In May 2026, reports surfaced that Anthropic was in talks to raise at a $900 billion valuation. OpenAI had been valued at $852 billion in a March funding round. Cursor — three years old, $2 billion in ARR — was in talks at $50 billion, a 25x revenue multiple.
Then a Series B founder we work with came into our weekly meeting with a polite question: their last round had priced them at 12x ARR. The market was apparently paying 25–50x. Why was their valuation so low?
The honest answer was: it isn’t. The AI valuations are. And if you build your fundraising strategy around what the four most-watched AI companies in the world are being valued at, you’ll either burn a year chasing a number that doesn’t exist or end up taking a down round you didn’t need to.
This post is for founders, operators, and finance leads trying to make sense of the current valuation environment — and trying to figure out what their own company is actually worth.
The numbers everyone’s looking at
The AI valuation headlines are real. They’re also extreme outliers.
OpenAI was valued at $852 billion in late March after closing a $122 billion funding round. Anthropic raised $30 billion at a $380 billion post-money in February 2026, and as of May was in talks to raise as much as $50 billion at a $900 billion valuation. Anthropic’s annualized revenue is reported to be approaching $45 billion. Cursor (Anysphere) doubled from $29.3 billion to a reported $50 billion in about five months, on $2 billion in ARR.
Strip out the headline numbers and look at what’s actually being paid:
| Company | Reported Valuation | Reported ARR | Revenue Multiple |
|---|---|---|---|
| OpenAI | ~$852B | ~$13B | ~65x |
| Anthropic | ~$380B–$900B | ~$45B | ~8x–20x |
| Cursor (Anysphere) | ~$50B | ~$2B | ~25x |
For the rest of the venture-backed software market, the comparable multiples are dramatically lower. Public SaaS companies in 2026 trade at roughly 6–10x forward revenue. Private Series B SaaS rounds are pricing at roughly 10–20x trailing ARR, with the highest growth and highest-quality companies at the top of that range. That’s the actual benchmark you should be using, not the AI mega-deals.
The gap between “AI infrastructure / frontier model / category-defining tooling” and “everything else” is the largest divergence the venture market has seen in a decade.
Why these AI valuations don’t apply to you
Three reasons, in order of importance.
1. They’re growing at rates that have no precedent
Cursor crossed $100 million ARR in January 2025. It hit $500 million by June. It cleared $1 billion in November. By February 2026, that figure had doubled again to $2 billion. Slack took five years to reach its first billion. Zoom took nine. Snowflake took six.
This is the part that gets lost in the multiple conversation. A 25x multiple on a company growing 200% year-over-year prices in something fundamentally different from a 25x multiple on a company growing 50%. Investors aren’t paying for current revenue. They’re paying for what the revenue becomes in 18 months if the growth holds.
If you’re growing 80–100% year-over-year at the Series B stage — strong by any historical standard — you’re growing one-third to one-half as fast as these AI companies in percentage terms, and a tiny fraction of their dollar growth. The math doesn’t transfer.
2. They have structural advantages no SaaS company has
The frontier AI labs are pricing in three things that almost no other startup has:
- Massive AI training and deployment costs that act as a moat. OpenAI and Anthropic spend more on training a single model than most Series B companies will raise in their lifetimes. That cost is also their barrier — if the cost of building a frontier model drops to $100M, the moat is gone. As of 2026, it hasn’t.
- Distribution as the product. OpenAI’s API and Anthropic’s API are infrastructure layers underneath every other AI startup in the market. They get a cut of the value created by the entire downstream ecosystem.
- Optionality on the value of the underlying technology. Investors paying $852 billion for OpenAI aren’t really pricing in $13 billion of ARR. They’re pricing in the possibility that AGI gets built and that OpenAI owns some share of the most valuable technology in human history. That’s a binary bet structure that doesn’t apply to vertical SaaS, dev tools, or B2B software.
A Series B SaaS company has none of these. Comparing valuations is like comparing the price of a lottery ticket to the price of a savings bond.
3. The capital pool funding these deals is different
Singapore’s sovereign wealth fund GIC and U.S. investment firm Coatue are leading Anthropic’s funding round. Sequoia Capital is expected to make a large commitment. These are not the funds that lead Series A and B rounds for normal software companies. They’re sovereign wealth funds, crossover hedge funds, and growth equity firms writing $5–10 billion checks looking for AGI-scale outcomes. They’re price-insensitive in a way Series A/B funds aren’t.
When you talk to a Series A or B VC, you’re talking to someone who needs to return their fund within 7–10 years and is pricing your company against 50 other companies they could fund instead. That’s an entirely different buyer.
What your valuation actually depends on
If you take the AI mega-deals off the table, the math for normal venture-backed startups in 2026 looks like this:
Growth rate is the single largest multiplier. A SaaS company growing 100% year-over-year at $10M ARR will price at roughly 20–30x. The same company at $10M ARR growing 30% prices at 6–10x. That’s a 3–5x difference in valuation driven by one variable.
Net dollar retention is the second. Best-in-class SaaS sits at 120%+ NDR. At 110–120%, you’re average. Under 100%, you’re discounted aggressively. The market in 2026 has gotten ruthless about retention — investors learned hard lessons in 2022–2023 about the difference between churning revenue and durable revenue.
Gross margin is the third. 75%+ gross margin SaaS gets the full multiple. 50–60% (typical for AI-heavy products with high inference costs, or for services-heavy companies) gets a haircut. Many AI startups in the application layer have margin problems they’re not yet disclosing — worth understanding if your competitors include AI-native products.
Burn multiple is the fourth. Net new ARR divided by net burn. Best-in-class is under 1.5x. Over 2x at Series B raises questions. The market wants growth AND efficiency, not growth at any cost. The 2021 era when burn didn’t matter is gone.
For a Series B SaaS company in 2026, a typical valuation range is:
| Growth + Efficiency Profile | Multiple Range |
|---|---|
| 100%+ YoY, 120%+ NDR, 75%+ GM, <1.5x burn multiple | 20–30x ARR |
| 60–100% YoY, 110–120% NDR, 70%+ GM | 12–20x ARR |
| 40–60% YoY, 100–110% NDR | 6–12x ARR |
| Under 40% YoY, sub-100% NDR | 3–6x ARR or harder to raise |
The Series B founder mentioned at the top of this post — pricing at 12x — was right in the middle of that “60–100% YoY, healthy retention” band. The price was correct for their profile. The frustration was about benchmarking against a market that doesn’t apply to them.
What founders actually get wrong
A few patterns we see consistently from operators trying to make sense of the current market:
Mistake 1: Treating “AI” as a magic premium. Adding AI features to your product doesn’t get you AI valuation multiples. Investors learned to distinguish “AI-native infrastructure” from “SaaS company with AI features” sometime in 2024. If your gross margin is being eaten by inference costs and your AI differentiation is the same as everyone else’s, the AI label is a liability, not an asset, in valuation conversations.
Mistake 2: Using the wrong comp set. Comparing yourself to OpenAI and Anthropic is meaningless. Comparing yourself to the 8–10 venture-backed Series B companies in your category that raised in the last 12 months is meaningful. Make sure you know who those comps actually are and what they actually priced at — published headline numbers often exclude the structuring (liquidation preferences, ratchets, warrant coverage) that materially affects the real deal.
Mistake 3: Anchoring on your last valuation. A round you raised at a $200M valuation in 2021 doesn’t entitle you to a $400M valuation in 2026. The market changed. The right benchmark is what the next round would actually price at today, based on your current metrics and current comparable transactions.
Mistake 4: Assuming the AI valuation environment will trickle down. It might. It also might be a top-of-cycle phenomenon that compresses sharply once the public markets recalibrate. Most fractional CFOs and Series B investors we talk to are pricing this risk in — they’re not extrapolating mega-deal multiples into the broader market.
Mistake 5: Conflating valuation with success. A Series B raised at 20x ARR with a 1.5x liquidation preference and 25% dilution can be a worse outcome than a Series B raised at 10x ARR with clean terms and 18% dilution. The number on the front page of the term sheet is not the deal.
What to do instead
If you’re a founder evaluating where your company actually stands, the work is straightforward — but it requires honesty about your own metrics.
-
Build your real comp set. Find 5–10 venture-backed companies in your stage and category that raised in the last 12–18 months. Get their reported valuations (TechCrunch, The Information, Crunchbase, PitchBook) and their reported ARR if disclosed. Their multiples — not the AI mega-deals — are your benchmark.
-
Stress-test your metrics against the four levers. Growth, NDR, gross margin, burn multiple. If you’re best-in-class on all four, you should expect the high end of the range for your category. If you’re average or below on any, expect to be priced accordingly.
-
Sequence your fundraise around metric strength, not vibes. If your NDR is the strongest part of your story, raise when you have a clean expansion cohort to show. If your growth rate is the story, raise after a strong quarter. Don’t raise into a weak month thinking the market will give you credit.
-
Negotiate terms, not just valuation. The headline number is the smallest part of a Series B deal in 2026. Liquidation preferences, board composition, anti-dilution, and option pool refresh all materially change the economic outcome. A higher headline valuation with worse terms is often a worse deal.
-
Get a real 409A done. If you’re hiring senior people in 2026, the 409A valuation that sets your strike price matters enormously for recruiting. We cover this in detail in our equity compensation guide — but the short version is: a poorly-priced 409A creates downstream tax problems that outlive the round.
What this looks like in practice
A Series B SaaS company we worked with this year had been told by a well-known fund that they should be raising at $300M based on AI comps. Their actual metrics: $11M ARR, 85% YoY growth, 118% NDR, 76% gross margin, 1.7x burn multiple. Strong by historical standards, but not best-in-class.
After modeling out the realistic range — 15–18x ARR given their profile — we landed at a $180M–$200M target. They raised at $190M with clean terms and a strategic lead they wanted. The fund that pitched $300M never produced a term sheet at that number.
The lesson: the gap between “the headline price the market is paying for AI infrastructure” and “the price your specific company can actually clear” is enormous in 2026. Founders who anchor on the wrong reference point lose months chasing rounds that don’t materialize.
When to bring in operator support
Valuation is where the gap between what founders read about and what actually closes is widest. A fractional CFO or experienced finance partner adds value in three specific places:
- Building your real comp set — finding the right reference transactions for your stage, category, and metric profile, not the ones that get the most headlines.
- Modeling the realistic valuation range before you go to market, so you know whether $200M is achievable or fantasy.
- Negotiating the structure of the deal, not just the headline number — the liquidation preferences, option pool refresh, and board terms that determine whether the round is actually a good outcome.
You probably don’t need outside help if you’ve raised before, have a strong sense of your comp set, and have a clean view of your own metrics. You likely do want operator support if you’re raising your first Series B, you’re confused about the current market, or you’ve been told a number that feels too good to be true and want a second opinion.
Pegacorn Group works with venture-backed Series A and B startups on exactly these problems. If you’re trying to figure out what your company is actually worth in the current market — not what AI infrastructure is worth, what you are worth — let’s talk.
This post pairs with: Equity compensation 101, The Series A finance stack, and The Series A/B benefits stack.