Learn from AI startups that shut down
TinySet builds a dataset of AI startups that closed in 2026. TinyResearch finds the patterns across them.
TinySet
Dataset · 29 rowsAI startups that shut down in 2026, with what they built and funding raised
| Website URL | Name | What they built | Funding raised (USD) |
|---|---|---|---|
| https://www.phind.com/ | Phind | An AI-powered search engine built specifically for developers, using multi-step reasoning to answer coding questions and cite sources from Stack Overflow, GitHub issues, and technical documentation. | 11000000 |
| Cartwheel Robotics | Yogi, a compact bipedal toddler-like social humanoid robot for companionship and public spaces (plus Speedy variant) | 3000000 | |
| https://www.soulmachines.com/ | Soul Machines | AI 'Digital People' avatars — human-like AI assistants for customer service, powered by a Digital Brain | 135000000 |
| https://www.kintsugihealth.com/ | Kintsugi | Voice biomarker technology to detect depression and anxiety from how people speak | 30000000 |
| Cydoc | A HIPAA-compliant AI-native EHR for web, Android, and iOS, featuring a Smart Patient Intake Form that automatically takes a patient's history and generates a clinical note for the doctor. | 265000 | |
| https://vercept.com | Vercept | Vy, a cloud computer-use AI agent (built on the VyUI interface-recognition model) that operates apps and controls a remote MacBook via screenshots, without APIs | 50000000 |
| https://www.yupp.ai/ | Yupp | A free crowdsourced AI model-comparison playground that let users test and compare responses from 800+ AI models side by side, vote on the best answers, and earn cryptocurrency for their feedback. | 33000000 |
| https://weights.gg | Weights | Community voice-cloning and AI-music platform (weights.gg) | 4000000 |
TinyResearch
ReportWhy are AI startups shutting down? Look for patterns across these companies.
AI startups are failing at an accelerating rate, but the causes are more structural and systemic than the popular narrative of "too much hype" suggests. Across hundreds of documented shutdowns from 2023–2026, six dominant failure patterns emerge: (1) catastrophic unit economics driven by compute costs, (2) product cannibalization by Big Tech foundation model providers, (3) the "wrapper trap" — thin differentiation on top of third-party models, (4) talent extraction through reverse acquihires that hollow out startups, (5) litigation and regulatory exposure concentrated in consumer-facing AI, and (6) a global funding concentration that leaves non-U.S. startups starved of capital. These patterns are not independent — they interact and accelerate each other. The canonical case study is Character.AI, which filed for Chapter 11 bankruptcy on October 12, 2026, illustrating nearly every failure mode simultaneously. Decision-makers evaluating AI startup investments, partnerships, or competitive strategy should treat each of these patterns as a live structural risk, not a historical artifact.
