VC
Value Add VC
⚡HomePulse⚡Helpful Apps📝Blog
← Value Add PulseFUNDING$100M run-rate

Arena, the AI Leaderboard Everyone Uses, Hits $100M Revenue Run-Rate

Arena -- the crowdsourced AI-model leaderboard born as a UC Berkeley research project and now relied on across the industry -- has reached a $100 million annualized revenue run-rate, up from $30 million in January, on the back of its commercial AI-evaluations business. The company raised a $150 million Series A in January at a $1.7 billion valuation and has turned 10 million-plus human model comparisons into a paid product for labs and enterprises.

$100M (from $30M in Jan)
Revenue Run-Rate
$150M (Jan 2026)
Series A
$1.7B
Valuation
$250M
Total Raised
10M+
Human Evaluations
TC
Trace Cohen
Early-stage VC & angel · Founder, New York Venture Partners
June 29, 2026
2 min read
KEY TAKEAWAYS FOR VCs & FOUNDERS
1

Independent evaluation is becoming critical infrastructure as models proliferate

2

A research project monetizing into a $100M business validates the eval category

3

Whoever owns the trusted benchmark shapes how every lab is perceived

4

Consumption revenue tripling in five months shows real enterprise demand

TC
The VC Read · Trace's TakeTrace Cohen

The pick-and-shovel play of the model wars: when every lab claims to be state-of-the-art, the neutral scoreboard becomes the scarce asset. Arena turned a Berkeley research project and 10 million crowd votes into a $100M business because trust doesn't scale and can't be faked overnight. The genius is distribution -- it's already the leaderboard everyone cites, so the commercial product sells itself. The structural tension worth watching: it's hard to be both the referee and a paid vendor to the teams you're scoring. If Arena keeps that neutrality, evals become a procurement standard and this is cheap at $1.7B.

💰 Funding Tracker →🤖 AI Landscape →

Arena, the crowdsourced leaderboard that has become the AI industry's default scoreboard, has reached a $100 million annualized revenue run-rate as of June 29, 2026 -- more than tripling from roughly $30 million in January -- according to TechCrunch. The growth comes from a commercial AI-evaluations service, launched in September 2025, that turns the platform's more than 10 million human model comparisons into paid benchmarking and testing for model labs and enterprises.

Arena's path is a textbook research-to-company story. It began in 2023 as a UC Berkeley project where users compared two anonymous model outputs and voted on the better one, producing the crowd-sourced rankings that labs now cite in launch announcements. It incorporated in April 2025, and its founders include CEO Anastasios Angelopoulos, CTO Wei-Lin Chiang and Databricks co-founder and Berkeley professor Ion Stoica.

“It incorporated in April 2025, and its founders include CEO Anastasios Angelopoulos, CTO Wei-Lin Chiang and Databricks co-founder and Berkeley professor Ion Stoica.”

The business thesis is that as models multiply and every lab claims state-of-the-art, independent, trusted evaluation becomes scarce and valuable infrastructure. Arena raised a $150 million Series A in January 2026 at a $1.7 billion post-money valuation -- bringing total funding to $250 million -- from a deep bench including Andreessen Horowitz, Lightspeed, Kleiner Perkins, Felicis, The House Fund, Laude Ventures and UC Investments.

The competitive landscape spans open benchmarks, internal lab evals, and a wave of AI-testing startups, but Arena's moat is mindshare: its leaderboard is the one developers and labs already point to, giving it distribution and credibility that are hard to manufacture. The consumption-based revenue model means growth tracks real usage rather than seat licenses, which cuts both ways.

The bear case is durability: benchmarks can be gamed, leaderboard prominence can fade, and a consumption model is volatile if a few large labs pull back. There is also the awkward tension of being both the neutral referee and a paid vendor to the teams it ranks. What to watch: whether Arena's evaluations become a procurement standard, how it preserves perceived neutrality, and whether the $100 million run-rate holds as the novelty of public leaderboards matures.

ShareXLinkedInEmail

Originally reported by TechCrunch. Analysis and editorial commentary by Value Add Pulse.

← Back to Pulse

Markets Now

live
S&P 500▲+1.13%
7,440.43
NASDAQ▲+1.82%
25,820.15
NVDA▼-0.39%
$194.97
MSFT▲+4.46%
$368.57
AAPL▲+2.40%
$281.74
GOOGL▲+2.89%
$353.65
META▲+3.63%
$562.60
AMZN▲+5.78%
$240.14

Read Next

FUNDING$250M

Peregrine Raises $250M at $6.8B to Power Public-Safety AI -- and World Cup Security

Peregrine Technologies, the Palantir-alumni-founded platform that unifies government data for police and public-safety agencies, raised a $250 million Series D at a $6.8 billion valuation -- nearly tripling its $2.5 billion mark from 15 months earlier. The San Francisco company serves 400-plus agencies covering 125 million people and will run security operations for eight of 11 U.S. host cities at the 2026 FIFA World Cup.

FUNDING$200M

Mirendil Raises a $200M Seed for Foundational AI in R&D, Backed by a16z, Kleiner and Nvidia

Mirendil, a stealthy startup building foundational AI systems for scientific and industrial research and development, launched with a staggering $200 million seed round co-led by Andreessen Horowitz and Kleiner Perkins, with Nvidia among its backers. The size of a seed once reserved for Series B underscores how aggressively top investors are pre-funding teams chasing AI that can accelerate discovery in chemistry, materials and the physical sciences.

FUNDING$200M

Quantifind Raises $200M to Scale AI for Financial-Crime Detection

Quantifind, which builds an AI-driven risk-intelligence platform that helps banks and government agencies detect money laundering, fraud and other financial crime, raised a $200 million growth round led by Summit Partners. The capital backs a category where regulatory pressure and the sheer volume of transactions make AI-powered screening less a luxury than a compliance necessity.

@Trace_Cohen·t@nyvp.com