A European company is raising over $1 billion before even launching its product. Recently, the global startup funding cycle was quietly dominated by one European AI company, starting with a $1 billion seed round without a finished product.
That company is Ineffable Intelligence, now raising a $1 billion seed at a $4 billion pre-money valuation—the continent’s largest seed round ever. Founded by David Silver, the former lead scientist at Google DeepMind, the lab is explicitly rejecting the LLM scaling path.
For anyone outside the research bubble, the obvious question is: why would investors fund $1 billion into a company with no product? Silver believes today’s transformer playbook has structural limits.
Current frontier models predict the next word by training on massive amounts of internet text. Scale the data, scale the parameters, scale the compute—and performance improves. But high-quality human data is finite, predicting text does not equal understanding the world, and these systems remain passive observers.
Ineffable’s answer is reinforcement learning at scale. Instead of static datasets, agents live inside environments—games, simulations, eventually the real world. They act, receive rewards or penalties, and optimise policies over millions of iterations.
Through self-play and autonomous experimentation, the system discovers strategies humans never found. Intelligence is grown through experience rather than compressed from the internet.
Investors are betting that active learning systems capable of planning, adapting, and generalising across environments can surpass static LLMs. If that happens, the dominant AI training paradigm changes.
This $1 billion seed isn’t just backing a company. It is funding a different theory of how superintelligence is built—one where agents learn by doing, not just predicting.