in 36 months. By 2022, that dropped to under 9%.
And when companies do graduate? It takes much, much longer. At current median times between rounds, it would take a company 10 years to go from Seed to Series D—45% longer than in 2022.
Even the best companies are bridging. Among SVB’s top-performing portfolio companies, 25% have done a bridge or extension round in the last 12 months.
The implication? Plan for longer runways. Assume graduation takes twice as long as historical norms. And don’t be ashamed of that bridge round—everyone is doing them.
#4. One-Fifth of VC-Backed Companies Are Both Not Growing AND Unprofitable. That’s “Dead in the Water.”
SVB ran the numbers on the health of VC-backed tech companies, and the breakdown is sobering:
- 75% are growing
- Of those growing, 63% are either profitable or improving
- But 25% are not growing
- Of those not growing, 91% are also unprofitable
That last bucket—not growing AND burning cash—represents about 20% of all VC-backed tech companies. SVB calls them “dead in the water.”
When a venture-backed company continues to burn cash while revenue declines, the fundraising journey is effectively over. The options narrow to soft-landing M&A, acquihire, assignment for benefit of creditors, or simply shutting down.
If you’re in this bucket, it’s time to have the hard conversation. The market isn’t coming back to save you.
#5. Revenue Benchmarks Keep Climbing. Top Quartile Series C Now Needs $45M.
Investors are in “risk off” mode, and they’re demanding more revenue for their money.
The top quartile benchmarks by series:
- Series A: $6.5M revenue (stable from 2023-2024)
- Series B: $14.9M revenue (up 27% from 2023)
- Series C: $45M revenue (up 65% from 2023)
That Series C number is striking. In 2023, top quartile was $27.2M. Now it’s $45M. If you’re planning a Series C, you need nearly twice the revenue you would have needed two years ago to be considered a top-quartile company.
Revenue multiples have stabilized (and even expanded slightly at Series A and B), but the absolute revenue bar keeps getting higher.
#6. AI Valuations Are in Bubble Territory—And Everyone Knows It.
SVB doesn’t mince words here: “Are we in a bubble? The answer is almost certainly yes.”
The AI valuation premiums are staggering:
- Seed: 62% premium over non-AI
- Series A: 31% premium
- Series B: 33% premium
- Series C: 73% premium
- Series D: 85% premium
The top five US AI unicorns (OpenAI, Anthropic, Databricks, Scale, xAI) are worth more than $500B combined—greater than the cumulative value of all IPOs during the entire dot-com boom.
SVB draws the Netscape/ChatGPT parallel: In the 1990s, the Nasdaq had a five-year bull run after Netscape before the bubble burst. If the AI timeline follows the same pattern, we’re about halfway through the run-up.
The real question isn’t whether there’s a bubble. It’s whether the biggest winners will be big enough to offset the many losers. The VCs are betting yes. History will tell us if they’re right.
#7. Foreign-Born Founders Drive Outsized Value—And Immigration Policy Threatens That.
This data point should be in every policy discussion: 59 of the top 100 highest-valued US unicorns have at least one foreign-born founder. This group represents:
- 59% of companies
- 77% of post-money value
- 71% of funds raised
The top three US unicorns—OpenAI, SpaceX, and Stripe—all have foreign-born founders. In fact, 19 of the top 20 US unicorns have at least one foreign-born founder.
Meanwhile, F-1 student visa applications peaked in 2015 and have dropped 20% since. The issuance rate has fallen by 37 percentage points from its peak.
The pipeline of future founders is at risk. If the US turns off this spigot, the effects may take years to show up—but they will show up.
#8. The IPO Window is Open—But the Rules Have Changed.
After a 216-day freeze (tied for longest since 2010), VC-backed tech IPOs returned in early 2025. Ten tech companies went public in H1 2025.
But today’s IPO is different:
- Bigger: Average revenue at IPO is now $526M (up from $196M in 2010-2013)
- Slower: Average annual revenue growth in the year before IPO is just 9% (down from 28% in 2010-2013)
- More down rounds: 7 of 17 IPOs in 2024-2025 were priced below the last private round
The message from public markets is clear: Show me scale, not just growth. Investment bankers still say the Rule of 40 should be driven by 30% growth, but the market is accepting 9% growth from companies with massive revenue bases.
For unicorns considering going public: Only 25% have the $300M+ revenue typically needed. Only 5% hit both that revenue threshold AND meet the Rule of 40.
#9. LPs Are Under Pressure—And That Affects Everything.
Here’s something founders don’t think about enough: 39% of the typical VC fund’s capital comes from “liquidity sensitive” LPs—pensions, endowments, and foundations that have annual payout obligations.
These LPs have a problem. Paper returns are high, but actual distributions are low. For top quartile funds, the majority of value remains undistributed—even for vintages going back to 2014.
Meanwhile, endowments are seeing spending needs climb. Annual withdrawals grew 6%+ in 2024. About half goes to student aid. And endowment taxes are increasing from 1.4% to 8.0% for the largest institutions.
The “endowment model”—high allocation to private markets predicated on long-term patience—is being stress-tested. LPs don’t want to exit venture, but they need it to pay them back.
This creates downstream pressure for liquidity, earlier secondary sales, and alternative fund structures designed to return capital faster.
#10. M&A is Picking Up—But Outcomes Are Getting Worse.
There’s good news and bad news on M&A.
Good news: The ratio of M&A deals to VC deals hit the highest level in seven years. There are now 8 M&A deals for every 100 VC deals, up from a low of 1 in 20.
More interestingly, 46% of M&A deals in 2025 involved a VC-backed buyer. Unicorns like OpenAI (5 acquisitions since 2023), Databricks (17 acquisitions), Stripe, and Figma are becoming major acquirers.
Bad news: Outcomes are getting worse. In 2025, 90% of deals had undisclosed terms—typically a signal of unfavorable outcomes. Only 7% of deals were sold for a known price at least 3x higher than total VC raised. That’s down from 22% in 2021.
Acquihires are also creating controversy. Google’s $2.4B acquisition of Windsurf’s leadership (but not the company) left shareholders frustrated. Expect more of these structures—and more backlash.
