"The best time to notice a company is before everyone else does. That's what a daily signal is for."
Every AI startup claims momentum. The AI250 measures it, scoring every company on six dimensions (revenue, product depth, technical moat, talent, customer adoption, and investment) and refreshing the scores every night. Here's who's standing out among private companies that haven't yet reached late-stage rounds (Series D or later in our funding records), as of October 19, 2026. DeepSeek (#4) would top this list, but it's funded by its parent hedge fund rather than venture capital, so we've left it out.
The Leaders
Together AI (#8, 71.4). The highest-ranked venture-stage infrastructure company in the index. Its cloud for training and serving open-weight models sits at the center of the open-model boom, and it scores 9.0 or higher on both Product Depth and Talent & Research.
Moonshot AI (#15, 69.1). The Kimi lab raised $3.5 billion in Q3 and scores 9.3 on Talent & Research, one of the highest marks in the index.
Perplexity AI (#17, 69.1). AI-native search with consumer pull that most incumbents envy.
Krutrim (#19, 67.4). India's first AI unicorn, building models and infrastructure for Indian languages. It's the clearest sign that the breakout list is no longer just a Silicon Valley list.
SambaNova (#18, 67.7). A full-stack AI chip and inference company, scoring 8.6 on both Data & Model Moat and Talent & Research.
Fireworks AI (#24, 66.0). One of the fastest inference platforms for open models, scoring 8.7 on Product Depth.
SandboxAQ (#27) and Lightning AI (#28). Two very different infrastructure plays (AI-driven quantitative simulation, and a developer platform for building and training models). SandboxAQ scores 8.8 on Data & Model Moat; Lightning AI scores 9.4 on Talent & Research.
Names to Watch Next
Sakana AI (#33). Tokyo's research-first lab, and one of the strongest research signals outside the US, at 9.3 on Talent & Research.
ElevenLabs (#38). Voice AI has gone from novelty to infrastructure for customer service, media, and agents.
Physical Intelligence (#40) is building foundation models for robots, one of the most ambitious bets in physical AI. Sarvam AI (#47) is another strong Indian foundation model player. Qdrant (#53) from Germany is a leading open-source vector database, and Applied Intuition (#56) supplies simulation and autonomy software to automakers and defense.
How We Built This List
We filtered the AI250 to private, venture-backed companies below Series D in our funding records and ranked them by Bluescore. Bluescore measures publicly observable signals, and revenue for private companies is estimated from public reporting rather than read from financials, so treat this list as a map of momentum, not a verdict on business quality. Each company profile shows its six dimension scores, so you can see exactly what's driving the number.
Methodology v4.0 took effect on October 18, 2026, so we've left out rank changes for now. We'll add the biggest climbers back once a full week of v4.0 history is in place.
Breakouts move fast. Follow the AI250 daily to catch the next one.