The Race for European AI: Why GPU Control is Essential for Data Sovereignty
The ongoing discourse around AI in Europe has evolved from foundational models to a pressing necessity: access to GPUs and infrastructure that ensure data remains within European borders. Dublin-based TensorX is positioning itself at the forefront of this critical battleground by providing private AI inference solutions while maintaining full data residency.
Founded by Shane Morton, who transitioned from creating financial trading software to acquiring ICT Services, one of Ireland’s top data center operators, TensorX emerged from a prevalent market need. Morton frequently encountered requests from fintech companies eager to adopt AI but cautious about data leaving jurisdictional limits. Executive Chairman Tim Grant articulated this challenge well:
“We realised there was no viable way for many enterprises to adopt AI without a sovereign, zero-data-retention solution."
Solving AI Adoption Challenges
TensorX combines hardware and software capabilities to deliver GPU clusters focused on secure AI inference for its clients. This setup ensures that sensitive information stays within the specified legal frameworks. Grant elaborates:
“We buy the GPUs, optimise them for today's leading AI models, and sell that capability to customers who need complete control over their data.”
Supporting over 33 AI models through an OpenAI-compatible API, TensorX allows businesses to leverage generative AI without risking data privacy violations. The current client base spans across various sectors:
- Regulated enterprises in finance, healthcare, and legal fields, where compliance with regulations like GDPR shapes AI utilization;
- AI marketplaces such as OpenRouter that connect developers with sovereign GPU compute solutions;
- Software companies developing proprietary AI products leveraging TensorX's architecture, including APEX:E3 and TradeLocker.
Navigating the Bottlenecks
Grant identifies three primary hurdles in the European AI landscape: acquiring GPUs, securing financing for these GPUs, and obtaining adequate power. TensorX benefits from its ownership of ICT, a firm with over two decades of expertise in data center infrastructures and supply chains, enhancing its procurement process.
Through its strategic partnership with Dell as a Titanium Partner and its NVIDIA Inception partner status, TensorX has successfully secured an allocation of NVIDIA Blackwell GPUs, essential in a tight supply environment. Yet, obtaining hardware is only part of the equation. Grant points out the complexities involved in financing:
“That's why we're focused not only on sourcing GPUs but also on developing new financing models for AI infrastructure. Traditional infrastructure financing already exists, but AI hardware is still so new that lenders don't yet understand depreciation cycles or long-term asset values. That's an important problem we're trying to solve.”
Investing in Future Infrastructure
TensorX claims a significant edge with its own capital backing, positioning it to make rapid decisions. With an initial commitment of €8 million earmarked for NVIDIA Blackwell GPUs, the company aims for a more extensive investment of around €100 million in GPU infrastructure to solidify its status as a key player in the European market.
Power: The Next Challenge
As Europe's AI infrastructure expands, power supply is increasingly becoming a limiting factor. Grant stresses the fast-paced changes in AI infrastructure, making future predictions challenging:
“We don't yet know how supply chains will evolve or how much new capacity will come online. There is significant effort across the industry to build more sustainable infrastructure, but it's still early.”
In preparation for this growth, TensorX has locked in enough capacity for its projected development, aiming for €50-100 million in deployed infrastructure in the near future. On the horizon, they're also exploring possible partnerships or developing their own data centers to secure energy resources.
A Vision Beyond European Borders
TensorX’s ambition extends beyond Europe; they aspire to establish themselves as experts in sovereign AI infrastructure on a global scale, aiding multinational enterprises in navigating complex data residency issues. Early indicators suggest a strong market in Germany, buoyed by the country’s strict regulatory focus.
Grant highlights this unexpected demand: “We've seen a cluster of independent enquiries from German businesses looking specifically for sovereign AI infrastructure. That wasn't something we expected, but it reflects the strong focus Germany places on regulation, compliance and data sovereignty.”
Currently, TensorX is focused on deploying the GPU infrastructures it has secured, with systems already in operation and more implementations in the pipeline. Alongside its dedicated capacities in Dublin and Helsinki, they are expanding further into Germany, France, and the Nordics.
In a strategic move to bolster its infrastructure financing capabilities, the company has partnered with Solstice, a blockchain-based firm targeting the financing of tangible assets. This partnership is designed to create a facility with up to $1 billion in capacity for AI hardware financing, responding to the urgent demand for sovereign compute resources across Europe. Solstice will manage on-chain financing and launch aiUSX, a yield asset to broaden infrastructure lending to firms investing in AI.