French cloud computing firm Scaleway has announced plans to build an AI supercomputer in Sweden, its first such system outside France. The machine will be powered by AMD's upcoming Instinct MI355X chips, and the company is positioning the project as a European alternative to the software ecosystem built around Nvidia's hardware.
The announcement comes as demand for AI computing capacity continues to surge, and as European businesses and governments look for ways to reduce reliance on a handful of dominant technology providers.
Why the software layer matters
In the world of artificial intelligence, the physical chips get most of the attention, but the software that runs on them can be just as important. Nvidia's CUDA platform has become the default environment for AI developers over the past decade. It is a set of tools and libraries that make it easier to write and run AI programs on Nvidia graphics processing units (GPUs).
Once developers build their applications on CUDA, moving that code to another type of chip can be difficult and costly. That creates what is known as a "switching cost" — a barrier that keeps customers locked into Nvidia even when other hardware might be cheaper or more available.
Scaleway is betting that AMD's open-source ROCm platform can lower that barrier. ROCm supports many of the same popular AI frameworks that developers already use, such as PyTorch and TensorFlow. If it works as advertised, it could make it easier for companies to mix and match Nvidia and AMD hardware, rather than being tied to a single vendor.
A European angle
Scaleway says it expects to offer access to the new supercomputer to startups and enterprises across Europe. The company also highlighted Sweden's relatively low-carbon electricity supply and its growing local AI scene. Scaleway entered the Swedish market through a partnership with telecom operator Tele2.
The broader backdrop is intensifying competition in the AI chip market. Nvidia still powers the vast majority of AI computing worldwide, but AMD has been gaining credibility. Its chips are already used in major research systems, including the LUMI supercomputer in Finland, one of the most powerful in Europe. US cloud providers have also begun renting AMD-based machines to European customers.
For European companies, having more options for AI compute could be significant. Many have expressed concerns about relying on US-based providers for critical infrastructure, especially as geopolitical tensions and export controls reshape the technology landscape.
What it means for investors
For investors, the real story here may not be the specific specifications of AMD's MI355X chip, but rather the role of ROCm in challenging Nvidia's dominance.
If ROCm makes it simpler to run the same AI workloads across different chips, cloud customers gain more credible "multi-sourcing" options. Instead of being locked into a CUDA-only stack, they could choose to rent capacity from multiple providers based on price, availability, or other factors.
That could, over time, weaken Nvidia's pricing power on scarce GPU capacity. Nvidia has been able to command premium prices for its chips because demand far outstrips supply, and because customers have few alternatives. If AMD and others can offer viable substitutes, cloud providers may be forced to compete more aggressively on price and contract terms.
The beneficiaries would likely be AI startups and large companies that rent computing power in Europe. They could see lower costs and more flexibility. Providers like Scaleway, which can differentiate on flexibility rather than sheer scale, could also gain a foothold in a market dominated by giants like Amazon Web Services, Microsoft Azure, and Google Cloud.
That said, Nvidia is not standing still. The company continues to release new chips and expand its software offerings, and its CUDA ecosystem remains deeply entrenched. Breaking that hold will not happen overnight.
For now, Scaleway's move is a sign that the competitive landscape for AI infrastructure is shifting. As more players enter the field and as software tools improve, the balance of power in the AI chip market could gradually change. Investors should watch how quickly ROCm adoption grows and whether major cloud providers begin to offer AMD-based options more widely.
In the meantime, the announcement adds to a growing list of developments in the European tech sector, which has been under pressure from slower economic growth and rising competition from the US and Asia. It also comes as companies across the region restructure around AI, and as investors weigh the implications of corporate write-downs and other headwinds.
For everyday investors, the key takeaway is that the AI boom is not just about one company or one technology. The race to build the infrastructure that powers AI is creating opportunities and risks across the tech sector, and the choices companies make today could shape the market for years to come.
