Seedcamp, one of the first investors to bet on Revolut, Wise, and UiPath when they were barely more than an idea, has announced a new $320 million fund. Its stated goal: to build a "transatlantic bridge" connecting the European and US startup ecosystems. The news goes beyond financial headlines. At a time when infrastructure for Large Language Models has become a geopolitical battleground, venture capital flows are drawing the map of technological power for the years ahead.

A transatlantic bridge with AI in its sights

Seedcamp’s portfolio already includes names such as Synthesia, an AI-driven video generation platform, and Fluidstack, which provides GPU cloud access for machine learning workloads. The new fund does not declare an exclusive focus on artificial intelligence, but the presence of these companies in its track record suggests that a significant share of the capital may flow toward startups operating at the intersection of models, hardware, and orchestration platforms. The transatlantic bridge, in this light, is more than a commercial metaphor: it can become the channel through which know-how, semiconductors, and software architectures move between two continents competing for digital sovereignty.

Fluidstack and the GPU cloud conundrum

Fluidstack deserves particular scrutiny. The startup, backed also by other top-tier funds, supplies on-demand GPU clusters but has recently explored hybrid solutions that allow data to remain within defined geographical boundaries. In a landscape where large enterprises weigh the Total Cost of Ownership of on-premise deployment against the cloud, such platforms can reduce technical and regulatory friction. If Seedcamp continues to back similar ventures, we may see new tools emerge that simplify local hosting of models – from quantization to VRAM management – reducing dependence on US Big Tech.

Digital sovereignty: the role of patient capital

The fund’s timing coincides with the rollout of regulations like the EU AI Act and a growing focus on GDPR compliance. Organizations handling sensitive data – healthcare, finance, public administration – are increasingly looking for air-gapped or self-hosted alternatives to avoid egress costs and compliance risks. In this scenario, venture capital acts as an accelerator: startups funded today could tomorrow deliver the tools needed to run LLM inference on local hardware with the same agility as the cloud, bridging the gap that still holds many CIOs back. AI-RADAR, in its mapping of on-premise stacks, has repeatedly noted how the availability of fine-tuning pipelines optimized for consumer hardware is already shifting cost-benefit assessments.

What to watch next

The fund’s size matters less than the destination of its first checks. If Seedcamp directs capital toward startups developing frameworks for model serving on bare metal infrastructure or networking technologies optimized for AI workloads, the transatlantic bridge could strengthen European technological autonomy. Conversely, investments focused purely on SaaS-only models would leave the power balance with the major cloud providers unchanged. For anyone designing deployment architectures today, tracking these signals is not an academic exercise: venture capital allocation choices anticipate the market’s direction and, ultimately, the concrete options available to companies that want to retain full control of their data.