Apex Raises $50M for Self-Improving AI
Beijing-based Apex Intelligence has raised nearly $50 million to develop foundation models focused on recursive self-improvement and scientific discovery.
Most AI systems still improve through large amounts of human-generated data and feedback. Apex Intelligence has raised nearly $50 million around a more ambitious research direction: models designed to improve their own scientific and reasoning processes.
What happened
The Beijing-based startup raised the capital across angel and angel-plus rounds.
IDG Capital, LinkX Capital and XtalPi co-led the first round, while Zhongguancun Science City Fund, SCGC and Shanghai Engine Fund co-led the follow-on financing.
Founded only in June, Apex is developing foundation models focused on recursive self-improvement and scientific discovery.
Why it matters
Self-improving systems are one of the most important unresolved directions in frontier AI. If models can generate hypotheses, test them and learn from the results with less human supervision, progress in areas such as materials, biology and mathematics could accelerate significantly.
The same capability also raises difficult questions about evaluation and control.
The bigger picture
China's AI ecosystem is producing a growing number of well-funded labs pursuing alternatives to the dominant US frontier-model companies.
Apex's early capital shows investors are willing to fund high-risk research before a conventional product exists. Whether that translates into commercially useful systems will depend on demonstrating measurable improvement rather than simply using self-improvement as a research narrative.