EU AI chip startups raise funds to rival Nvidia chips

European AI chip startups are seeking larger funding rounds as investors look for alternatives to Nvidia GPUs for AI inference.

Dutch startup Euclyd is in talks to raise at least 100 million euros (US$118 million) to scale its technology and reach customers.

Euclyd, founded in 2024 by former ASML director Bernardo Kastrup, has raised less than 10 million euros (US$11.8 million).

The company is building a multi-chiplet inference system, though its chips have not yet been proven in commercial deployment at scale.

Other startups including Optalysys, Fractile, and Arago are also raising funds.

European AI chip startups have raised about US$800 million so far in 2026, compared with US$4.7 billion for US peers, according to Dealroom.

🔗 Source: CNBC

🧠 Food for thought

Implications, context, and why it matters.

Startups such as Fractile are building in-memory computing, which processes data inside memory to cut delays and power use. This architecture made up 22.5% of the generative AI chip market in 2025 1.
Investor interest includes former Intel chief executive Pat Gelsinger, who said in January 2025 that he had invested in Fractile 2.
Many teams are targeting AI inference, the step where trained models produce responses, instead of AI training. Inference led the generative AI chip market in 2025 at about US$1.39 billion, which leaves room for niches outside Nvidia’s training chips 1.
Public funding also plays a role, including the 43 billion euros (US$50.7 billion) European Chips Act, a European Union plan to boost semiconductor manufacturing. It aims to raise the region’s global semiconductor market share to 20% by 2030 3.

Funding helps, yet many European chip startups still face Nvidia’s CUDA platform, a widely used programming system for AI chips with more than 4 million developers 1.
Rivals can post strong hardware results, such as Groq’s LPU, its chip architecture for running AI models quickly, at more than 500 tokens per second in benchmarks. Customer adoption can stall when the software stack is new 1.
Cloud hyperscalers, meaning the largest cloud computing providers, accounted for 42.3% of the generative AI chip market in 2025. They have often been slow to add fragmented hardware into standardized infrastructure 1.
Some startups may end up as acquisition targets for larger companies hunting new chip designs, as happened with UK-based Graphcore, acquired for just over US$600 million in 2024 4.