1
00:00:04,590 --> 00:00:07,829
Trump signs 30 day voluntary AI model review EO.

2
00:00:08,229 --> 00:00:10,399
Labs now face a quiet choice on timing.

3
00:00:10,600 --> 00:00:14,649
OpenAI can absorb the delay because its release calendar already has slack.

4
00:00:14,949 --> 00:00:17,670
Anthropic runs tighter cycles and will likely need to

5
00:00:17,670 --> 00:00:20,739
duplicate its safety team just for the government channel.

6
00:00:21,030 --> 00:00:24,770
The real pressure lands on smaller frontier players who lack the headcount.

7
00:00:25,000 --> 00:00:28,409
Expect at least one to ship without review inside 9 months,

8
00:00:28,629 --> 00:00:32,490
testing whether voluntary turns into a hard gate for federal cloud access.

9
00:00:33,139 --> 00:00:36,450
Microsoft launches always on Scout autopilot agent.

10
00:00:36,740 --> 00:00:39,279
Background agents change the cost equation fast.

11
00:00:39,779 --> 00:00:42,049
Scout keeps a session alive across teams,

12
00:00:42,220 --> 00:00:43,459
Outlook and SharePoint,

13
00:00:43,659 --> 00:00:47,360
which means every user now burns GPU minutes even when idle.

14
00:00:47,580 --> 00:00:48,220
Most CIOs.

15
00:00:48,419 --> 00:00:51,200
will notice the new line item on their next Azure bill.

16
00:00:51,479 --> 00:00:54,720
OpenAI will have to decide whether to ship its own always on

17
00:00:54,720 --> 00:00:58,290
runtime or lose the seat to whatever ships default in Windows.

18
00:00:58,580 --> 00:01:00,860
The margin hit lands first on third party

19
00:01:00,860 --> 00:01:03,500
co-pilots that still wait for an explicit trigger.

20
00:01:04,349 --> 00:01:07,800
Microsoft unveils 7 new in-house MAI models.

21
00:01:07,959 --> 00:01:10,300
7 models at once signals the end of

22
00:01:10,300 --> 00:01:13,690
Microsoft treating OpenAI as its sole frontier supplier.

23
00:01:13,940 --> 00:01:17,239
The specialized ones in transcription and voice will move first

24
00:01:17,440 --> 00:01:19,540
because they slot directly into existing office.

25
00:01:19,764 --> 00:01:20,514
Pipelines.

26
00:01:20,754 --> 00:01:23,735
OpenAI loses the default path on those workloads.

27
00:01:23,934 --> 00:01:27,694
Expect the next contract renewal to carry lower volume commitments as a result.

28
00:01:28,175 --> 00:01:31,385
Smaller rivals building vertical agents now have a second cloud

29
00:01:31,385 --> 00:01:34,995
native option that ships with enterprise compliance already attached.

30
00:01:35,440 --> 00:01:39,160
Nvidia debuts RTX Spark ARM AI PC chips.

31
00:01:39,480 --> 00:01:42,239
One petaflop in a laptop flips the assumption that

32
00:01:42,239 --> 00:01:44,919
personal agents need any cloud round trip at all.

33
00:01:45,160 --> 00:01:47,400
Enterprise security teams will push for local

34
00:01:47,400 --> 00:01:50,180
execution on anything involving internal data.

35
00:01:50,470 --> 00:01:54,599
That shift cuts directly into open AI and anthropic inference revenue.

36
00:01:54,959 --> 00:01:58,889
Nvidia gains the default silicon for every Windows agent deployment.

37
00:01:59,129 --> 00:02:02,769
AMD and Intel now have 6 quarters to ship competitive arm

38
00:02:02,769 --> 00:02:07,169
parts or accept a permanent backseat in the AIPC refresh cycle.

39
00:02:07,610 --> 00:02:12,789
Anthropic files confidential IPO at $965 billion valuation.

40
00:02:13,169 --> 00:02:16,910
Anthropic's move puts every remaining private AI lab on the clock.

41
00:02:17,369 --> 00:02:20,809
OpenAI now has to decide whether to file within 90 days

42
00:02:20,809 --> 00:02:24,009
or watch its best people chase the new public stock options.

43
00:02:24,360 --> 00:02:27,110
The second order effect hits inference costs.

44
00:02:27,360 --> 00:02:30,440
Public markets will demand margin expansion that only comes

45
00:02:30,440 --> 00:02:33,270
from hiking API rates six months after the bell.

46
00:02:33,600 --> 00:02:35,720
Smaller labs lose the ability to compete

47
00:02:35,720 --> 00:02:38,100
on price without similar liquidity events.

48
00:02:38,309 --> 00:02:41,710
Expect rate card changes right after earnings season begins.

49
00:02:42,039 --> 00:02:45,399
Nvidia introduces Vera CPU for agentic compute.

50
00:02:45,639 --> 00:02:49,479
X86 vendors just lost their default slot in the agent rack.

51
00:02:49,839 --> 00:02:51,479
Vera's memory bandwidth advantage.

52
00:02:51,604 --> 00:02:55,755
Compounds when agents run continuous loops instead of one-off inferences.

53
00:02:56,044 --> 00:02:59,164
The performance delta shows up fastest in multi-turn agent

54
00:02:59,164 --> 00:03:03,184
sessions that X86 memory subsystems were never sized for.

55
00:03:03,445 --> 00:03:07,085
Intel and AMD must now either license similar high bandwidth

56
00:03:07,085 --> 00:03:10,714
tech or accept shrinking design wins in the inference tier.

57
00:03:11,005 --> 00:03:15,785
Cloud providers who bet on X86 cycles now face mismatched roadmaps.

58
00:03:15,914 --> 00:03:18,684
Expect design wins to flip within two quarters.

59
00:03:19,149 --> 00:03:23,820
Nvidia advances physical AI with Cosmos 3 and GR 00T.

60
00:03:24,270 --> 00:03:27,710
Simulation fidelity now determines who ships robots first.

61
00:03:28,029 --> 00:03:30,830
Cosmos 3 cuts the synthetic data gap enough that

62
00:03:30,830 --> 00:03:34,110
humanoid startups can skip two full hardware iterations.

63
00:03:34,429 --> 00:03:39,750
Figure and Boston Dynamics must either integrate the GR 00T reference stack or

64
00:03:39,750 --> 00:03:43,770
explain to backers why their custom controllers still justify the extra years.

65
00:03:44,070 --> 00:03:46,979
The second order squeeze lands on sensor makers whose inner

66
00:03:47,110 --> 00:03:49,809
Faces get standardized inside the Nvidia platform.

67
00:03:50,309 --> 00:03:54,070
Watch contract negotiations with LiDAR suppliers tighten once

68
00:03:54,070 --> 00:03:56,779
reference designs lock in camera first stacks.

69
00:03:57,169 --> 00:04:00,369
Microsoft reveals Majorana 2 quantum processor.

70
00:04:00,550 --> 00:04:02,910
Error rates drop enough that hybrid quantum

71
00:04:02,910 --> 00:04:05,830
classical workflows finally clear the prototype stage.

72
00:04:06,149 --> 00:04:07,949
Google and IBM must now defend

73
00:04:07,949 --> 00:04:11,130
their superconducting bets against a topological alternative

74
00:04:11,470 --> 00:04:14,869
that Microsoft claims avoids the dilution problem entirely.

75
00:04:15,100 --> 00:04:19,500
The 2nd order impact hits cloud HPC teams whose Q priorities

76
00:04:19,500 --> 00:04:23,899
get reordered once quantum nodes move from experimental to reserved capacity.

77
00:04:24,260 --> 00:04:26,660
Chemical simulation workloads migrate first.

78
00:04:26,989 --> 00:04:29,380
Pharma partners running molecular modeling get the

79
00:04:29,380 --> 00:04:31,920
first Azure quantum credits next fiscal year.

80
00:04:32,339 --> 00:04:34,970
Nvidia launches secure agent workspaces.

81
00:04:35,260 --> 00:04:38,100
Secure isolation layers on AI factories raise the

82
00:04:38,100 --> 00:04:40,649
bar for anyone handling regulated data flows.

83
00:04:40,940 --> 00:04:43,519
The move locks hardware choices for the next cycle

84
00:04:43,670 --> 00:04:47,839
because only Nvidia's stack ships the certified confidential compute path today.

85
00:04:48,309 --> 00:04:48,779
Except.

86
00:04:49,045 --> 00:04:50,885
Market vendors to lose deals once

87
00:04:50,885 --> 00:04:53,524
compliance teams start requiring these workspaces.

88
00:04:53,964 --> 00:04:56,725
Salesforce integration teams will feel the pinch first when

89
00:04:56,725 --> 00:05:00,144
agent workflows hit financial services by Q2 next year.

90
00:05:00,445 --> 00:05:02,445
Smaller players without access to the same

91
00:05:02,445 --> 00:05:05,484
silicon will watch their enterprise pipeline shrink.

92
00:05:06,579 --> 00:05:09,820
Taiwan president pledges AI resources at CompuTex.

93
00:05:10,059 --> 00:05:12,660
Energy allocations announced today will decide which

94
00:05:12,660 --> 00:05:15,769
AI clusters actually come online in 2027.

95
00:05:16,019 --> 00:05:19,500
The government support masks a tighter constraint on grid capacity

96
00:05:19,500 --> 00:05:23,049
that foreign chip makers must now negotiate directly with Thaipower.

97
00:05:23,299 --> 00:05:25,850
Watch how this plays out for TSMC's capacity

98
00:05:25,850 --> 00:05:28,820
commitments once power purchase agreements get signed.

99
00:05:29,089 --> 00:05:32,054
Intel's foundry push in the region faces an Unexpected

100
00:05:32,054 --> 00:05:35,515
delay risk if local priorities shift toward domestic models.

101
00:05:35,744 --> 00:05:39,734
AMD must accelerate alternative sites or accept slower ramps.

102
00:05:40,054 --> 00:05:43,204
Microsoft pushes Project Solara agent-first vision.

103
00:05:43,494 --> 00:05:47,174
Agent-first computing flips the resource model so that background tasks

104
00:05:47,174 --> 00:05:50,244
now consume the majority of cycles on any given device.

105
00:05:50,575 --> 00:05:53,654
Microsoft is betting Enterprise SLAs will absorb the

106
00:05:53,654 --> 00:05:56,505
added latency costs without pushing back on pricing.

107
00:05:56,760 --> 00:05:59,440
The 2nd order effect hits 3rd party app stores

108
00:05:59,440 --> 00:06:03,720
hardest when autonomous agents bypass traditional install flows entirely.

109
00:06:04,000 --> 00:06:07,000
OpenAI faces pressure to embed similar controls in

110
00:06:07,000 --> 00:06:10,070
their desktop clients before Windows defaults take hold.

111
00:06:10,390 --> 00:06:13,910
This forces a choice on pricing models within the next release cycle.

112
00:06:14,200 --> 00:06:17,429
Microsoft debuts RTX Spark Dev Box hardware.

113
00:06:17,709 --> 00:06:20,600
Developers get early access to agent testing hardware

114
00:06:20,600 --> 00:06:24,040
that already includes the isolation primitives Nvidia announced.

115
00:06:24,290 --> 00:06:27,170
The real constraint shifts to software frameworks that must

116
00:06:27,170 --> 00:06:30,640
now target this specific platform to hit performance targets.

117
00:06:31,000 --> 00:06:33,510
Apple loses ground in the AI dev community

118
00:06:33,640 --> 00:06:37,910
if these boxes become the default for agent iteration inside Windows shops.

119
00:06:38,130 --> 00:06:40,809
Expect training data pipelines to standardize around

120
00:06:40,809 --> 00:06:43,119
the Spark stack faster than anticipated.

121
00:06:43,440 --> 00:06:47,589
Competitors will see their developer ecosystems erode within 12 months.