WEBVTT

1
00:00:00.096 --> 00:00:03.883
$3 billion in 11 days. That is what

2
00:00:03.883 --> 00:00:07.654
two Munich defence companies, Helsing and Quantum Systems,

3
00:00:07.654 --> 00:00:10.992
raised in the first two weeks of

4
00:00:11.730 --> 00:00:15.357
July. On September 5, Isar Aerospace reached

5
00:00:15.357 --> 00:00:19.112
orbit and deployed its payloads, the first commercial

6
00:00:19.369 --> 00:00:23.188
space company from Europe to deliver satellites into

7
00:00:23.236 --> 00:00:26.991
orbit. That was actually a historic event. And

8
00:00:27.472 --> 00:00:31.142
the same ecosystem from below: German financing

9
00:00:31.142 --> 00:00:34.731
rounds were already down 11% for the

10
00:00:34.763 --> 00:00:38.176
first half of 2026. Berlin pre-seed funding

11
00:00:38.289 --> 00:00:42.086
was down 65% compared with 2022, according to

12
00:00:42.615 --> 00:00:45.772
the city's own Business Angels association.

13
00:00:45.804 --> 00:00:49.153
And a $14.8 billion takeover offer for

14
00:00:49.410 --> 00:00:52.919
Delivery Hero from Uber. Germany passed the Strategic

15
00:00:53.272 --> 00:00:56.678
Necessity Test. This quarter raised a harder question:

16
00:00:56.727 --> 00:01:00.343
who ends up owning the companies that pass

17
00:01:00.698 --> 00:01:03.996
it? I call this the Ownership Test.

18
00:01:14.838 --> 00:01:17.282
Hello and welcome, everybody. This

19
00:01:17.282 --> 00:01:21.117
is episode 780 of Startuprad.io, the review of

20
00:01:21.327 --> 00:01:24.713
the third quarter of 2026. I'm Joe

21
00:01:24.713 --> 00:01:27.733
Menninger, joining you from Frankfurt am Main,

22
00:01:28.569 --> 00:01:32.312
Germany. This is the quarterly review for the

23
00:01:32.505 --> 00:01:36.168
startup ecosystems of Germany, Austria, and Switzerland, covering

24
00:01:36.666 --> 00:01:40.457
everything that happened between July 1 and September

25
00:01:40.843 --> 00:01:44.522
28, the day of recording. The factual claims

26
00:01:44.554 --> 00:01:48.362
are sourced; the interpretations and predictions are mine.

27
00:01:49.149 --> 00:01:52.795
So let's talk startups. Three numbers define

28
00:01:52.972 --> 00:01:56.088
this quarter. $18 billion: Helsing's valuation,

29
00:01:56.506 --> 00:01:59.750
by our count Germany's second decacorn after

30
00:01:59.895 --> 00:02:03.557
Celonis, and worth more than Lufthansa. 354:

31
00:02:03.991 --> 00:02:06.143
the number of German

32
00:02:06.705 --> 00:02:10.496
financing rounds in the first half of 2026,

33
00:02:10.496 --> 00:02:14.319
down 11%, with 67% of all capital in

34
00:02:14.400 --> 00:02:17.691
rounds above €50 million, per EY. And

35
00:02:17.820 --> 00:02:21.545
two: the number of flights it took

36
00:02:21.721 --> 00:02:25.222
Isar Aerospace's Spectrum to reach orbit when

37
00:02:25.222 --> 00:02:28.626
it deployed its payloads on September 5.

38
00:02:29.027 --> 00:02:32.383
A quick word on method. This review

39
00:02:33.218 --> 00:02:36.590
covers July 1 through September 28. We

40
00:02:36.911 --> 00:02:40.733
record a few days before quarter-end by design.

41
00:02:40.749 --> 00:02:44.346
Anything that lands between this recording and September

42
00:02:44.378 --> 00:02:48.127
30, of course, goes into the October news

43
00:02:48.626 --> 00:02:52.411
episode and into our Q4 and annual reviews. Yes,

44
00:02:52.411 --> 00:02:54.488
we do all of that

45
00:02:56.346 --> 00:02:59.961
for you. Announced deals are labelled announced. Completed

46
00:03:00.219 --> 00:03:02.979
deals are labelled completed. Company equity,

47
00:03:03.690 --> 00:03:07.410
secondaries, fund closes, and public grants are never

48
00:03:07.507 --> 00:03:11.236
added into one number. And where a figure

49
00:03:11.236 --> 00:03:15.075
comes from an interest group's own paper rather than

50
00:03:15.075 --> 00:03:17.896
neutral statistics, I will say so.

51
00:03:18.762 --> 00:03:22.529
Three things coming up. First, the prediction card.

52
00:03:22.529 --> 00:03:26.183
The H1 review put five predictions on record

53
00:03:26.216 --> 00:03:29.261
with deadlines and confidence levels, and one

54
00:03:29.486 --> 00:03:33.317
confirmed faster than I expected. Second, what the

55
00:03:33.429 --> 00:03:36.779
software companies that survived this market actually

56
00:03:36.779 --> 00:03:40.195
have in common. And third, the question

57
00:03:40.259 --> 00:03:43.763
nobody in the ecosystem wants to answer: when

58
00:03:44.344 --> 00:03:46.799
strategically necessary German companies win,

59
00:03:48.478 --> 00:03:51.828
who ends up owning them? The prediction

60
00:03:51.959 --> 00:03:55.519
card. Accountability first. The H1 2026 review made

61
00:03:55.552 --> 00:03:59.091
five predictions. Here is where each one stands

62
00:03:59.107 --> 00:04:02.433
as of this recording. Prediction one: KNDS

63
00:04:03.092 --> 00:04:06.525
formally restarts its IPO process before March

64
00:04:06.663 --> 00:04:10.458
31, 2027. Confidence was 7 out of 10.

65
00:04:10.458 --> 00:04:13.711
Status: live and volatile. Bottom line, we're

66
00:04:16.360 --> 00:04:20.189
waiting. KNDS postponed the listing on July 1.

67
00:04:20.383 --> 00:04:24.002
The Financial Times and CNBC reported that the

68
00:04:24.082 --> 00:04:27.930
company struggled to convince investors of a valuation

69
00:04:28.203 --> 00:04:31.257
above €12 billion. On August 26, Bloomberg

70
00:04:31.498 --> 00:04:35.305
reported KNDS was restarting investor meetings, targeting a

71
00:04:35.580 --> 00:04:38.953
possible relaunch in the second half of

72
00:04:39.773 --> 00:04:43.564
September. On September 16, Bloomberg reported the company

73
00:04:43.629 --> 00:04:47.067
and its owners were considering delaying again, possibly

74
00:04:47.260 --> 00:04:50.939
beyond this year. So: process activity, yes. A

75
00:04:50.939 --> 00:04:54.682
formal restart, announcement, filing, or timetable? Not yet.

76
00:04:55.068 --> 00:04:58.635
The clock runs to March 2027. The

77
00:04:58.667 --> 00:05:02.089
deeper signal is unchanged: private capital finances

78
00:05:02.362 --> 00:05:06.159
defence at record valuations while public markets

79
00:05:06.432 --> 00:05:10.213
keep saying not yet. Prediction two was

80
00:05:10.213 --> 00:05:14.027
the big one: at least two German

81
00:05:14.059 --> 00:05:17.392
companies in defence, robotics, space, quantum, or

82
00:05:17.969 --> 00:05:21.718
strategic infrastructure announce priced equity rounds above

83
00:05:21.750 --> 00:05:24.939
€1 billion before the end of

84
00:05:25.420 --> 00:05:29.185
2026. Confidence: 8 out of 10, the

85
00:05:29.250 --> 00:05:32.891
highest on the card. Status: confirmed within

86
00:05:32.891 --> 00:05:35.922
the first two weeks of the

87
00:05:36.419 --> 00:05:39.692
half. Quantum Systems announced $1.2 billion

88
00:05:40.028 --> 00:05:43.525
on July 2, just over €1.05 billion

89
00:05:43.718 --> 00:05:47.134
at that day's exchange rate. Helsing announced

90
00:05:47.487 --> 00:05:50.791
$1.8 billion on July 13. Both German.

91
00:05:51.176 --> 00:05:54.673
Both priced equity. Both above €1 billion.

92
00:05:55.026 --> 00:05:58.522
The prediction had until December. It needed

93
00:05:58.875 --> 00:06:01.922
only 11 days. Prediction three: SAP

94
00:06:02.243 --> 00:06:05.755
announces at least one further material AI

95
00:06:05.915 --> 00:06:09.010
acquisition beyond Prior Labs, n8n, and

96
00:06:09.683 --> 00:06:13.276
Parloa before December 31, 2026. Confidence: 6

97
00:06:13.372 --> 00:06:15.761
out of 10. Status:

98
00:06:16.852 --> 00:06:20.299
open, with a nuance. SAP completed two

99
00:06:21.117 --> 00:06:24.502
AI acquisitions inside Q3: Prior Labs

100
00:06:24.598 --> 00:06:28.019
closed July 16, with completion announced

101
00:06:28.051 --> 00:06:31.826
July 17, and the data-lakehouse platform Dremio

102
00:06:31.826 --> 00:06:35.504
on July 6. But both were announced

103
00:06:35.777 --> 00:06:38.637
in May, before the prediction

104
00:06:39.134 --> 00:06:42.652
was made. Completions are not new announcements.

105
00:06:42.909 --> 00:06:46.266
No genuinely new SAP AI acquisition

106
00:06:46.636 --> 00:06:50.314
was announced in Q3. The clock runs

107
00:06:50.314 --> 00:06:53.861
to December 31. Prediction four: DeepL announces

108
00:06:54.309 --> 00:06:57.820
a material strategic transaction, financing, or restructuring

109
00:06:57.916 --> 00:07:00.609
before June 30, 2027. Confidence:

110
00:07:01.523 --> 00:07:05.034
6 out of 10. Status: still open.

111
00:07:05.290 --> 00:07:08.529
DeepL did announce a significant restructuring,

112
00:07:08.529 --> 00:07:12.007
a workforce reduction of roughly 25% in

113
00:07:12.119 --> 00:07:15.566
its shift to an AI-centric operating model,

114
00:07:15.630 --> 00:07:19.222
but that announcement came on May 7,

115
00:07:19.318 --> 00:07:22.912
in the second quarter, before this prediction

116
00:07:22.944 --> 00:07:26.088
existed. Pre-existing signals do not confirm

117
00:07:26.153 --> 00:07:29.955
predictions. What it tells us is that

118
00:07:30.164 --> 00:07:33.982
the direction of the call was right.

119
00:07:34.094 --> 00:07:37.431
The qualifying event is still outstanding. Prediction

120
00:07:38.009 --> 00:07:41.330
five: at least one member of the

121
00:07:41.715 --> 00:07:45.052
Profitability Cohort, N26, Personio, Flink, or FINN,

122
00:07:45.100 --> 00:07:48.903
announces an IPO mandate, a regulatory filing, or a

123
00:07:49.016 --> 00:07:52.595
formal sale process before the end of 2026.

124
00:07:52.643 --> 00:07:56.206
Media speculation does not count. Confidence: 5 out

125
00:07:56.544 --> 00:08:00.059
of 10, deliberately the lowest on the card.

126
00:08:00.717 --> 00:08:04.553
Status: still open. No mandate, no filing, no process.

127
00:08:04.762 --> 00:08:08.598
The adjacent signal is Scalable Capital, €30 billion in

128
00:08:08.598 --> 00:08:11.503
assets under management and already pre-marketing its

129
00:08:11.808 --> 00:08:15.082
Altersvorsorgedepot, a German retirement-investment product. But Scalable

130
00:08:15.082 --> 00:08:18.696
is not in the cohort, and IPO chatter

131
00:08:18.857 --> 00:08:22.183
is exactly the speculation the prediction excludes.

132
00:08:22.183 --> 00:08:25.429
Interim scorecard: one confirmed at the highest

133
00:08:25.605 --> 00:08:29.237
confidence rating, zero missed, four on the clock.

134
00:08:29.237 --> 00:08:32.900
And from the monthly episodes, one more confirmation

135
00:08:32.996 --> 00:08:36.612
to log. In June, we predicted Isar Aerospace

136
00:08:36.612 --> 00:08:40.323
would reach orbit before the end of 2026.

137
00:08:40.323 --> 00:08:43.848
They did, on September 5. On its second

138
00:08:43.945 --> 00:08:46.431
flight, Spectrum delivered its payloads to

139
00:08:47.497 --> 00:08:51.291
orbit from Andøya. Sorry, it's in Norway; I butchered

140
00:08:51.291 --> 00:08:54.326
the name, I know. Confirmed with almost

141
00:08:56.179 --> 00:08:59.882
four months to spare. In the Q1 review,

142
00:08:59.979 --> 00:09:03.570
we said Germany's startup market had become a

143
00:09:03.618 --> 00:09:07.181
selection event. In the H1 review, we formalised

144
00:09:07.181 --> 00:09:10.509
the selection rule as the Strategic Necessity Test.

145
00:09:10.653 --> 00:09:14.434
Capital concentrates on companies that governments, industries,

146
00:09:15.219 --> 00:09:18.952
and supply chains really need to exist.

147
00:09:18.984 --> 00:09:22.829
And Q3 is the quarter that test

148
00:09:22.829 --> 00:09:26.066
got priced. The evidence: Helsing, $1.8

149
00:09:26.274 --> 00:09:30.103
billion at $18 billion, in a round

150
00:09:30.119 --> 00:09:32.811
led by Dragoneer with Lightspeed,

151
00:09:33.147 --> 00:09:36.255
ICONIQ, Growth Equity at Goldman

152
00:09:36.832 --> 00:09:40.552
Sachs Alternatives, JPMorganChase, CPP Investments, General Catalyst,

153
00:09:40.777 --> 00:09:44.087
Plural, and StepStone. Quantum Systems, $1.2

154
00:09:44.087 --> 00:09:47.735
billion at approximately $8 billion post-money,

155
00:09:47.992 --> 00:09:51.752
co-led by Blackstone, Noteus, Airbus, and Advent.

156
00:09:52.170 --> 00:09:55.994
Co-CEO Florian Seibel told the Financial Times

157
00:09:56.042 --> 00:09:59.786
that investors uncomfortable with a move into

158
00:09:59.883 --> 00:10:03.450
lethal systems had the opportunity to

159
00:10:03.578 --> 00:10:07.205
exit in the transaction. Proxima Fusion:

160
00:10:07.205 --> 00:10:10.860
€411 million at €2.4 billion, what

161
00:10:11.117 --> 00:10:14.659
the company calls the largest European

162
00:10:14.659 --> 00:10:18.394
fusion financing to date, with Google and

163
00:10:18.506 --> 00:10:21.921
RWE as strategic investors. RWE had

164
00:10:22.097 --> 00:10:25.896
already signed an agreement in February to

165
00:10:25.896 --> 00:10:29.583
work toward a stellarator plant at

166
00:10:29.647 --> 00:10:33.062
the former Gundremmingen nuclear site before

167
00:10:33.174 --> 00:10:36.669
it invested. The customer invested. That

168
00:10:36.861 --> 00:10:40.679
compresses the path from lab to grid.

169
00:10:40.727 --> 00:10:44.256
And on September 5, Isar Aerospace reached

170
00:10:44.465 --> 00:10:48.234
orbit on Spectrum's second flight and deployed its

171
00:10:48.234 --> 00:10:51.732
payloads, the first commercial company from Europe

172
00:10:51.844 --> 00:10:55.261
to deliver satellites into orbit and the

173
00:10:55.309 --> 00:10:58.277
first provider to hit the European

174
00:10:58.822 --> 00:11:02.255
Launcher Challenge orbital milestone ahead of the

175
00:11:02.865 --> 00:11:06.430
2027 deadline. Germany's government is backing a

176
00:11:06.462 --> 00:11:09.483
reported €35 billion space build-out by

177
00:11:09.660 --> 00:11:13.227
2030. The Strategic Necessity thesis does not

178
00:11:13.597 --> 00:11:17.052
get cleaner evidence than a rocket. Note

179
00:11:17.052 --> 00:11:20.731
what the lead investors of the quarter's biggest

180
00:11:20.747 --> 00:11:24.524
rounds have in common: Dragoneer, Blackstone, Goldman Sachs,

181
00:11:24.540 --> 00:11:27.882
Google. Global institutional allocators, not European venture

182
00:11:28.316 --> 00:11:32.085
funds. Hold that thought, but hold it carefully.

183
00:11:32.085 --> 00:11:35.883
It is the door into segment three,

184
00:11:36.075 --> 00:11:39.888
and it is not as simple as

185
00:11:40.257 --> 00:11:44.022
it looks. Participation collapse. Now the other

186
00:11:44.022 --> 00:11:47.788
side of the ledger. The August episode

187
00:11:47.852 --> 00:11:51.489
named it the concentration economy. At quarter

188
00:11:51.553 --> 00:11:55.223
scale, I want to be more precise:

189
00:11:55.239 --> 00:11:58.861
it is a participation collapse. The baseline

190
00:11:59.181 --> 00:12:02.933
numbers from EY's half-year data: €5.3 billion

191
00:12:03.237 --> 00:12:07.053
invested in German startups in the first half

192
00:12:07.245 --> 00:12:10.708
of 2026, up 14%, but rounds fell

193
00:12:10.708 --> 00:12:14.282
11% to 354, and 67% of all

194
00:12:14.539 --> 00:12:18.082
capital went into rounds above €50 million.

195
00:12:18.547 --> 00:12:22.378
Compare the peak: the first half of 2021

196
00:12:22.650 --> 00:12:26.161
had 588 rounds. Roughly 40% fewer financings

197
00:12:26.161 --> 00:12:29.513
than five years ago, with more money.

198
00:12:30.027 --> 00:12:33.768
Those are half-year figures. The Q3-specific count is

199
00:12:33.897 --> 00:12:37.558
not yet published, and we will score it

200
00:12:37.558 --> 00:12:41.155
in the annual review. At the base

201
00:12:41.187 --> 00:12:44.944
of the funnel, the picture darkens, and here

202
00:12:45.361 --> 00:12:48.621
the source matters. This is the Business

203
00:12:48.814 --> 00:12:52.555
Angels Club Berlin-Brandenburg's own position paper, an advocacy

204
00:12:52.844 --> 00:12:56.408
document, but with concrete numbers. According to

205
00:12:56.521 --> 00:13:00.294
the BACB, Berlin pre-seed funding is down 65%

206
00:13:01.337 --> 00:13:04.870
since 2022, back to roughly 2017 levels.

207
00:13:04.870 --> 00:13:08.531
The median angel ticket is down from

208
00:13:08.563 --> 00:13:12.095
€50,000 to €38,000. Startup insolvencies nearly doubled

209
00:13:12.256 --> 00:13:15.997
since 2022, even while formations hit a

210
00:13:16.029 --> 00:13:19.577
record 3,053 in the first half, per

211
00:13:19.802 --> 00:13:23.609
the Startup-Verband. And labour is concentrating too. Bitkom

212
00:13:23.657 --> 00:13:27.351
surveyed 102 German tech startups, a small

213
00:13:27.737 --> 00:13:31.174
sample and not representative, but directionally striking.

214
00:13:31.286 --> 00:13:34.467
27% had forgone new hires in

215
00:13:34.563 --> 00:13:38.273
the past 12 months because of AI.

216
00:13:38.739 --> 00:13:42.530
7% cut jobs because of it. 16% added

217
00:13:42.530 --> 00:13:46.194
staff because of it. The average startup

218
00:13:46.210 --> 00:13:49.738
in the survey now employs 12 people,

219
00:13:50.093 --> 00:13:53.685
down from 13. The ecosystem grows in

220
00:13:53.943 --> 00:13:56.644
value without growing in participation.

221
00:13:58.290 --> 00:14:02.086
Now look at the software companies that still won

222
00:14:02.102 --> 00:14:05.529
this quarter and what they have in common.

223
00:14:05.690 --> 00:14:09.490
Camunda, the Berlin process-orchestration company: annual recurring revenue nearing

224
00:14:09.539 --> 00:14:13.029
$200 million. The first $100 million took roughly

225
00:14:13.547 --> 00:14:17.239
10 years; the second is expected in about

226
00:14:17.741 --> 00:14:21.222
two, without a single new venture round since

227
00:14:21.222 --> 00:14:24.847
2021. Valued above $1 billion since 2025: a

228
00:14:24.960 --> 00:14:28.424
hidden unicorn built on revenue, not a raise.

229
00:14:28.909 --> 00:14:32.200
Pliant: past $100 million in annual recurring revenue.

230
00:14:32.296 --> 00:14:36.036
Moss: a unicorn at a €1 billion valuation

231
00:14:36.036 --> 00:14:39.474
on a Series C of just €35 million.

232
00:14:39.586 --> 00:14:42.475
In the H1 review, we noted

233
00:14:43.519 --> 00:14:46.857
software was absent from the top funding

234
00:14:47.194 --> 00:14:50.742
tier. In Q3, the software story is not

235
00:14:50.742 --> 00:14:54.433
about round size at all. It is about

236
00:14:54.690 --> 00:14:58.526
proof. My read: the door for software is

237
00:14:58.526 --> 00:15:01.512
open, but the currency at the door

238
00:15:02.041 --> 00:15:05.075
changed from story to revenue. Augustus is

239
00:15:05.910 --> 00:15:09.281
the other face of selectivity: a $180

240
00:15:09.297 --> 00:15:13.021
million Series B at a $1 billion valuation,

241
00:15:13.294 --> 00:15:16.745
raised while progressing toward a US national-bank launch,

242
00:15:16.745 --> 00:15:20.245
with preliminary conditional approval from the OCC and

243
00:15:20.277 --> 00:15:23.648
further US regulatory approvals reported since. Regulated

244
00:15:24.146 --> 00:15:27.742
infrastructure is the fintech lane that still prices.

245
00:15:27.742 --> 00:15:31.322
And the casualty list from the same system

246
00:15:31.514 --> 00:15:34.854
inside the quarter: Sono Motors, insolvency proceedings

247
00:15:34.966 --> 00:15:38.516
opened July 31. Glow25, roughly 120 job cuts,

248
00:15:38.661 --> 00:15:42.451
with management naming AI-enabled automation as a driver.

249
00:15:42.451 --> 00:15:45.872
Enpal, closing its Hamburg site, affecting up to

250
00:15:45.872 --> 00:15:49.390
85 roles, while guiding to over €1.3 billion

251
00:15:49.502 --> 00:15:53.004
of full-year revenue. Selection is not a phase

252
00:15:53.213 --> 00:15:56.602
of this market. It is the market.

253
00:15:56.730 --> 00:16:00.263
The Ownership Test. Now the thread that ties

254
00:16:00.263 --> 00:16:03.693
the quarter together, and the uncomfortable one. I

255
00:16:03.693 --> 00:16:07.268
want to be precise here because the evidence

256
00:16:08.021 --> 00:16:11.644
points in more than one direction, and that

257
00:16:11.644 --> 00:16:14.898
is exactly what makes it interesting. First,

258
00:16:15.556 --> 00:16:19.226
the anchor statistic from the September episode: according

259
00:16:19.339 --> 00:16:23.058
to KfW, more than three quarters of the

260
00:16:23.122 --> 00:16:26.616
venture money that went into German startups in

261
00:16:26.616 --> 00:16:30.287
the first quarter of 2026 came from abroad.

262
00:16:30.303 --> 00:16:33.237
Keep that number in mind. Now

263
00:16:33.509 --> 00:16:37.180
the outward-pointing evidence. Delivery Hero: Uber announced a

264
00:16:37.212 --> 00:16:40.034
voluntary takeover offer on July 16,

265
00:16:40.402 --> 00:16:44.153
€41.50 per share, an implied equity value of

266
00:16:44.330 --> 00:16:48.129
$14.8 billion for 100% of the company, or

267
00:16:48.177 --> 00:16:51.687
$13.7 billion adjusted for Uber's existing stake.

268
00:16:51.864 --> 00:16:55.647
Announced, not closed. The acceptance period runs until

269
00:16:55.711 --> 00:16:59.318
November 5, and closing is expected only

270
00:16:59.318 --> 00:17:03.037
in the second half of 2027. But if

271
00:17:03.037 --> 00:17:06.561
it closes, Berlin's biggest platform company belongs

272
00:17:07.160 --> 00:17:10.138
to San Francisco. AtaiBeckley: merger agreement

273
00:17:13.364 --> 00:17:16.358
with Eli Lilly dated July 15,

274
00:17:16.989 --> 00:17:20.759
roughly $2.8 billion upfront plus up to $1

275
00:17:20.873 --> 00:17:24.199
billion in milestone payments. Announced, not closed.

276
00:17:26.023 --> 00:17:29.525
Aleph Alpha, once the face of German

277
00:17:29.605 --> 00:17:33.216
AI sovereignty: on September 16, Cohere and

278
00:17:33.216 --> 00:17:36.743
Aleph Alpha signed a definitive business-combination agreement.

279
00:17:36.967 --> 00:17:40.494
Signed, not closed, but the flag on

280
00:17:40.815 --> 00:17:43.957
Germany's flagship AI lab is changing.

281
00:17:44.278 --> 00:17:47.452
Hugging Face, French-founded and US-headquartered, the

282
00:17:47.869 --> 00:17:51.717
open-model infrastructure half of Europe's AI stack

283
00:17:51.717 --> 00:17:55.260
depends on: NVIDIA announced the acquisition on

284
00:17:55.340 --> 00:17:59.011
September 3 at approximately $12.9 billion. And

285
00:17:59.428 --> 00:18:03.200
HappyRobot, the $1.2 billion unicorn we profiled

286
00:18:03.200 --> 00:18:06.655
in episode 770: formed in Munich, owned

287
00:18:06.655 --> 00:18:09.981
in San Francisco. And now the

288
00:18:11.235 --> 00:18:13.998
counter-evidence, and it is real.

289
00:18:14.850 --> 00:18:18.353
Helsing raised from Dragoneer, Goldman, and JPMorganChase.

290
00:18:19.413 --> 00:18:22.932
The company itself states it remained predominantly European-owned

291
00:18:23.109 --> 00:18:26.949
after the round. Foreign-led capital is not the

292
00:18:27.142 --> 00:18:30.846
same thing as foreign ownership. Langdock, which we

293
00:18:30.990 --> 00:18:34.213
flagged in July for tension over its

294
00:18:34.598 --> 00:18:37.854
US parent, reportedly did the opposite of

295
00:18:38.351 --> 00:18:42.103
the outbound trend, re-domiciling its parent to Germany

296
00:18:42.376 --> 00:18:46.160
as an SE, with the large majority now

297
00:18:46.160 --> 00:18:49.672
owned by founders and employees in the EU.

298
00:18:49.736 --> 00:18:53.168
SAP acted as a domestic acquirer twice, Prior

299
00:18:53.200 --> 00:18:56.823
Labs and Dremio, and committed more than €1

300
00:18:56.823 --> 00:19:00.125
billion over four years to build Prior

301
00:19:00.414 --> 00:19:04.261
Labs into a frontier AI lab in Europe. September

302
00:19:04.326 --> 00:19:07.804
delivered the cleanest counter-pattern of the whole quarter

303
00:19:07.852 --> 00:19:11.571
in 48 hours: on September 8, The Exploration

304
00:19:11.571 --> 00:19:15.066
Company, the German-parent space startup, raised $450 million.

305
00:19:15.146 --> 00:19:18.817
On September 10, the European Space Agency signed

306
00:19:19.266 --> 00:19:22.153
it for the ALADDIN programme, worth

307
00:19:22.153 --> 00:19:25.731
up to €760 million including optional service missions.

308
00:19:25.731 --> 00:19:29.532
Capital and customers, back to back, with ownership staying

309
00:19:29.821 --> 00:19:33.495
home. HyImpulse raised more than €50 million on

310
00:19:33.944 --> 00:19:37.777
September 2. On September 17, the German government and

311
00:19:37.777 --> 00:19:41.451
KfW started the next phase of the WIN

312
00:19:41.451 --> 00:19:44.980
initiative, the push to mobilise domestic scale-up capital.

313
00:19:44.996 --> 00:19:48.184
When Europe is the customer and the

314
00:19:48.460 --> 00:19:52.161
LP, ownership stays close. That is the pattern

315
00:19:53.004 --> 00:19:56.697
to build on. So here is the framework.

316
00:19:56.745 --> 00:20:00.416
The Strategic Necessity Test asks: does this company

317
00:20:00.529 --> 00:20:03.637
need to exist for governments, industries, and

318
00:20:04.216 --> 00:20:07.530
supply chains? Germany now passes it; the capital

319
00:20:07.530 --> 00:20:11.258
proves it. The Ownership Test asks the next question:

320
00:20:11.290 --> 00:20:14.970
when a strategically necessary company succeeds, who ends

321
00:20:14.970 --> 00:20:16.994
up owning it? Q3 did

322
00:20:19.028 --> 00:20:22.266
not settle that test. Q3 exposed it.

323
00:20:22.490 --> 00:20:25.841
The quarter's biggest pending exits point outward, Delivery

324
00:20:26.241 --> 00:20:29.913
Hero and AtaiBeckley, while its most strategic growth

325
00:20:30.249 --> 00:20:33.904
companies, Helsing, Proxima, and Isar, stay European-controlled on

326
00:20:34.033 --> 00:20:37.078
global capital. Necessity attracts capital. Ownership is

327
00:20:37.543 --> 00:20:41.295
decided by infrastructure, and that is the test Europe

328
00:20:41.391 --> 00:20:45.095
has not yet built for. What people are

329
00:20:45.095 --> 00:20:48.944
getting wrong: this is not a sellout morality tale,

330
00:20:48.944 --> 00:20:52.537
and it is not an argument against American

331
00:20:52.617 --> 00:20:56.290
capital. Dragoneer and Blackstone financing Munich defence companies

332
00:20:56.402 --> 00:20:59.626
is the system working. My thesis is

333
00:21:00.476 --> 00:21:04.261
that the gap sits on the European side of

334
00:21:04.261 --> 00:21:07.806
the table: a thin domestic LP base at

335
00:21:07.806 --> 00:21:11.463
pension scale, a public market where a KNDS

336
00:21:12.265 --> 00:21:15.878
could not price and no German tech listing

337
00:21:16.297 --> 00:21:20.122
of comparable scale happened, and SAP as the quarter's

338
00:21:20.493 --> 00:21:24.134
most visible domestic acquirer, carrying that role largely

339
00:21:24.134 --> 00:21:27.764
alone. Ownership follows infrastructure. Europe built the companies

340
00:21:27.861 --> 00:21:31.571
before it built the balance sheets to keep

341
00:21:31.748 --> 00:21:35.479
them. That is an interpretation, but it is

342
00:21:35.849 --> 00:21:39.371
the one the quarter's evidence keeps pointing at.

343
00:21:39.468 --> 00:21:43.015
The single thread of Q3 2026: the structural

344
00:21:43.368 --> 00:21:46.961
rotation completed and got priced. And the moment

345
00:21:46.961 --> 00:21:50.554
it was priced, the question moved. Q1 asked

346
00:21:51.067 --> 00:21:54.484
which companies survive: the selection event. H1

347
00:21:54.612 --> 00:21:57.820
asked which companies attract capital: the Strategic

348
00:21:58.446 --> 00:22:02.199
Necessity Test. Q3 asks who ends up owning

349
00:22:02.199 --> 00:22:05.552
the winners: the Ownership Test, posed, not

350
00:22:06.210 --> 00:22:09.740
yet decided. Three quarters, one arc, from survival

351
00:22:09.836 --> 00:22:13.593
to necessity to ownership. The fourth question, for Q4

352
00:22:13.674 --> 00:22:16.949
and beyond, is participation: whether an ecosystem of

353
00:22:17.206 --> 00:22:20.658
354 rounds, collapsing pre-seed, and 12-person startups keeps

354
00:22:20.995 --> 00:22:24.801
producing companies worth owning at all. For the framework

355
00:22:25.074 --> 00:22:28.751
being applied here, go back to two episodes: the

356
00:22:28.783 --> 00:22:32.444
H1 2026 review, 'Europe's Startup Recovery Was a

357
00:22:32.492 --> 00:22:36.152
Structural Rotation,' where the Strategic Necessity Test was

358
00:22:36.152 --> 00:22:39.667
defined, and Unicorn Atlas number one on Helsing,

359
00:22:39.940 --> 00:22:43.599
where the defence-capital thesis started. Links in the

360
00:22:43.615 --> 00:22:47.291
show notes. Takeaways: operators, investors, and ecosystem. If I

361
00:22:48.125 --> 00:22:51.704
were a founder or operator: three things. First,

362
00:22:51.961 --> 00:22:54.593
the lead investors in this quarter's

363
00:22:55.813 --> 00:22:59.489
mega-rounds were American and global institutions. If you build

364
00:23:00.066 --> 00:23:03.705
in a strategic-necessity category, plan your Series B

365
00:23:03.705 --> 00:23:07.280
and beyond with that map and decide before

366
00:23:07.296 --> 00:23:10.631
the term sheet what ownership outcome you are

367
00:23:10.983 --> 00:23:14.751
building toward, because that choice gets made whether you

368
00:23:14.815 --> 00:23:18.406
make it or not. Second, if you build

369
00:23:18.678 --> 00:23:22.173
software, the quarter's message is proof over narrative.

370
00:23:22.430 --> 00:23:25.139
The software names that mattered, Camunda, Pliant,

371
00:23:25.797 --> 00:23:29.040
and Moss, mattered for revenue, not round size.

372
00:23:29.651 --> 00:23:33.295
Third, run the AI headcount audit. In Bitkom's survey,

373
00:23:33.681 --> 00:23:37.502
half of startups already changed their staffing because of

374
00:23:37.759 --> 00:23:41.484
AI in one direction or another. If I were

375
00:23:41.613 --> 00:23:44.904
an investor, and these are observations, not recommendations,

376
00:23:44.904 --> 00:23:48.260
the interesting ends of the market are the

377
00:23:48.758 --> 00:23:50.749
two the mega-rounds ignore. At

378
00:23:53.918 --> 00:23:57.396
the bottom, if the BACB's numbers hold, pre-seed competition

379
00:23:57.541 --> 00:24:01.003
in Berlin is the thinnest in a decade

380
00:24:01.051 --> 00:24:04.286
for whoever still writes angel tickets. HTGF launched

381
00:24:05.234 --> 00:24:08.753
fundraising for its fifth seed fund on August 26,

382
00:24:08.785 --> 00:24:12.287
but public capital alone will not refill that base.

383
00:24:12.641 --> 00:24:16.457
At the top, assets coming out of restructurings, Sono's

384
00:24:16.539 --> 00:24:20.125
solar business and IP is up for sale, may

385
00:24:20.350 --> 00:24:24.197
be priced very differently from their long-term technical value.

386
00:24:24.197 --> 00:24:27.996
That is a market-inefficiency observation, the same one

387
00:24:28.141 --> 00:24:31.475
we made in August. And watch the

388
00:24:32.068 --> 00:24:35.675
Ownership Test as a return driver. If the

389
00:24:35.835 --> 00:24:38.705
Delivery Hero and AtaiBeckley transactions close

390
00:24:38.945 --> 00:24:42.023
in 2027, a significant wave of

391
00:24:42.023 --> 00:24:45.342
liquidity reaches German founders, employees, and angels,

392
00:24:45.374 --> 00:24:48.728
the recycling this ecosystem has been waiting

393
00:24:48.744 --> 00:24:52.227
for. If. If I were building the

394
00:24:52.757 --> 00:24:56.095
ecosystem, policy, institutions, infrastructure, my thesis from

395
00:24:56.127 --> 00:24:59.658
this quarter's evidence is that the binding

396
00:24:59.658 --> 00:25:03.477
constraint is shifting from company formation and capital

397
00:25:03.573 --> 00:25:06.992
attraction toward retention of ownership. The levers

398
00:25:07.104 --> 00:25:10.779
are known. The Altersvorsorgedepot launches January 1, 2027

399
00:25:11.773 --> 00:25:15.204
and could become a domestic capital channel

400
00:25:15.220 --> 00:25:18.698
over time; a public market where a

401
00:25:19.131 --> 00:25:22.690
KNDS or a DeepL can actually price;

402
00:25:22.738 --> 00:25:26.249
and procurement as anchor customer, the way

403
00:25:26.505 --> 00:25:30.048
ESA just demonstrated with The Exploration Company.

404
00:25:30.048 --> 00:25:33.863
Every quarter those levers stay unpulled, the Ownership Test keeps

405
00:25:34.296 --> 00:25:38.000
getting answered abroad. Our Q4 predictions, on the record.

406
00:25:38.128 --> 00:25:41.884
We have three new predictions, each with a deadline and

407
00:25:43.184 --> 00:25:45.896
a confidence level. Number one: KNDS does

408
00:25:46.490 --> 00:25:50.117
not complete its listing this year; the IPO slips

409
00:25:50.181 --> 00:25:53.760
into 2027. Confidence: 6 out of 10. Prediction two:

410
00:25:54.033 --> 00:25:57.772
Germany ends 2026 with at least 12 new unicorns

411
00:25:57.965 --> 00:26:01.319
for the year, up from 10 by late August.

412
00:26:01.736 --> 00:26:05.569
Confidence: 7 out of 10. Prediction three: at least

413
00:26:05.601 --> 00:26:09.122
one more DACH company valued above $1 billion agrees

414
00:26:09.494 --> 00:26:13.127
to an acquisition by a non-European buyer before December

415
00:26:13.594 --> 00:26:17.047
31. Confidence: 7 out of 10. These are

416
00:26:17.144 --> 00:26:19.225
forecasts, not facts. Check me

417
00:26:22.129 --> 00:26:25.406
in the annual review. And the four predictions

418
00:26:25.519 --> 00:26:29.150
still on the clock from the H1 card, KNDS

419
00:26:29.295 --> 00:26:33.053
restart, SAP's next AI acquisition, DeepL, and the Profitability

420
00:26:33.085 --> 00:26:36.867
Cohort, stay live. The annual review settles them. That

421
00:26:37.411 --> 00:26:41.225
is Q3 2026, the quarter the rotation got priced,

422
00:26:41.642 --> 00:26:45.312
participation kept narrowing, and the Ownership Test was posed.

423
00:26:45.424 --> 00:26:48.581
The annual review goes live in early January

424
00:26:49.110 --> 00:26:52.795
and will score every prediction on this card. I'll

425
00:26:53.821 --> 00:26:56.737
see you next quarter. If this episode

426
00:26:57.266 --> 00:27:01.099
sharpened how you see the market, send it to

427
00:27:01.131 --> 00:27:04.899
one founder who is still operating on last cycle's

428
00:27:04.899 --> 00:27:08.314
assumptions. This is Startuprad.io. That's all, folks. Thank

429
00:27:08.346 --> 00:27:10.142
you very much. Find more

430
00:27:10.142 --> 00:27:13.461
news, streams, events, and interviews at www.startuprad.io. Remember,

431
00:27:13.782 --> 00:27:14.808
sharing is caring.
