WEBVTT

1
00:00:00.040 --> 00:00:05.240
<v Speaker 1>AI Daily briefing. I'm ed sharp, thanks for joining me today.

2
00:00:05.480 --> 00:00:08.960
<v Speaker 1>Hardware wars Apple versus open Ai and the new silicon race.

3
00:00:09.960 --> 00:00:12.839
<v Speaker 1>Apple has sued open Ai. The claim is that open

4
00:00:12.880 --> 00:00:17.359
<v Speaker 1>ai recruited former Apple executives and extracted confidential hardware design

5
00:00:17.559 --> 00:00:21.640
<v Speaker 1>and manufacturing information. That's not a partnership dispute. That's an

6
00:00:21.640 --> 00:00:25.320
<v Speaker 1>allegation of industrial espionage, and it signals that the competition

7
00:00:25.399 --> 00:00:28.719
<v Speaker 1>between these two companies has moved well past product rivalry

8
00:00:28.960 --> 00:00:33.399
<v Speaker 1>into legal warfare. The important distinction here is timing. Apple

9
00:00:33.439 --> 00:00:37.759
<v Speaker 1>and OpenAI were partners until recently integrating chat GPT directly

10
00:00:37.840 --> 00:00:41.280
<v Speaker 1>into Apple devices. What changed is that both companies are

11
00:00:41.280 --> 00:00:44.439
<v Speaker 1>now competing for the same consumer AI hardware market. The

12
00:00:44.520 --> 00:00:48.320
<v Speaker 1>lawsuit is the moment that rivalry became formal. The signal

13
00:00:48.359 --> 00:00:51.479
<v Speaker 1>is that trade secrets in chip design and device manufacturing

14
00:00:51.679 --> 00:00:55.079
<v Speaker 1>are now the most contested assets in AI. The hardware

15
00:00:55.159 --> 00:00:58.520
<v Speaker 1>race isn't just Apple versus OpenAI. Meta is moving its

16
00:00:58.520 --> 00:01:02.320
<v Speaker 1>customer Iris chip into productction by September, a timeline designed

17
00:01:02.320 --> 00:01:06.000
<v Speaker 1>to double its internal compute capacity. This is vertical integration.

18
00:01:06.120 --> 00:01:09.680
<v Speaker 1>In practice, Meta is choosing to own its silicon rather

19
00:01:09.719 --> 00:01:12.840
<v Speaker 1>than depend on in Nvidia or anyone else. The strategic

20
00:01:12.879 --> 00:01:16.920
<v Speaker 1>logic is straightforward. Whoever controls the hardware controls the cost curve.

21
00:01:17.400 --> 00:01:20.480
<v Speaker 1>That matters because the companies building their own chips are

22
00:01:20.480 --> 00:01:25.159
<v Speaker 1>insulating themselves from supply chain shocks, an external pricing pressure meta.

23
00:01:25.239 --> 00:01:28.959
<v Speaker 1>Joining Apple and Google in custom silicon development isn't a footnote.

24
00:01:29.359 --> 00:01:32.640
<v Speaker 1>It's a structural shift in how big tech competes. Open

25
00:01:32.680 --> 00:01:36.480
<v Speaker 1>Ai launched GPT five point six this week along with

26
00:01:36.560 --> 00:01:40.400
<v Speaker 1>a chat GPT work desktop app, after completing a formal

27
00:01:40.519 --> 00:01:44.200
<v Speaker 1>US South cybersecurity review. That second part is the more

28
00:01:44.200 --> 00:01:48.599
<v Speaker 1>consequential detail. Government approval is now a precondition for frontier

29
00:01:48.599 --> 00:01:53.560
<v Speaker 1>model access. This isn't informal consultation. It's a regulatory checkpoint

30
00:01:53.599 --> 00:01:57.719
<v Speaker 1>embedded in the release cycle. The implication is clear, model

31
00:01:57.760 --> 00:02:01.799
<v Speaker 1>release velocity is now partially determined by government review timelines,

32
00:02:02.159 --> 00:02:04.920
<v Speaker 1>and there's no published guidance on how long those reviews

33
00:02:04.959 --> 00:02:08.639
<v Speaker 1>will take. For open ai and its competitors, that's an

34
00:02:08.719 --> 00:02:13.240
<v Speaker 1>unknown variable sitting inside every future product roadmap. In Munich,

35
00:02:13.360 --> 00:02:16.080
<v Speaker 1>defense ai start up Helsing closed a one point eight

36
00:02:16.120 --> 00:02:19.400
<v Speaker 1>billion dollar funding round at an eighteen billion dollar valuation.

37
00:02:20.039 --> 00:02:24.240
<v Speaker 1>That's Europe's largest defense tech rays. Helsing builds autonomous drone

38
00:02:24.280 --> 00:02:27.560
<v Speaker 1>and battlefield intelligence systems, and its valuation is now a

39
00:02:27.599 --> 00:02:31.360
<v Speaker 1>benchmark for the entire defense AI category. The key context

40
00:02:31.400 --> 00:02:35.439
<v Speaker 1>is doctrine. The Pentagon has been shifting toward attributable drone systems,

41
00:02:35.800 --> 00:02:40.080
<v Speaker 1>meaning cheaper, expendable platforms that trade precision for mass and autonomy.

42
00:02:40.479 --> 00:02:43.840
<v Speaker 1>That doctrine creates a procurement market, and Helsing is positioned

43
00:02:43.879 --> 00:02:47.439
<v Speaker 1>directly inside it. An eighteen billion dollar valuation in this

44
00:02:47.479 --> 00:02:50.439
<v Speaker 1>sector tells you where institutional money thinks. The next decade

45
00:02:50.439 --> 00:02:54.879
<v Speaker 1>of defense spending lands. Two major semiconductor expansions confirmed this week.

46
00:02:55.319 --> 00:02:58.000
<v Speaker 1>Intel is committing five billion euros to expand its chip

47
00:02:58.000 --> 00:03:02.319
<v Speaker 1>production facility in Lexlip Island. Samsung is accelerating its Korean

48
00:03:02.400 --> 00:03:05.240
<v Speaker 1>plant timeline by two years as part of a one

49
00:03:05.280 --> 00:03:08.719
<v Speaker 1>point three to five trillion dollar national AI manufacturing corridor.

50
00:03:09.199 --> 00:03:13.199
<v Speaker 1>Both moves reflect the same underlying logic. Governments and companies

51
00:03:13.199 --> 00:03:16.960
<v Speaker 1>are treating domestic chip capacity as a strategic necessity, not

52
00:03:17.039 --> 00:03:21.919
<v Speaker 1>a commercial preference. Europe reducing reliance on Asian manufacturing and

53
00:03:22.000 --> 00:03:27.000
<v Speaker 1>South Korea competing directly with TSMC aren't isolated decisions. They're

54
00:03:27.039 --> 00:03:30.479
<v Speaker 1>part of a coordinated fragmentation of the global semiconductor supply

55
00:03:30.599 --> 00:03:34.800
<v Speaker 1>chain into regional blocks. Consider this one day of Iran

56
00:03:34.919 --> 00:03:38.240
<v Speaker 1>US self tension pushed oil prices higher and triggered sharp

57
00:03:38.240 --> 00:03:42.319
<v Speaker 1>declines across AI and semiconductor stocks. That reaction speed tells

58
00:03:42.319 --> 00:03:45.879
<v Speaker 1>you something. Energy costs are now a recognized material risk

59
00:03:45.960 --> 00:03:50.360
<v Speaker 1>inside the AI infrastructure investment thesis, not an afterthought. The

60
00:03:50.400 --> 00:03:53.439
<v Speaker 1>build out of data centers at scale requires enormous power,

61
00:03:53.800 --> 00:03:56.800
<v Speaker 1>and any geopolitical shock that moves energy prices can move

62
00:03:56.879 --> 00:04:01.800
<v Speaker 1>valuations almost immediately. The unresolved is whether this week's pullback

63
00:04:01.919 --> 00:04:04.520
<v Speaker 1>was a one day correction or an early signal of

64
00:04:04.520 --> 00:04:08.560
<v Speaker 1>sustained capital flight. That's worth watching closely over the next

65
00:04:08.599 --> 00:04:11.319
<v Speaker 1>few weeks. Pull back and the through line across all

66
00:04:11.360 --> 00:04:15.319
<v Speaker 1>of this is consistent. The competition in AI has relocated

67
00:04:15.319 --> 00:04:22.040
<v Speaker 1>from software capabilities to physical infrastructure. Chips, manufacturing facilities, energy supply,

68
00:04:22.399 --> 00:04:26.279
<v Speaker 1>trade secrets, and government approvals are now the terrain. That's

69
00:04:26.279 --> 00:04:28.360
<v Speaker 1>a different kind of race than the one being run

70
00:04:28.399 --> 00:04:31.839
<v Speaker 1>two years ago. The near term watch points are straightforward.

71
00:04:32.160 --> 00:04:36.439
<v Speaker 1>Will Apple's lawsuit expose meaningful IP and discovery? How quickly

72
00:04:36.439 --> 00:04:40.319
<v Speaker 1>will U S south cyber security reviews process future model releases?

73
00:04:40.959 --> 00:04:44.560
<v Speaker 1>And can Sansum actually hit an accelerated chip production timeline

74
00:04:44.600 --> 00:04:48.600
<v Speaker 1>without running into electricity and labor constraints. Those three questions

75
00:04:48.600 --> 00:04:50.759
<v Speaker 1>will tell us more about where this industry is heading

76
00:04:50.920 --> 00:04:54.720
<v Speaker 1>than any benchmark score. Thanks for listening. This podcast was

77
00:04:54.720 --> 00:04:56.199
<v Speaker 1>built using AI technology.