This deal's real test won't play out in Metzingen or Arezzo. It will show up years from now, when someone opens a datasheet, finds a compute-on-module part number in the bill of materials for a robot limb, and has to decide whether that part will still ship, at volume, in 2030.
On that question SECO's numbers are more useful than the announcement. First-half revenue was 98.8m euros against 98.4m a year earlier [6], growth of 0.4% [1]. Edge computing did 91.4m of it, about 92.5% of the half [7][2]. The line that improved is margin, 54% from 53.4%, and it improved while memory prices were rising [9]. That is a competently run components business with a record order book [17], which is a different animal from a business holding idle capacity for a humanoid ramp.
The quarters show where the movement is. Second-quarter revenue of 50m euros beat guidance [8], which puts the first quarter at 48.8m [3]. Guidance of about 60m for the third quarter, described as up 25% and the company's best ever [8], implies roughly 48m in the same quarter last year [4] and about 158.8m across nine months [5]. Set that beside NEURA's stated intent to build millions of cognitive robots by 2030 [10]. NEURA's Smart Limbs approach spreads sensing and compute through the machine rather than centralising it in a torso, so millions of robots means tens of millions of modules [12].
Technological sovereignty and a resilient European supply chain [11] is the framing NEURA and SECO use for a contract to design, engineer and manufacture boards in Europe [1] around processors Qualcomm designs in San Diego and supplies under a separate long-term collaboration the two firms announced in March, covering the Brain and Nervous System reference architectures [3]. NEURA founder David Reger says roughly this himself in the announcement, that Physical AI "will not be built by one company alone", and that the aim is European industrial knowledge combined with the best technologies in the world [14]. Coherent, and narrower than the word sovereignty usually carries.
What Europe supplies here is module engineering and qualification at manufacturable volume, from a listed supplier in Arezzo whose existing customers include medical devices, transport, energy, defence and vending machines [4][5]. That is a scarce capability, and it is one assembled out of an existing industrial supplier base rather than built for humanoids. The stack in this deal is a German designer, an Italian board house and an American chip vendor.
The forcing function, for anyone running the same diligence on their own supply chain, is two axes rather than a slogan. The relevant test has two parts: whether the supplier controls the scarce component or only integrates it, and whether the supplier has shipped your volume before or only a volume two orders of magnitude below it. SECO controls the integration and not the processor, and it has not shipped humanoid volumes, which puts it in the same box as most good industrial suppliers. That box is a fine place to buy from and an expensive place to discover in year three, when the answer to a capacity question is a new line and somebody's capital.
