ctg. morning brief, 9 Oct 2026
On 9 Oct 2026, Munich's Tytan Technologies is building manufacturing capacity for up to 3,000 low-cost drone interceptors a month, Kerry Hoppe (business development and government affairs) told the Apex Europe conference in London on Wed 7 Oct (UK Defence Journal). Founded in 2023 after a Kyiv hackathon against Shaheds, it now has three interceptor types for NATO Class I and Class II threats, with sites in Munich, Berlin and Kyiv. Its €30m Series A in February, co-led by Armira and the NATO Innovation Fund, took total funding to €46m.
Why it matters: Hoppe's maths: missile defence at about a million per shot against drones that cost 30 to 50 grand. Tytan's own Series A release counts Shahed raids of more than 500 drones in a single night. You only fix that ratio with volume, and volume is a factory problem, not a prototype problem. 3,000 a month is a production number. The next test is whether orders arrive at the same rate.
Source: https://ukdefencejournal.org.uk/tytan-scales-up-low-cost-drone-interceptor-production/
French AI infrastructure company Sesterce plans an AI campus worth more than €10bn on the site of the former Kaipola paper mill in Jämsä, Finland (8 Oct). Phase one is 200 MW with works starting in 2026, phase two 600 MW, and the long-run aim is more than 1 GW in Finland. It says it has an anchor customer but won't name it. About 2,000 people during construction, around 300 once running. It commits to enabling 1 MW of new renewable capacity on the Finnish grid for every 1 MW it uses, within ten years of operation, and to no drinking water for cooling. Le Monde/AFP says the campus is due online in 2027.
Why it matters: Two days after Finland's LVV stopped Google's site work over missing environmental assessments, a French company shows the other route: an old mill that already has the grid connection, water and logistics, plus commitments audited by a third party and published every year. Europe's compute will get built on brownfield industrial sites. Mill towns that used to export paper can export compute. The number still missing is the customer.
Source: https://www.sesterce.com/newsroom/sesterce-plans-ai-campus-kaipola-jamsa-finland
Universal Quantum, the trapped-ion quantum computer builder from Haywards Heath with a hub in Hamburg, raised over $100m in a Series A co-led by DCVC and Firgun Ventures (8 Oct). The company calls it the largest Series A ever for a UK-headquartered quantum firm. It is delivering a $77m contract for the German Aerospace Center (DLR), has more than 100 technical staff, and is opening its first R&D centre outside Europe in Singapore.
Why it matters: Universal Quantum tiles many small chips into one machine, which is a manufacturing bet, not a physics paper. A German government contract carried the roadmap this far. The syndicate that scales it runs from Silicon Valley (DCVC) to Singapore's sovereign EDBI, Japan and Australian pension money. Good for the company. Europe should notice who writes the big quantum cheques.
Source: https://universalquantum.com/knowledge-hub/universal-quantum-raises-over-100m-series-a-to-scale-quantum-computing-globally
Tallinn's Golbriak Space raised a €4m seed, co-led by Join Capital and PhotonVentures, to industrialise its laser communication terminals and take production to thousands of units (Tech Funding News, 8 Oct). Two product lines: ERMESAI+ for space and ARES for aircraft, at 2.5 Gbps today with 10 Gbps planned for 2027. Before this round it had raised about $130,000, per CB Insights.
Why it matters: Germany's Mynaric, once Europe's big laser terminal hope, went through a restructuring that left shareholders with nothing and ended up with Rocket Lab. Constellations will need terminals by the thousand. A small team designing for volume and loose pointing is the kind of supplier Europe should start ordering from early.
Source: https://techfundingnews.com/golbriak-space-raises-4m-seed-optical-laser-terminals-join-capital-photonventures/
SemiAnalysis (8 Oct) built a dataset of 857 model releases from nine leading Chinese developers, 2021 to 15 Sep 2026. In SemiAnalysis's numbers, only 31 (3.6%) ever came with a published safety result from the developer, and just 9 (1.1%) had one at or before release. Beijing's AI Safety Governance Framework 3.0 (14 Sep) warns about loss of control and models that deceive evaluators, yet opens its principles with innovation as "the first priority."
Why it matters: China regulates outputs hard and the frontier barely. The EU does the opposite, with systemic-risk duties from 10²⁵ FLOP. Stop treating that as a handicap and sell it as a product: models a bank, a ministry or an army can audit. Aleph Alpha shipping Kolibri (3 Oct) with a documented AI Act, GDPR and copyright trail is what that looks like.
Source: https://newsletter.semianalysis.com/p/beijing-will-not-pace-the-frontier