On sovereign hardware, the people of Europe, and an open invitation
to build the edge the world actually needs
In 2013, it was revealed that Angela Merkel's phone had been tapped by the NSA. The leader of Europe's largest economy. An ally. The response was diplomatic embarrassment, some strongly worded statements, and then everyone carried on using the same infrastructure.
Because there was no alternative. The hardware, the network layer, the operating system — all of it ran through systems where the answer to "who can access this?" was ultimately not European.
Europe has since built the world's strongest data protection framework. GDPR. The AI Act. NIS2. The Data Act. A comprehensive legal architecture for digital sovereignty. And then left it sitting on top of hardware it does not control, running on chips manufactured in facilities where the largest single shareholder is the United States Treasury.
The US government converted its CHIPS Act grant to Intel into an $8.9 billion equity position. Intel is now, structurally and literally, a US government asset. The Magdeburg fab was not delayed because the engineering was hard. It was delayed because Intel's largest shareholder has a clear preference for where advanced node capacity gets built — and it is Ohio, not Saxony-Anhalt.
Europe cannot build its sovereign hardware strategy around a company that is a US government equity position. This is not a political statement. It is a structural observation. The ownership makes it true before any policy is written.
In Veldhoven, the Netherlands, there is a factory that makes the most precise machines ever built by human hands. ASML's EUV lithography systems are the only tools on earth capable of manufacturing advanced semiconductors at sub-5nm nodes. One factory. One company. One small Dutch city. The entire global semiconductor industry depends on what they build there.
They were laughed at. Canon and Nikon had the market. ASML was a tiny joint venture nobody took seriously. They made a bet on extreme ultraviolet light — wavelengths so short they are absorbed by air, requiring mirrors polished to atomic precision, in a vacuum, with plasma generated by firing lasers at tin droplets 50,000 times per second. The physics alone took decades.
They believed. When every contract went to Canon, when the funding was tight, when the physics kept throwing up new impossibilities — they believed. That belief, that specific flavour of Dutch engineering stubbornness, is now the single most strategically important industrial capability on earth.
ASML's facility has down cycles. Test systems. Calibration infrastructure. Engineering workstations. The entire apparatus of a world-class precision facility between production runs. This is the most strategically located brownfield compute resource in the world — and it sits largely idle between builds.
Europe's semiconductor heritage is not a corporate asset. It is a human one. The engineers at ASML, at imec in Leuven, at NXP in Eindhoven, at Infineon in Munich — they carry knowledge that took decades to accumulate and that no acquisition can transfer. This is Europe's actual strategic asset. It is time to use it.
Jensen — we are not coming for your data centre. We are not coming for your training clusters, your H100 market, your Blackwell architecture, your hyperscaler relationships. We have no interest in that competition and no capability to win it even if we did.
We are going somewhere you have chosen not to go. The families. The communities. The people who will never buy an H100 but who deserve frontier intelligence in their homes. The gap between a 5060 Ti with 16GB and a device that runs a 70B model at full precision — that gap exists because your business model requires it, not because the engineering does.
LPDDR6 is coming. The memory capacity constraint on consumer cards is temporary. Someone will build the 96GB edge card. It will run every model a family needs. It will be sovereign, local, owned outright. The question is whether Nvidia's name is on it.
We propose: put your name on the card. Take the margin. We handle the manufacturing through European facilities that already have the engineering capability and the infrastructure. You get the royalties, the partnership, a hundred million new customers who will upgrade to the next generation. We get the sovereign edge network Europe needs.
Or don't come. We will go anyway. But the first conversation is the olive branch — genuine, without hidden agenda. Because Nvidia winning in the data centre and Europe winning at the edge are not in conflict. They are the right division of labour.
Nvidia's entire revenue depends on a handful of hyperscaler customers. If Amazon, Microsoft, or Google fully commits to custom silicon — which all three are actively developing — Nvidia feels it immediately. A hundred million distributed family nodes is a fundamentally different customer base. Distributed. Resilient. No single relationship that can be lost overnight.
The olive branch is also, quietly, pointing at their concentration risk. Without saying so out loud.
The olive branch is genuine. The partnership proposition is real. But the direction of travel is equally clear, and it does not end with Nvidia.
Open source silicon — RISC-V — is already here. The instruction set is free. The toolchain is maturing. What does not yet exist is an edge inference chip designed from first principles for the workload a family node actually runs: not training, not rendering, not general compute — inference, sovereign, local, owned. Optimised for 70B parameter models at 96GB. Efficient enough to run on solar. Cheap enough for every household.
No corporate roadmap constrains what it can do. No licensing overhead inflates the cost. No artificial memory segmentation preserves enterprise margins. The chip exists to serve the person, not the business model.
Path A funds Path B. Path A proves the market that Path B will eventually own. They are not in competition — they are in sequence. The engineers in Veldhoven's down cycles are already working on the question, even if they do not yet know what they are building. That is how it should be. That is how ASML began.
ASML did not become ASML because a government programme decided Europe needed EUV capability. It became ASML because engineers who understood physics deeply were given space, time, and the freedom to attempt something everyone else thought was impossible.
The proposition is simple: open the down cycles. Give the next generation of European engineers access to the world's most sophisticated hardware facility between production runs. Not a structured programme. Not a curriculum. Not a grant application. Access and freedom.
What do they build? We genuinely do not know. That is the point. You create the conditions — the knowledge, the infrastructure, the proximity to people who have already done the impossible — and you find out. The Spectrum was not designed from a policy paper. ARM was not born from a government grant. They emerged from people with access and freedom and a constraint that forced them to think differently.
The constraint here is sovereignty. The freedom is the down cycles. The knowledge is already in the room.
The engineers who built EUV are still there. The knowledge that took decades to accumulate lives in the people of Veldhoven. Put young engineers in the same room and let the transmission happen. Not through lectures. Through proximity, access, and the cultural inheritance of a community that believed when no one else did.
This document is open. It can be copied, forwarded, forked, disagreed with, built upon. It carries no corporate ownership, no government imprimatur, no institutional authority. It is written by a network engineer from Aberdeenshire who commutes past subsea cables and offshore wind every morning and sees the infrastructure of sovereignty already in place, waiting to be connected.
The proposition is not: Europe should build a European Intel. Not: Europe should compete with TSMC. Not: Europe should match US hyperscale spend.
The proposition is: Europe should build for its people. Starting from the household. Starting from the family that deserves frontier intelligence without surrendering their data to a foreign sovereign. Starting from the engineers in Veldhoven who already proved that the impossible is just a matter of belief and time.
The Spectrum asked: what if every family owned a computer? ARM asked: what if power matters more than speed? Nokia asked: what if the phone is the network?
We ask: what if the intelligence lives with the person?
Build that. Everything else will follow.