WEBVTT

1
00:00:00.040 --> 00:00:05.000
<v Speaker 1>AI Daily Briefing. I'm ed sharp, thanks for joining me today.

2
00:00:05.240 --> 00:00:08.759
<v Speaker 1>The forty x valuation question. Can Cognition justify its price?

3
00:00:09.919 --> 00:00:12.359
<v Speaker 1>Cognition Ai is in early talks to raise at a

4
00:00:12.400 --> 00:00:15.960
<v Speaker 1>forty billion dollar valuation. That's a fifty percent jump from

5
00:00:16.000 --> 00:00:18.559
<v Speaker 1>the twenty six billion dollar raise it closed just ninety

6
00:00:18.640 --> 00:00:21.519
<v Speaker 1>days ago, and the revenue backing it is approaching one

7
00:00:21.559 --> 00:00:25.559
<v Speaker 1>billion dollars annualized. The question isn't whether those numbers are real,

8
00:00:26.079 --> 00:00:30.399
<v Speaker 1>it's whether they can last. Cognition builds devon, the autonomous

9
00:00:30.399 --> 00:00:34.119
<v Speaker 1>coding agent that attracted genuine commercial interest when most AI

10
00:00:34.200 --> 00:00:38.039
<v Speaker 1>tools were still fighting for developer attention. The company reportedly

11
00:00:38.079 --> 00:00:41.719
<v Speaker 1>doubled its sales in a single quarter. That's the kind

12
00:00:41.759 --> 00:00:45.240
<v Speaker 1>of growth that makes investors move fast and ask questions later.

13
00:00:45.880 --> 00:00:49.880
<v Speaker 1>Here's the key implication. At forty billion dollars against roughly

14
00:00:49.920 --> 00:00:52.960
<v Speaker 1>one billion in annualized revenue, you're looking at a forty

15
00:00:53.039 --> 00:00:57.280
<v Speaker 1>times revenue multiple. That's not a technology valuation. That's a

16
00:00:57.320 --> 00:01:02.079
<v Speaker 1>bet that quarterly doubling continues indefinitely. In practice, every company

17
00:01:02.079 --> 00:01:05.640
<v Speaker 1>that's ever growing this fast has eventually slowed. The question

18
00:01:05.719 --> 00:01:09.280
<v Speaker 1>for Cognition is when, not if, the talks are early

19
00:01:09.760 --> 00:01:12.799
<v Speaker 1>terms could shift, the deal could fall apart, but even

20
00:01:12.840 --> 00:01:15.560
<v Speaker 1>the fact that these conversations are happening tells us something

21
00:01:15.599 --> 00:01:19.120
<v Speaker 1>about where AI coding sits right now. The broader sector

22
00:01:19.239 --> 00:01:22.480
<v Speaker 1>is moving just as fast. Before AI, the venture rounds

23
00:01:22.480 --> 00:01:26.159
<v Speaker 1>in enterprise software typically came eighteen to twenty four months apart.

24
00:01:26.680 --> 00:01:30.879
<v Speaker 1>Founders needed time to show the growth investor FOMO an

25
00:01:30.920 --> 00:01:34.120
<v Speaker 1>annualized run rate math are doing the work that sustained

26
00:01:34.120 --> 00:01:38.040
<v Speaker 1>performance used to do that. Compression is not purely irrational.

27
00:01:38.640 --> 00:01:44.760
<v Speaker 1>Measurable outputs, working code, shipped features closed bugs give enterprise

28
00:01:44.840 --> 00:01:48.480
<v Speaker 1>buyers something they can price quickly. That's different from AI

29
00:01:48.599 --> 00:01:51.920
<v Speaker 1>jewels in categories where the value is fuzzier. The unit

30
00:01:52.000 --> 00:01:56.280
<v Speaker 1>economics in coding automation are closer to the surface. The

31
00:01:56.319 --> 00:01:58.959
<v Speaker 1>signal here is that when you can show a CFO

32
00:01:59.120 --> 00:02:02.599
<v Speaker 1>exactly what the soft is doing, the sale cycle shortens,

33
00:02:03.040 --> 00:02:06.400
<v Speaker 1>and when the sale cycle shortens, the growth numbers accelerate,

34
00:02:06.799 --> 00:02:11.000
<v Speaker 1>and when those numbers accelerate, the fundraising logic follows. It's

35
00:02:11.039 --> 00:02:14.680
<v Speaker 1>a tighter feedback loop than most other AI verticals have managed.

36
00:02:15.159 --> 00:02:18.759
<v Speaker 1>Here's the thing. On the acquisition side. SpaceX is expected

37
00:02:18.759 --> 00:02:21.879
<v Speaker 1>to close its deal to acquire Cursor this week. Cursor

38
00:02:21.960 --> 00:02:24.120
<v Speaker 1>is a code editor that reached two billion dollars an

39
00:02:24.159 --> 00:02:26.800
<v Speaker 1>annualized revenue, which is a number that would have seemed

40
00:02:26.840 --> 00:02:30.159
<v Speaker 1>implausible for a developer tool a few years ago. The

41
00:02:30.240 --> 00:02:34.479
<v Speaker 1>important distinction is what this acquisition signals. SpaceX is not

42
00:02:34.560 --> 00:02:38.240
<v Speaker 1>a software company by identity, but elil Musk has consistently

43
00:02:38.240 --> 00:02:41.479
<v Speaker 1>moved into adjacent infrastructure when it suits a larger strategy.

44
00:02:42.199 --> 00:02:45.039
<v Speaker 1>A two billion dollar revenue coding tool isn't just a product,

45
00:02:45.599 --> 00:02:48.520
<v Speaker 1>it's a distribution channel into every development team using it.

46
00:02:49.000 --> 00:02:52.080
<v Speaker 1>That's a different kind of asset. Anthropic is now being

47
00:02:52.120 --> 00:02:55.599
<v Speaker 1>discussed at a two trillion dollar IPO evaluation, with a

48
00:02:55.599 --> 00:02:58.960
<v Speaker 1>potential debut as early as October of next year that

49
00:02:59.000 --> 00:03:01.919
<v Speaker 1>would exceed space x his own record EDPO at one

50
00:03:02.000 --> 00:03:05.560
<v Speaker 1>point seven seven trillion, and Thropic hasn't officially set a date,

51
00:03:05.960 --> 00:03:09.520
<v Speaker 1>and the two trillion figure is investor expectation rather than

52
00:03:09.560 --> 00:03:14.840
<v Speaker 1>confirmed pricing. That distinction matters. Investor expectations at this stage

53
00:03:14.840 --> 00:03:17.879
<v Speaker 1>in a private company's life tend to move with sentiment

54
00:03:18.039 --> 00:03:22.759
<v Speaker 1>as much as fundamentals. Anthropics revenue trajectory is strong, but

55
00:03:22.879 --> 00:03:27.439
<v Speaker 1>two trillion requires that trajectory to continue without meaningful deceleration

56
00:03:27.919 --> 00:03:31.599
<v Speaker 1>through a public listing process that takes months. Consider this,

57
00:03:32.000 --> 00:03:34.599
<v Speaker 1>Deep Seek made a different kind of move this week.

58
00:03:34.960 --> 00:03:38.400
<v Speaker 1>The Chinese Ai Lab raised V four dash pro output

59
00:03:38.439 --> 00:03:42.039
<v Speaker 1>token prices from eighty seven cents per million to three

60
00:03:42.080 --> 00:03:44.879
<v Speaker 1>dollars in ninety six cents at peak, a jump of

61
00:03:44.960 --> 00:03:48.919
<v Speaker 1>roughly three hundred and fifty five percent. They also introduced

62
00:03:49.000 --> 00:03:52.439
<v Speaker 1>peak and off peak billing tiers and expanded thinking effort

63
00:03:52.560 --> 00:03:56.400
<v Speaker 1>levels to three tiers. Earlier this year, Deep Seek's aggressive

64
00:03:56.439 --> 00:03:58.879
<v Speaker 1>pricing was read as the start of a sustained AI

65
00:03:59.000 --> 00:04:04.479
<v Speaker 1>price war. This reversal complicates that narrative competitors including Noonshot,

66
00:04:04.520 --> 00:04:07.879
<v Speaker 1>Ali Baba and byte Dance have closed the capability gap.

67
00:04:07.960 --> 00:04:12.120
<v Speaker 1>Since Deep seeks are one moment went viral with less differentiation,

68
00:04:12.439 --> 00:04:16.120
<v Speaker 1>the pressure to hold margin increases. Price hikes at this

69
00:04:16.160 --> 00:04:19.720
<v Speaker 1>scale suggest Deep seek is betting its users won't leave easily.

70
00:04:20.439 --> 00:04:24.279
<v Speaker 1>The regulatory picture is worth tracking. A columnist, Money Baschizard,

71
00:04:24.319 --> 00:04:26.680
<v Speaker 1>has made a direct critique of how AI policy is

72
00:04:26.720 --> 00:04:29.879
<v Speaker 1>being built right now. The argument is that current regulation

73
00:04:30.000 --> 00:04:33.560
<v Speaker 1>rests on worst case predictions rather than observed effects. The

74
00:04:33.639 --> 00:04:37.000
<v Speaker 1>mass economic displacement that policy frameworks are designed to prevent

75
00:04:37.360 --> 00:04:41.560
<v Speaker 1>hasn't materialized in the data. Labour statistics and productivity numbers

76
00:04:41.600 --> 00:04:44.319
<v Speaker 1>have not moved in the ways that AI pessimists projected.

77
00:04:44.879 --> 00:04:48.160
<v Speaker 1>The tension here isn't between being for or against regulation.

78
00:04:48.839 --> 00:04:52.160
<v Speaker 1>It's about what counts is sufficient evidence to act. Predictive

79
00:04:52.199 --> 00:04:55.439
<v Speaker 1>caution has a cost if the predictions are wrong. Empirical

80
00:04:55.480 --> 00:04:57.720
<v Speaker 1>patience has a cost if the predictions turn out to

81
00:04:57.720 --> 00:05:01.639
<v Speaker 1>be early. That's a genuinely unresolved question, and its shaping

82
00:05:01.680 --> 00:05:04.959
<v Speaker 1>how governments approach the next wave of AI policy decisions.

83
00:05:05.839 --> 00:05:09.000
<v Speaker 1>Pulling back the thread connecting most of this is timing

84
00:05:09.079 --> 00:05:13.680
<v Speaker 1>and proof. Cognitions forty times revenue multiple is sustainable only

85
00:05:13.720 --> 00:05:18.399
<v Speaker 1>of doubling continues. Anthropics two trillion target depends on sustained

86
00:05:18.439 --> 00:05:22.560
<v Speaker 1>acceleration through a month's long IPO process. Deep seeks price

87
00:05:22.600 --> 00:05:25.800
<v Speaker 1>hike tests whether its users have real switching costs or

88
00:05:25.839 --> 00:05:28.720
<v Speaker 1>were just there for the price, and the regulation debate

89
00:05:28.879 --> 00:05:31.879
<v Speaker 1>is fundamentally asking how long you wait for evidence before

90
00:05:31.879 --> 00:05:34.800
<v Speaker 1>you act. Each of those is an unresolved proof point.

91
00:05:35.319 --> 00:05:38.879
<v Speaker 1>Watch for Cognitions round to close or store watch Anthropics

92
00:05:38.920 --> 00:05:42.279
<v Speaker 1>official IPO filing when it comes Watch, Deep seeks user

93
00:05:42.279 --> 00:05:45.600
<v Speaker 1>attention numbers, and Watch whether any major jurisdiction moves toward

94
00:05:45.680 --> 00:05:49.319
<v Speaker 1>evidence based AI policy frameworks in the months ahead. Those

95
00:05:49.360 --> 00:05:53.000
<v Speaker 1>are the signals that matter. Thanks for listening. This podcast

96
00:05:53.079 --> 00:05:56.240
<v Speaker 1>was built using AI technology, a yes we production