In April 2026, Anysphere — the company behind Cursor — entered talks to raise a fresh $2B at a $50B pre-money, co-led by Andreessen Horowitz and Thrive Capital with Nvidia tagging in. That deal effectively doubled the $29.3B mark Cursor printed at Series D five months earlier, and reset the ceiling for everyone else in this segment. The 144 startups we cover here have absorbed $76.1B between them, and the gap between the leader and the long tail keeps widening.
This category is for software where a human engineer is the operator — IDE assistants, prompt-to-app builders, terminal agents, code-search layers, and the proprietary code labs racing to beat general-purpose LLMs on engineering work. Hosted runtimes belong in AI Developer Platforms, GPU layers belong in AI Infrastructure, and computer-use agents belong in AI Agents. The boundary holds because the buyer is different.
Where the IDE wars stand
Cursor crossed $1B ARR by late 2025 and is reportedly tracking $2B by year-end 2026 — the fastest B2B SaaS ramp on record. Cognition acquired Windsurf in July 2025, repriced to $10.2B by September, and is now in talks at $25B; Devin and the Windsurf editor now sit under one roof. Codeium's $2.97B in cumulative funding got partially absorbed in that deal. Replit hit a $9B Series F on the back of Replit Agent. JetBrains AI Assistant, Tabnine ($55M, $1.5B valuation), Sourcegraph ($228M Series D), and Augment Code ($252M Series B at $977M) round out the assisted-IDE bench.
The hyperscaler squeeze
GitHub Copilot, Gemini Code Assist, Amazon Q Developer, and Amazon's Kiro all ride existing seats — the floor for paid AI coding is now zero dollars if you already have a Microsoft, Google, or AWS contract. That bundling is what makes Cursor's premium-tier ARR thesis so unusual. The other independent bet is vertical: Magic raising on a code-specific foundation model, Poolside at $500M Series B targeting enterprise SWE workloads, Lovable's $6.6B implied valuation on a seed round for prompt-to-app. Bolt.new (StackBlitz) and v0 (Vercel) anchor the prompt-to-app surface. The long tail — Sweep, Phind, Bloop, Zed, Supermaven (acquired by Cursor) — fights for niches the incumbents do not bother defending.