This week the hard part of deep tech was production. A Munich isotope maker sold itself for its supply chain, and Eindhoven's photonics foundries added capacity.
Where production faltered, so did the story. Europe installed fewer factory robots in 2025 while Asia and the Americas installed more, and a Lyon microbiome company learned that a single-arm trial would not carry it to market.
Lead
Telix Buys ITM for the Part That Is Hardest to Build
Telix Pharmaceuticals agreed on 20 September to buy ITM Isotope Technologies Munich for $1.65 billion upfront. Up to $700 million more depends on approvals and sales of ITM-11, and Telix also takes on more than $300 million of debt.
ITM supplies lutetium-177 and other isotopes to nuclear medicine in more than 65 countries and booked $273 million in revenue last year. Analysts at William Blair said the deal "de-risks one of the most complex aspects of operating a radiopharmaceutical company".
The FDA rejected ITM-11 in August over manufacturing issues at a third-party facility; no safety or efficacy problems were raised. Telix paid for a German production network, the step in radiopharma that is hardest to copy.
(BioSpace)
Biotech & Life Sciences
EMA Backs Zealand's Short Bowel Drug
The CHMP recommended Zeydovio, Zealand Pharma's long-acting GLP-2 analogue glepaglutide, for short bowel syndrome at its 14 to 17 September meeting. Zealand calls it the first major advance in the condition in Europe in over a decade.
In the Phase 3 EASE-1 trial, a twice-weekly 10 mg autoinjector cut weekly parenteral support by 5.13 litres against 2.85 on placebo, and 14 per cent of treated patients came off intravenous feeding entirely.
A Commission decision is expected by mid-November. Zealand is seeking a commercial partner, and a US filing waits on the EASE-5 trial.
(EMA)
Quantum & Photonics
Pasqal Reports Its First Half as a Listed Company
Pasqal, which listed on Nasdaq on 28 August through a SPAC, reported first-half revenue of EUR 4.9 million, up 14 per cent. The operating loss grew to EUR 59.2 million, of which EUR 37.5 million was share-based pay and one-off listing costs.
Booked and awarded business, including grants and tax credits, stood at EUR 70.4 million, and cash reached about EUR 312.9 million after the deal closed. In the half it delivered a 140-qubit machine to Italy for integration with the Leonardo supercomputer and put a 200-qubit system in Saudi Arabia.
Pasqal has hardware installed at customers, and revenue still smaller than its share-based pay. Public investors will now watch that gap every six months.
Eindhoven Adds the Laser to Someone Else's Silicon
Smart Photonics and GlobalFoundries launched a joint foundry service on 22 September that flip-chips Smart Photonics' indium phosphide lasers, modulators, and amplifiers into cavities on GlobalFoundries' silicon photonics wafers. General availability is planned for the second half of 2027.
Silicon photonics rides CMOS manufacturing but cannot make light; indium phosphide can. Until now designers had to integrate the two themselves.
TNO said the same week that its Eindhoven InP pilot line could double to 20,000 six-inch wafers a year. The Netherlands is positioning itself as the place other companies' photonics gets made.
Space
Ramses Powers Up in Milan With No Room to Slip
ESA and JAXA's Ramses planetary defence spacecraft was switched on for the first time at OHB Italia's Milan cleanroom, with its onboard computer and subsystems running together inside the spacecraft body.
Ramses must launch on a Japanese H3 by spring 2028 to meet the asteroid Apophis as it passes Earth on 13 April 2029. That is less than three and a half years from contract to lift-off, so the two halves of the spacecraft are being built in parallel.
The asteroid sets the deadline, which makes Ramses a test of whether ESA can hold a schedule that would be tight for a commercial builder.
(ESA)
Robotics & Automation
The EU Installed Fewer Robots While Everyone Else Installed More
The International Federation of Robotics said on 24 September that EU installations fell 11 per cent in 2025 to 60,500 units. Asia rose 14 per cent to 457,315 and the Americas rose 14 per cent to 57,044.
The installed base still passed 700,000. Germany took 41 per cent of installations and Italy came second with 7,800 units, most of them in metalworking.
Europe's physical AI startups raise money on a thesis about factory demand. In 2025 that demand paused, and it still leans on car and metal plants.
Semiconductors
Nexperia's European Side Places Its First Big Bet in India
Nexperia signed a partnership with Tata Electronics on 18 September to make part of its MOSFET range at Tata's 300 mm fab in Dholera, Gujarat, with assembly and test at Tata's plant in Jagiroad, Assam. Both sites are still being built, a combined $14 billion investment.
The ownership dispute has split Nexperia's European and Chinese operations, and the Chinese side is working towards fully domestic production. In July we asked whether Nexperia would join France's packaging initiative; its first large new manufacturing partner is in India.
Poland's State Bank Names Its Chip Niche
Bank Gospodarstwa Krajowego, Poland's state development bank, published a report on 25 September forecasting the Polish semiconductor market will grow from $5.1 billion to $23.7 billion by 2035. It recommends specialising in design, power electronics, back-end assembly and test, and gallium nitride.
The GaN case rests on a crystal growth method from the Polish Academy of Sciences' high-pressure institute that yields record-low defect density.
Intel and Trumpf Hüttinger employ 92 per cent of Poland's 5,000-plus back-end workers. The report is a plan to widen a very narrow base.
One to Watch
MaaT Pharma Needs a Randomised Trial and the Money to Run It
The CHMP confirmed on 18 September its refusal of Xervyteg, MaaT Pharma's pooled faecal microbiota therapy for acute graft-versus-host disease. A single-arm trial did not let it characterise benefit and risk.
In June we called this vote microbiome therapy's first European test.
MaaT now plans PHOENIX, a randomised Phase 3 trial of about 138 patients against best available therapy in the US and Europe, with FDA feedback backing the design. Enrolment is targeted for the first half of 2027, but only if it can raise the money, and its cash lasts until December 2026.
Watch the financing before year end. The regulator has asked for the evidence a small company can least afford, and MaaT has three months of cash to find it.