Exclusive: Former H Company CTO seeks $30m for new startup

We have entered the era of the rapid-cycle AI pivot. Davor Ilic, the former technical lead who recently departed from the much-hyped H Company, is reportedly back on the fundraising trail. He is looking for $30 million to seed a new venture. If you feel like you just read about H Company’s $220 million seed round last month, it is because you did. The ink is barely dry on the cap table, and the technical leadership is already splintering to start fresh.

For those who missed the drama, H Company—formerly known as Holistic AI—was the darling of the French tech scene, positioned as the next big rival to OpenAI and Mistral. Then, a few weeks ago, news broke that three of the founding members, including Ilic, were leaving due to internal disagreements. Now we see the fallout: a quest for tens of millions of dollars to build another foundation model startup. This is the new normal in the AI space, and builders need to pay close attention to what this means for the market.

The Multi-Million Dollar Seed is the New Standard

In the traditional startup world, a $30 million seed round is an outlier. It is usually reserved for repeat founders with a decade of exits. In the AI world, it is the entry fee. The capital intensity of training large-scale models means that talent doesn't just ask for a runway; they ask for a war chest. Ilic is betting that his pedigree from H Company and DeepMind is enough to bypass the traditional proof-of-concept phase.

From a founder’s perspective, this trend is a double-edged sword. On one hand, it validates that there is still massive liquidity for technical talent. On the other hand, it creates a massive barrier to entry for anyone who isn't coming out of a top-tier lab. If you are a builder trying to compete in the general-purpose model space without $50 million in the bank, you are playing a game you have already lost. The signal here is clear: the market is consolidating around the 'elite' researcher class.

The Fragmentation of Foundation Labs

We are seeing a repeat of the 'PayPal Mafia' effect, but at warp speed. Usually, it takes years for a company to mature before employees spin off to start their own thing. With H Company, the split happened before they even released a flagship product. This suggests that the internal friction in these mega-funded labs is reaching a boiling point early on.

When you raise hundreds of millions before you have a product, the pressure to deliver something 'revolutionary' is suffocating. Disagreements over technical roadpoints or commercialization strategies become existential crises. Ilic’s new venture represents a bet that he can move faster or more efficiently outside of the H Company structure. For builders, the takeaway is that big funding does not equal stability. In fact, it often equals the opposite.

What This Means for the AI Stack

The quest for $30 million suggests Ilic isn't just building a wrapper; he is likely going back to the well of foundational research. But the question we have to ask is: how many foundation models does the world actually need? We already have GPT-4, Claude, Gemini, Mistral, and Llama. Each new entry into this space has to justify its existence by being either significantly cheaper, significantly more capable, or specialized in a way the giants are not.

If Ilic is chasing the same general-purpose crown, he is facing a steep uphill battle. The compute costs alone will eat through that $30 million in a matter of months. Builders who are watching this should look for the 'pick and shovel' opportunities created by this fragmentation. Every new model needs new benchmarks, new integration tools, and new security layers. The more these labs split and multiply, the more fragmented the ecosystem becomes, and the more valuable 'bridge' technology becomes.

The Investor Appetite for Risk

It is worth noting that investors are still willing to write these checks despite the high turnover in the sector. This tells us that the fear of missing out (FOMO) is still the primary driver of AI venture capital. Investors are not necessarily betting on the specific business plan; they are betting on the talent density. They want to be in the room with the people who know how to talk to the GPUs.

However, as a founder, you have to be wary of this kind of 'pedigree-only' investing. When the hype settles, the companies that survive won't be the ones with the most famous CTOs; they will be the ones that found a way to make the technology useful for people who don't care about parameters or weights. The move from H Company to a new $30 million seed play is a move for more autonomy, but it also resets the clock on actual delivery.

Takeaway for Builders

The lesson here isn't to go out and try to raise $30 million. The lesson is to watch the movement of talent. When the technical leads of the biggest labs start jumping ship to build their own versions of the same thing, it usually means the technology is becoming commoditized or the original vision was too bloated to execute.

Focus on the application layer. While the researchers fight over who gets to train the next massive transformer, the real value for most founders lies in taking these models and making them work for specific industries. Let the foundation labs burn through VC cash to build the engines; your job is to build the car that people actually want to drive.

The AI space is currently a revolving door of high-stakes bets. Don't get distracted by the dollar amounts; focus on the utility.

We will see if Ilic can close this round in a tightening market. If he does, it confirms that the AI bubble hasn't popped—it’s just fragmenting. For everyone else, it’s a reminder that in this industry, your technical reputation is your primary currency, but even that has a shelf life if you don't ship something tangible.

Read the original at Sifted →