H Company: Why Investors Gave 180 Million Dollars to a Startup Without a Product

Imagine this situation: you come to investors, you have no ready product, no massive cluster of thousands of GPUs, and not even a website with fancy charts. Yet they still write you a check for $180 million. It sounds like delusion from the dotcom bubble era, if not for one "but."

Behind this project stand people who literally created modern AI within the walls of DeepMind. Startup H Company, formerly known as Holistic AI, became a new sensation in the industry not because it promises to make GPT-5 faster, but because it questions the very approach to creating intelligence. The industry has become accustomed to "Scaling Laws."

The logic is simple: give the model more texts, build a bigger cluster, and it will become smarter. But the founders of H Company — Charles Le Lain, Laurent Sifre, Karl Tules, and Julien Perolat — decided that this path leads to a dead end. They're pushing the concept of "Small Data."

The idea is that the human brain doesn't need to read the entire internet to learn logic. We learn from examples, understand cause-and-effect relationships, and can act under uncertainty. This is exactly what modern LLMs lack, which remain merely very advanced statistical parrots.

Andrey Karpathy, one of OpenAI's founders and former head of AI at Tesla, is lavish in his praise for the new team. In an industry where every second person tries to resell an API from Microsoft, H Company's approach looks like an attempt to return to real science. Karpathy notes that the focus on Multi-agent systems may become that very "silver key" to strong artificial intelligence (AGI).

Instead of one monolithic model, H Company is building a system where specialized agents interact with each other, test hypotheses, and correct errors in real time. Why is this important right now? We've hit a data ceiling.

Quality texts on the internet are running out, and training on synthetic data raises more questions than answers. If H Company proves that "Strong Intelligence" can be built on compact yet perfectly structured datasets, it will completely change the economics of the industry. Development costs will drop tenfold, and dependence on NVIDIA will cease to be fatal.

Investors, including Bernard Arnault and Eric Schmidt, are clearly betting that the future belongs to efficiency, not brute force. Of course, skeptics will remind us that overhyped expectations often lead to dramatic failures. H Company has nothing but the reputation of its founders and an ambitious manifesto.

But in a world where Mistral has already shown how a small French team can outpace tech giants from Silicon Valley, H Company deserves serious attention. They're not just building another neural network, they're trying to crack the very code of how a machine "thinks." The key point: Will "small" become the new "big"?

If H Company delivers on its vision, we face the sunset of the era of giant LLMs and a transition to compact yet truly intelligent agents.