By Pranil Arsid, Analyst
AI has crossed from novelty to norm: assistants produced 41% of code in 2024, about 256 billion lines. Drawing on GitHub’s billion-repository and public Q&A, the repetitive, well-commented data allows models to logically suggest additional lines of code; Copilot trials shows tasks are finished 55% faster. Open-source entrants like DeepSeek intensify price pressure.
Competition clusters into four camps. GitHub Copilot anchors Microsoft with 1.3 million paying developers and 50,000 enterprise seats. Google’s Gemini 2.5 Pro tops SWE-Bench at 63.8%, beating GPT-4.5’s 38%. Open-source DeepSeek-Coder V2 nears GPT-4-Turbo parity, while IDE-native Cursor’s $900 million Series C set a $9.9 billion valuation and reach $300M ARR within 2 years.
Trajectory appears exponential. Anthropic CEO Dario Amodei predicts AI could write 90% of code within a year. Gartner expects 80% of enterprises will have utilized GenAI APIs or models by 2026. As open-source models keep closing gaps, advantage hinges on secure integration, workflow coverage, cost control, talent re-skilling, governance, and developer oversight.
Stepmark Partners 