Latest / Deep Tech Germany – AI, Robotics & Frontier Innovation by Startuprad.io™ / Isar Aerospace Orbit & Proxima Fusion: Deep Tech Q3 2026

Transcript

Isar Aerospace Orbit & Proxima Fusion: Deep Tech Q3 2026 · Deep Tech Germany – AI, Robotics & Frontier Innovation by Startuprad.io™

• 0:00
  $3 billion in 11 days. That is what
• 0:03
  two Munich defence companies, Helsing and Quantum Systems,
• 0:07
  raised in the first two weeks of
• 0:11
  July. On September 5, Isar Aerospace reached
• 0:15
  orbit and deployed its payloads, the first commercial
• 0:19
  space company from Europe to deliver satellites into
• 0:23
  orbit. That was actually a historic event. And
• 0:27
  the same ecosystem from below: German financing
• 0:31
  rounds were already down 11% for the
• 0:34
  first half of 2026. Berlin pre-seed funding
• 0:38
  was down 65% compared with 2022, according to
• 0:42
  the city's own Business Angels association.
• 0:45
  And a $14.8 billion takeover offer for
• 0:49
  Delivery Hero from Uber. Germany passed the Strategic
• 0:53
  Necessity Test. This quarter raised a harder question:
• 0:56
  who ends up owning the companies that pass
• 1:00
  it? I call this the Ownership Test.
• 1:14
  Hello and welcome, everybody. This
• 1:17
  is episode 780 of Startuprad.io, the review of
• 1:21
  the third quarter of 2026. I'm Joe
• 1:24
  Menninger, joining you from Frankfurt am Main,
• 1:28
  Germany. This is the quarterly review for the
• 1:32
  startup ecosystems of Germany, Austria, and Switzerland, covering
• 1:36
  everything that happened between July 1 and September
• 1:40
  28, the day of recording. The factual claims
• 1:44
  are sourced; the interpretations and predictions are mine.
• 1:49
  So let's talk startups. Three numbers define
• 1:52
  this quarter. $18 billion: Helsing's valuation,
• 1:56
  by our count Germany's second decacorn after
• 1:59
  Celonis, and worth more than Lufthansa. 354:
• 2:03
  the number of German
• 2:06
  financing rounds in the first half of 2026,
• 2:10
  down 11%, with 67% of all capital in
• 2:14
  rounds above €50 million, per EY. And
• 2:17
  two: the number of flights it took
• 2:21
  Isar Aerospace's Spectrum to reach orbit when
• 2:25
  it deployed its payloads on September 5.
• 2:29
  A quick word on method. This review
• 2:33
  covers July 1 through September 28. We
• 2:36
  record a few days before quarter-end by design.
• 2:40
  Anything that lands between this recording and September
• 2:44
  30, of course, goes into the October news
• 2:48
  episode and into our Q4 and annual reviews. Yes,
• 2:52
  we do all of that
• 2:56
  for you. Announced deals are labelled announced. Completed
• 3:00
  deals are labelled completed. Company equity,
• 3:03
  secondaries, fund closes, and public grants are never
• 3:07
  added into one number. And where a figure
• 3:11
  comes from an interest group's own paper rather than
• 3:15
  neutral statistics, I will say so.
• 3:18
  Three things coming up. First, the prediction card.
• 3:22
  The H1 review put five predictions on record
• 3:26
  with deadlines and confidence levels, and one
• 3:29
  confirmed faster than I expected. Second, what the
• 3:33
  software companies that survived this market actually
• 3:36
  have in common. And third, the question
• 3:40
  nobody in the ecosystem wants to answer: when
• 3:44
  strategically necessary German companies win,
• 3:48
  who ends up owning them? The prediction
• 3:51
  card. Accountability first. The H1 2026 review made
• 3:55
  five predictions. Here is where each one stands
• 3:59
  as of this recording. Prediction one: KNDS
• 4:03
  formally restarts its IPO process before March
• 4:06
  31, 2027. Confidence was 7 out of 10.
• 4:10
  Status: live and volatile. Bottom line, we're
• 4:16
  waiting. KNDS postponed the listing on July 1.
• 4:20
  The Financial Times and CNBC reported that the
• 4:24
  company struggled to convince investors of a valuation
• 4:28
  above €12 billion. On August 26, Bloomberg
• 4:31
  reported KNDS was restarting investor meetings, targeting a
• 4:35
  possible relaunch in the second half of
• 4:39
  September. On September 16, Bloomberg reported the company
• 4:43
  and its owners were considering delaying again, possibly
• 4:47
  beyond this year. So: process activity, yes. A
• 4:50
  formal restart, announcement, filing, or timetable? Not yet.
• 4:55
  The clock runs to March 2027. The
• 4:58
  deeper signal is unchanged: private capital finances
• 5:02
  defence at record valuations while public markets
• 5:06
  keep saying not yet. Prediction two was
• 5:10
  the big one: at least two German
• 5:14
  companies in defence, robotics, space, quantum, or
• 5:17
  strategic infrastructure announce priced equity rounds above
• 5:21
  €1 billion before the end of
• 5:25
  2026. Confidence: 8 out of 10, the
• 5:29
  highest on the card. Status: confirmed within
• 5:32
  the first two weeks of the
• 5:36
  half. Quantum Systems announced $1.2 billion
• 5:40
  on July 2, just over €1.05 billion
• 5:43
  at that day's exchange rate. Helsing announced
• 5:47
  $1.8 billion on July 13. Both German.
• 5:51
  Both priced equity. Both above €1 billion.
• 5:55
  The prediction had until December. It needed
• 5:58
  only 11 days. Prediction three: SAP
• 6:02
  announces at least one further material AI
• 6:05
  acquisition beyond Prior Labs, n8n, and
• 6:09
  Parloa before December 31, 2026. Confidence: 6
• 6:13
  out of 10. Status:
• 6:16
  open, with a nuance. SAP completed two
• 6:21
  AI acquisitions inside Q3: Prior Labs
• 6:24
  closed July 16, with completion announced
• 6:28
  July 17, and the data-lakehouse platform Dremio
• 6:31
  on July 6. But both were announced
• 6:35
  in May, before the prediction
• 6:39
  was made. Completions are not new announcements.
• 6:42
  No genuinely new SAP AI acquisition
• 6:46
  was announced in Q3. The clock runs
• 6:50
  to December 31. Prediction four: DeepL announces
• 6:54
  a material strategic transaction, financing, or restructuring
• 6:57
  before June 30, 2027. Confidence:
• 7:01
  6 out of 10. Status: still open.
• 7:05
  DeepL did announce a significant restructuring,
• 7:08
  a workforce reduction of roughly 25% in
• 7:12
  its shift to an AI-centric operating model,
• 7:15
  but that announcement came on May 7,
• 7:19
  in the second quarter, before this prediction
• 7:22
  existed. Pre-existing signals do not confirm
• 7:26
  predictions. What it tells us is that
• 7:30
  the direction of the call was right.
• 7:34
  The qualifying event is still outstanding. Prediction
• 7:38
  five: at least one member of the
• 7:41
  Profitability Cohort, N26, Personio, Flink, or FINN,
• 7:45
  announces an IPO mandate, a regulatory filing, or a
• 7:49
  formal sale process before the end of 2026.
• 7:52
  Media speculation does not count. Confidence: 5 out
• 7:56
  of 10, deliberately the lowest on the card.
• 8:00
  Status: still open. No mandate, no filing, no process.
• 8:04
  The adjacent signal is Scalable Capital, €30 billion in
• 8:08
  assets under management and already pre-marketing its
• 8:11
  Altersvorsorgedepot, a German retirement-investment product. But Scalable
• 8:15
  is not in the cohort, and IPO chatter
• 8:18
  is exactly the speculation the prediction excludes.
• 8:22
  Interim scorecard: one confirmed at the highest
• 8:25
  confidence rating, zero missed, four on the clock.
• 8:29
  And from the monthly episodes, one more confirmation
• 8:32
  to log. In June, we predicted Isar Aerospace
• 8:36
  would reach orbit before the end of 2026.
• 8:40
  They did, on September 5. On its second
• 8:43
  flight, Spectrum delivered its payloads to
• 8:47
  orbit from Andøya. Sorry, it's in Norway; I butchered
• 8:51
  the name, I know. Confirmed with almost
• 8:56
  four months to spare. In the Q1 review,
• 8:59
  we said Germany's startup market had become a
• 9:03
  selection event. In the H1 review, we formalised
• 9:07
  the selection rule as the Strategic Necessity Test.
• 9:10
  Capital concentrates on companies that governments, industries,
• 9:15
  and supply chains really need to exist.
• 9:18
  And Q3 is the quarter that test
• 9:22
  got priced. The evidence: Helsing, $1.8
• 9:26
  billion at $18 billion, in a round
• 9:30
  led by Dragoneer with Lightspeed,
• 9:33
  ICONIQ, Growth Equity at Goldman
• 9:36
  Sachs Alternatives, JPMorganChase, CPP Investments, General Catalyst,
• 9:40
  Plural, and StepStone. Quantum Systems, $1.2
• 9:44
  billion at approximately $8 billion post-money,
• 9:47
  co-led by Blackstone, Noteus, Airbus, and Advent.
• 9:52
  Co-CEO Florian Seibel told the Financial Times
• 9:56
  that investors uncomfortable with a move into
• 9:59
  lethal systems had the opportunity to
• 10:03
  exit in the transaction. Proxima Fusion:
• 10:07
  €411 million at €2.4 billion, what
• 10:11
  the company calls the largest European
• 10:14
  fusion financing to date, with Google and
• 10:18
  RWE as strategic investors. RWE had
• 10:22
  already signed an agreement in February to
• 10:25
  work toward a stellarator plant at
• 10:29
  the former Gundremmingen nuclear site before
• 10:33
  it invested. The customer invested. That
• 10:36
  compresses the path from lab to grid.
• 10:40
  And on September 5, Isar Aerospace reached
• 10:44
  orbit on Spectrum's second flight and deployed its
• 10:48
  payloads, the first commercial company from Europe
• 10:51
  to deliver satellites into orbit and the
• 10:55
  first provider to hit the European
• 10:58
  Launcher Challenge orbital milestone ahead of the
• 11:02
  2027 deadline. Germany's government is backing a
• 11:06
  reported €35 billion space build-out by
• 11:09
  2030. The Strategic Necessity thesis does not
• 11:13
  get cleaner evidence than a rocket. Note
• 11:17
  what the lead investors of the quarter's biggest
• 11:20
  rounds have in common: Dragoneer, Blackstone, Goldman Sachs,
• 11:24
  Google. Global institutional allocators, not European venture
• 11:28
  funds. Hold that thought, but hold it carefully.
• 11:32
  It is the door into segment three,
• 11:36
  and it is not as simple as
• 11:40
  it looks. Participation collapse. Now the other
• 11:44
  side of the ledger. The August episode
• 11:47
  named it the concentration economy. At quarter
• 11:51
  scale, I want to be more precise:
• 11:55
  it is a participation collapse. The baseline
• 11:59
  numbers from EY's half-year data: €5.3 billion
• 12:03
  invested in German startups in the first half
• 12:07
  of 2026, up 14%, but rounds fell
• 12:10
  11% to 354, and 67% of all
• 12:14
  capital went into rounds above €50 million.
• 12:18
  Compare the peak: the first half of 2021
• 12:22
  had 588 rounds. Roughly 40% fewer financings
• 12:26
  than five years ago, with more money.
• 12:30
  Those are half-year figures. The Q3-specific count is
• 12:33
  not yet published, and we will score it
• 12:37
  in the annual review. At the base
• 12:41
  of the funnel, the picture darkens, and here
• 12:45
  the source matters. This is the Business
• 12:48
  Angels Club Berlin-Brandenburg's own position paper, an advocacy
• 12:52
  document, but with concrete numbers. According to
• 12:56
  the BACB, Berlin pre-seed funding is down 65%
• 13:01
  since 2022, back to roughly 2017 levels.
• 13:04
  The median angel ticket is down from
• 13:08
  €50,000 to €38,000. Startup insolvencies nearly doubled
• 13:12
  since 2022, even while formations hit a
• 13:16
  record 3,053 in the first half, per
• 13:19
  the Startup-Verband. And labour is concentrating too. Bitkom
• 13:23
  surveyed 102 German tech startups, a small
• 13:27
  sample and not representative, but directionally striking.
• 13:31
  27% had forgone new hires in
• 13:34
  the past 12 months because of AI.
• 13:38
  7% cut jobs because of it. 16% added
• 13:42
  staff because of it. The average startup
• 13:46
  in the survey now employs 12 people,
• 13:50
  down from 13. The ecosystem grows in
• 13:53
  value without growing in participation.
• 13:58
  Now look at the software companies that still won
• 14:02
  this quarter and what they have in common.
• 14:05
  Camunda, the Berlin process-orchestration company: annual recurring revenue nearing
• 14:09
  $200 million. The first $100 million took roughly
• 14:13
  10 years; the second is expected in about
• 14:17
  two, without a single new venture round since
• 14:21
  2021. Valued above $1 billion since 2025: a
• 14:24
  hidden unicorn built on revenue, not a raise.
• 14:28
  Pliant: past $100 million in annual recurring revenue.
• 14:32
  Moss: a unicorn at a €1 billion valuation
• 14:36
  on a Series C of just €35 million.
• 14:39
  In the H1 review, we noted
• 14:43
  software was absent from the top funding
• 14:47
  tier. In Q3, the software story is not
• 14:50
  about round size at all. It is about
• 14:54
  proof. My read: the door for software is
• 14:58
  open, but the currency at the door
• 15:02
  changed from story to revenue. Augustus is
• 15:05
  the other face of selectivity: a $180
• 15:09
  million Series B at a $1 billion valuation,
• 15:13
  raised while progressing toward a US national-bank launch,
• 15:16
  with preliminary conditional approval from the OCC and
• 15:20
  further US regulatory approvals reported since. Regulated
• 15:24
  infrastructure is the fintech lane that still prices.
• 15:27
  And the casualty list from the same system
• 15:31
  inside the quarter: Sono Motors, insolvency proceedings
• 15:34
  opened July 31. Glow25, roughly 120 job cuts,
• 15:38
  with management naming AI-enabled automation as a driver.
• 15:42
  Enpal, closing its Hamburg site, affecting up to
• 15:45
  85 roles, while guiding to over €1.3 billion
• 15:49
  of full-year revenue. Selection is not a phase
• 15:53
  of this market. It is the market.
• 15:56
  The Ownership Test. Now the thread that ties
• 16:00
  the quarter together, and the uncomfortable one. I
• 16:03
  want to be precise here because the evidence
• 16:08
  points in more than one direction, and that
• 16:11
  is exactly what makes it interesting. First,
• 16:15
  the anchor statistic from the September episode: according
• 16:19
  to KfW, more than three quarters of the
• 16:23
  venture money that went into German startups in
• 16:26
  the first quarter of 2026 came from abroad.
• 16:30
  Keep that number in mind. Now
• 16:33
  the outward-pointing evidence. Delivery Hero: Uber announced a
• 16:37
  voluntary takeover offer on July 16,
• 16:40
  €41.50 per share, an implied equity value of
• 16:44
  $14.8 billion for 100% of the company, or
• 16:48
  $13.7 billion adjusted for Uber's existing stake.
• 16:51
  Announced, not closed. The acceptance period runs until
• 16:55
  November 5, and closing is expected only
• 16:59
  in the second half of 2027. But if
• 17:03
  it closes, Berlin's biggest platform company belongs
• 17:07
  to San Francisco. AtaiBeckley: merger agreement
• 17:13
  with Eli Lilly dated July 15,
• 17:16
  roughly $2.8 billion upfront plus up to $1
• 17:20
  billion in milestone payments. Announced, not closed.
• 17:26
  Aleph Alpha, once the face of German
• 17:29
  AI sovereignty: on September 16, Cohere and
• 17:33
  Aleph Alpha signed a definitive business-combination agreement.
• 17:36
  Signed, not closed, but the flag on
• 17:40
  Germany's flagship AI lab is changing.
• 17:44
  Hugging Face, French-founded and US-headquartered, the
• 17:47
  open-model infrastructure half of Europe's AI stack
• 17:51
  depends on: NVIDIA announced the acquisition on
• 17:55
  September 3 at approximately $12.9 billion. And
• 17:59
  HappyRobot, the $1.2 billion unicorn we profiled
• 18:03
  in episode 770: formed in Munich, owned
• 18:06
  in San Francisco. And now the
• 18:11
  counter-evidence, and it is real.
• 18:14
  Helsing raised from Dragoneer, Goldman, and JPMorganChase.
• 18:19
  The company itself states it remained predominantly European-owned
• 18:23
  after the round. Foreign-led capital is not the
• 18:27
  same thing as foreign ownership. Langdock, which we
• 18:30
  flagged in July for tension over its
• 18:34
  US parent, reportedly did the opposite of
• 18:38
  the outbound trend, re-domiciling its parent to Germany
• 18:42
  as an SE, with the large majority now
• 18:46
  owned by founders and employees in the EU.
• 18:49
  SAP acted as a domestic acquirer twice, Prior
• 18:53
  Labs and Dremio, and committed more than €1
• 18:56
  billion over four years to build Prior
• 19:00
  Labs into a frontier AI lab in Europe. September
• 19:04
  delivered the cleanest counter-pattern of the whole quarter
• 19:07
  in 48 hours: on September 8, The Exploration
• 19:11
  Company, the German-parent space startup, raised $450 million.
• 19:15
  On September 10, the European Space Agency signed
• 19:19
  it for the ALADDIN programme, worth
• 19:22
  up to €760 million including optional service missions.
• 19:25
  Capital and customers, back to back, with ownership staying
• 19:29
  home. HyImpulse raised more than €50 million on
• 19:33
  September 2. On September 17, the German government and
• 19:37
  KfW started the next phase of the WIN
• 19:41
  initiative, the push to mobilise domestic scale-up capital.
• 19:44
  When Europe is the customer and the
• 19:48
  LP, ownership stays close. That is the pattern
• 19:53
  to build on. So here is the framework.
• 19:56
  The Strategic Necessity Test asks: does this company
• 20:00
  need to exist for governments, industries, and
• 20:04
  supply chains? Germany now passes it; the capital
• 20:07
  proves it. The Ownership Test asks the next question:
• 20:11
  when a strategically necessary company succeeds, who ends
• 20:14
  up owning it? Q3 did
• 20:19
  not settle that test. Q3 exposed it.
• 20:22
  The quarter's biggest pending exits point outward, Delivery
• 20:26
  Hero and AtaiBeckley, while its most strategic growth
• 20:30
  companies, Helsing, Proxima, and Isar, stay European-controlled on
• 20:34
  global capital. Necessity attracts capital. Ownership is
• 20:37
  decided by infrastructure, and that is the test Europe
• 20:41
  has not yet built for. What people are
• 20:45
  getting wrong: this is not a sellout morality tale,
• 20:48
  and it is not an argument against American
• 20:52
  capital. Dragoneer and Blackstone financing Munich defence companies
• 20:56
  is the system working. My thesis is
• 21:00
  that the gap sits on the European side of
• 21:04
  the table: a thin domestic LP base at
• 21:07
  pension scale, a public market where a KNDS
• 21:12
  could not price and no German tech listing
• 21:16
  of comparable scale happened, and SAP as the quarter's
• 21:20
  most visible domestic acquirer, carrying that role largely
• 21:24
  alone. Ownership follows infrastructure. Europe built the companies
• 21:27
  before it built the balance sheets to keep
• 21:31
  them. That is an interpretation, but it is
• 21:35
  the one the quarter's evidence keeps pointing at.
• 21:39
  The single thread of Q3 2026: the structural
• 21:43
  rotation completed and got priced. And the moment
• 21:46
  it was priced, the question moved. Q1 asked
• 21:51
  which companies survive: the selection event. H1
• 21:54
  asked which companies attract capital: the Strategic
• 21:58
  Necessity Test. Q3 asks who ends up owning
• 22:02
  the winners: the Ownership Test, posed, not
• 22:06
  yet decided. Three quarters, one arc, from survival
• 22:09
  to necessity to ownership. The fourth question, for Q4
• 22:13
  and beyond, is participation: whether an ecosystem of
• 22:17
  354 rounds, collapsing pre-seed, and 12-person startups keeps
• 22:20
  producing companies worth owning at all. For the framework
• 22:25
  being applied here, go back to two episodes: the
• 22:28
  H1 2026 review, 'Europe's Startup Recovery Was a
• 22:32
  Structural Rotation,' where the Strategic Necessity Test was
• 22:36
  defined, and Unicorn Atlas number one on Helsing,
• 22:39
  where the defence-capital thesis started. Links in the
• 22:43
  show notes. Takeaways: operators, investors, and ecosystem. If I
• 22:48
  were a founder or operator: three things. First,
• 22:51
  the lead investors in this quarter's
• 22:55
  mega-rounds were American and global institutions. If you build
• 23:00
  in a strategic-necessity category, plan your Series B
• 23:03
  and beyond with that map and decide before
• 23:07
  the term sheet what ownership outcome you are
• 23:10
  building toward, because that choice gets made whether you
• 23:14
  make it or not. Second, if you build
• 23:18
  software, the quarter's message is proof over narrative.
• 23:22
  The software names that mattered, Camunda, Pliant,
• 23:25
  and Moss, mattered for revenue, not round size.
• 23:29
  Third, run the AI headcount audit. In Bitkom's survey,
• 23:33
  half of startups already changed their staffing because of
• 23:37
  AI in one direction or another. If I were
• 23:41
  an investor, and these are observations, not recommendations,
• 23:44
  the interesting ends of the market are the
• 23:48
  two the mega-rounds ignore. At
• 23:53
  the bottom, if the BACB's numbers hold, pre-seed competition
• 23:57
  in Berlin is the thinnest in a decade
• 24:01
  for whoever still writes angel tickets. HTGF launched
• 24:05
  fundraising for its fifth seed fund on August 26,
• 24:08
  but public capital alone will not refill that base.
• 24:12
  At the top, assets coming out of restructurings, Sono's
• 24:16
  solar business and IP is up for sale, may
• 24:20
  be priced very differently from their long-term technical value.
• 24:24
  That is a market-inefficiency observation, the same one
• 24:28
  we made in August. And watch the
• 24:32
  Ownership Test as a return driver. If the
• 24:35
  Delivery Hero and AtaiBeckley transactions close
• 24:38
  in 2027, a significant wave of
• 24:42
  liquidity reaches German founders, employees, and angels,
• 24:45
  the recycling this ecosystem has been waiting
• 24:48
  for. If. If I were building the
• 24:52
  ecosystem, policy, institutions, infrastructure, my thesis from
• 24:56
  this quarter's evidence is that the binding
• 24:59
  constraint is shifting from company formation and capital
• 25:03
  attraction toward retention of ownership. The levers
• 25:07
  are known. The Altersvorsorgedepot launches January 1, 2027
• 25:11
  and could become a domestic capital channel
• 25:15
  over time; a public market where a
• 25:19
  KNDS or a DeepL can actually price;
• 25:22
  and procurement as anchor customer, the way
• 25:26
  ESA just demonstrated with The Exploration Company.
• 25:30
  Every quarter those levers stay unpulled, the Ownership Test keeps
• 25:34
  getting answered abroad. Our Q4 predictions, on the record.
• 25:38
  We have three new predictions, each with a deadline and
• 25:43
  a confidence level. Number one: KNDS does
• 25:46
  not complete its listing this year; the IPO slips
• 25:50
  into 2027. Confidence: 6 out of 10. Prediction two:
• 25:54
  Germany ends 2026 with at least 12 new unicorns
• 25:57
  for the year, up from 10 by late August.
• 26:01
  Confidence: 7 out of 10. Prediction three: at least
• 26:05
  one more DACH company valued above $1 billion agrees
• 26:09
  to an acquisition by a non-European buyer before December
• 26:13
  31. Confidence: 7 out of 10. These are
• 26:17
  forecasts, not facts. Check me
• 26:22
  in the annual review. And the four predictions
• 26:25
  still on the clock from the H1 card, KNDS
• 26:29
  restart, SAP's next AI acquisition, DeepL, and the Profitability
• 26:33
  Cohort, stay live. The annual review settles them. That
• 26:37
  is Q3 2026, the quarter the rotation got priced,
• 26:41
  participation kept narrowing, and the Ownership Test was posed.
• 26:45
  The annual review goes live in early January
• 26:49
  and will score every prediction on this card. I'll
• 26:53
  see you next quarter. If this episode
• 26:57
  sharpened how you see the market, send it to
• 27:01
  one founder who is still operating on last cycle's
• 27:04
  assumptions. This is Startuprad.io. That's all, folks. Thank
• 27:08
  you very much. Find more
• 27:10
  news, streams, events, and interviews at www.startuprad.io. Remember,
• 27:13
  sharing is caring.