1
00:00:04,469 --> 00:00:07,940
FERC orders fast track grid links for AI data centers.

2
00:00:08,180 --> 00:00:11,390
Grid operators now face a 60-day clock to rewrite

3
00:00:11,390 --> 00:00:15,380
interconnection cues already holding up 3 hyperscale campuses.

4
00:00:15,590 --> 00:00:17,780
The order prioritizes AI loads,

5
00:00:17,790 --> 00:00:20,180
but leaves transmission upgrades unfunded,

6
00:00:20,389 --> 00:00:22,989
which means utilities will push those costs down.

7
00:00:23,229 --> 00:00:24,680
Stream into higher rates.

8
00:00:24,899 --> 00:00:28,260
Expect mid-market data center projects to slip another 18

9
00:00:28,260 --> 00:00:31,290
months while the big three lock in firm capacity.

10
00:00:31,569 --> 00:00:34,439
Smaller cloud providers lose their window entirely.

11
00:00:34,700 --> 00:00:38,299
This forces Google and Microsoft to accelerate on-site

12
00:00:38,299 --> 00:00:41,959
generation deals by mid-2026 or absorb the delays.

13
00:00:42,619 --> 00:00:45,909
Amazon drops nearly finished Sam Altman biopic.

14
00:00:46,139 --> 00:00:50,099
The poll comes right after Amazon committed $50 billion to OpenAI,

15
00:00:50,580 --> 00:00:53,290
suggesting the studio wants distance from any Altman

16
00:00:53,290 --> 00:00:56,009
spotlight that could turn awkward during sales cycles.

17
00:00:56,259 --> 00:00:57,819
Production sat weeks from wrap,

18
00:00:57,939 --> 00:01:00,209
yet the project moves without explanation.

19
00:01:00,500 --> 00:01:04,580
This timing signals Amazon's enterprise team now treats OpenAI

20
00:01:04,580 --> 00:01:07,360
leadership as a liability rather than an asset.

21
00:01:07,580 --> 00:01:12,099
Insiders expect the buyer to surface within 30 days with a very different framing.

22
00:01:13,449 --> 00:01:17,319
Chat GPT market share falls below 50% for the first time.

23
00:01:17,569 --> 00:01:19,819
The drop landed hardest in North America,

24
00:01:20,010 --> 00:01:23,110
where Gemini picked up 9 points in just 5 months.

25
00:01:23,370 --> 00:01:26,480
Enterprise pilots that once defaulted to chat GPT

26
00:01:26,730 --> 00:01:30,489
now split traffic across 3 models to test reliability under low.

27
00:01:31,035 --> 00:01:35,574
This fragments the data advantage OpenAI needs for continued fine tuning.

28
00:01:35,815 --> 00:01:40,014
Expect the company to push usage-based enterprise contracts by September

29
00:01:40,154 --> 00:01:43,904
or watch seat retention erode further against bundled competitors.

30
00:01:44,275 --> 00:01:47,955
Teams now demand multi-model clauses in every new deployment.

31
00:01:48,680 --> 00:01:52,989
OpenAI GPT 5.6 leaks point to late June launch.

32
00:01:53,190 --> 00:01:55,690
Yakub Petraki called the gains agentic,

33
00:01:55,930 --> 00:02:00,059
which points straight at workflow displacement rather than benchmark chasing.

34
00:02:00,250 --> 00:02:03,489
Early Access Partners already report 30% fewer

35
00:02:03,489 --> 00:02:05,790
tokens needed for the same task chains.

36
00:02:05,970 --> 00:02:08,639
That efficiency lands first in coding agents,

37
00:02:08,649 --> 00:02:11,589
where margins compress once inference costs fall.

38
00:02:11,770 --> 00:02:13,889
This forces Anthropic to either match the

39
00:02:13,889 --> 00:02:16,929
API price or pivot to enterprise-only contracts.

40
00:02:17,216 --> 00:02:18,805
Before the public release hits.

41
00:02:19,055 --> 00:02:23,296
Prediction markets price launch odds above 70% inside 4 weeks.

42
00:02:23,615 --> 00:02:27,406
AWS Bedrock Agent Corps harness those generally available.

43
00:02:27,656 --> 00:02:31,055
Production deployment just became a configuration checkbox rather

44
00:02:31,055 --> 00:02:33,326
than a months-long project for most teams.

45
00:02:33,615 --> 00:02:36,755
The isolated micro VMs create an audit boundary

46
00:02:36,856 --> 00:02:39,615
that lets companies trace exactly where an agent

47
00:02:39,615 --> 00:02:43,386
decision originated without relying on third-party monitoring tools.

48
00:02:43,615 --> 00:02:45,376
This forces OpenAI and.

49
00:02:45,522 --> 00:02:49,201
Tropic to ship comparable isolation layers by early next year,

50
00:02:49,382 --> 00:02:51,582
or watch Enterprise customers route agent

51
00:02:51,582 --> 00:02:54,932
traffic exclusively through AWS for compliance.

52
00:02:55,261 --> 00:02:58,421
Smaller developers gain immediate access to enterprise grade

53
00:02:58,421 --> 00:03:01,171
guardrails they previously had to build themselves.

54
00:03:01,421 --> 00:03:05,572
Gemini 3.5 Pro nears release amid capability leaks.

55
00:03:05,891 --> 00:03:08,462
Leaks around the 2M context window miss

56
00:03:08,462 --> 00:03:11,602
how this changes RAG costs for agent builders

57
00:03:11,742 --> 00:03:13,822
who no longer need constant chunking.

58
00:03:14,070 --> 00:03:16,270
The real pressure lands on OpenAI,

59
00:03:16,470 --> 00:03:19,789
which must either extend its own context or lose long

60
00:03:19,789 --> 00:03:22,050
form enterprise deals by the end of the year.

61
00:03:22,350 --> 00:03:26,410
Smaller labs with fixed context limits face a steeper climb to stay

62
00:03:26,410 --> 00:03:30,850
relevant in agent workflows that now favor massive single pass reasoning.

63
00:03:31,110 --> 00:03:35,110
Expect benchmark reports to shift once the model hits production traffic.

64
00:03:35,350 --> 00:03:39,029
Shopify opens universal commerce protocol to all developers.

65
00:03:39,309 --> 00:03:41,289
Merchants gain agent discovery without paying.

66
00:03:41,638 --> 00:03:44,548
For additional marketplace features that used to gate access,

67
00:03:44,968 --> 00:03:49,307
the protocol forces Amazon to either publish its own open commerce standard

68
00:03:49,488 --> 00:03:53,947
or see agent-driven sales migrate to Shopify stores within 18 months.

69
00:03:54,128 --> 00:03:57,848
Existing checkout optimization vendors lose their wedge once carts

70
00:03:57,848 --> 00:04:01,348
initialize directly from agent prompts across the open network.

71
00:04:01,688 --> 00:04:05,477
Transaction fees become the next battleground as volume shifts.

72
00:04:05,927 --> 00:04:08,888
Coinbase for agents enables autonomous crypto trades.

73
00:04:09,315 --> 00:04:12,186
Defined limits create a new control plane that banks

74
00:04:12,186 --> 00:04:14,615
have so far ignored in their own offerings.

75
00:04:14,945 --> 00:04:17,786
Traditional financial institutions must now expose

76
00:04:17,786 --> 00:04:20,265
similar agent endpoints or watch crypto

77
00:04:20,265 --> 00:04:24,246
transaction share move to Coinbase hosted agents over the next year.

78
00:04:24,545 --> 00:04:29,066
Compliance teams at every exchange face fresh questions about who owns the

79
00:04:29,066 --> 00:04:33,175
decision when an agent initiates a transfer outside human review cycles.

80
00:04:33,425 --> 00:04:36,346
The custody question will dominate regulatory filings.

81
00:04:36,523 --> 00:04:38,664
Long before trading volume numbers do.

82
00:04:38,884 --> 00:04:43,074
OpenAI Maria AI completes first autonomous drug discovery.

83
00:04:43,364 --> 00:04:47,764
Contract research organizations still routing every new target through human

84
00:04:47,764 --> 00:04:51,843
review boards now face timelines their current headcount cannot support.

85
00:04:52,084 --> 00:04:53,993
The human validation step remains,

86
00:04:54,204 --> 00:04:58,014
yet it now sits at the end of an automated pipeline rather than the beginning.

87
00:04:58,243 --> 00:05:00,523
Maria closed the loop from hypothesis to

88
00:05:00,523 --> 00:05:03,803
validated compound without waiting on committee calendars.

89
00:05:04,141 --> 00:05:05,902
By Q3 2026,

90
00:05:06,101 --> 00:05:11,221
any CRO lacking an integrated agent will watch project costs climb 30%

91
00:05:11,221 --> 00:05:14,851
while faster competitors lock in the next round of pharma partnerships.

92
00:05:15,101 --> 00:05:18,532
B2B content strategies shift for AI agent buyers.

93
00:05:18,821 --> 00:05:22,022
Marketing budgets are already migrating toward assets agents

94
00:05:22,022 --> 00:05:24,221
can parse instead of humans who skim.

95
00:05:24,502 --> 00:05:27,782
Agencies still optimizing for human scroll depth will lose

96
00:05:27,782 --> 00:05:31,261
briefs once agents start filtering on evidence density alone.

97
00:05:31,519 --> 00:05:35,779
The shift turns every case study into potential model fine tuning fodder.

98
00:05:36,000 --> 00:05:40,070
Structured data sets embedded in every asset become the new ranking factor.

99
00:05:40,320 --> 00:05:41,299
Within 18 months,

100
00:05:41,399 --> 00:05:43,440
the firms that treat content as training

101
00:05:43,440 --> 00:05:46,850
material will capture most agent-mediated evaluations,

102
00:05:47,119 --> 00:05:48,940
sidelining traditional formats.

103
00:05:49,200 --> 00:05:53,640
GLM 5.2 undercuts frontier models on price and coding.

104
00:05:53,959 --> 00:05:56,579
Teams running persistent agent fleets no longer need.

105
00:05:56,790 --> 00:06:00,670
To ration frontier models for only the highest value coding jobs.

106
00:06:00,829 --> 00:06:02,929
The gap is no longer raw performance,

107
00:06:03,070 --> 00:06:05,100
but cost per 1000 agent runs.

108
00:06:05,470 --> 00:06:11,970
GLM 5.2's parity at lower cost forces the first real margin pressure on API pricing.

109
00:06:12,269 --> 00:06:14,149
Anthropic must either match the new

110
00:06:14,149 --> 00:06:17,070
rate or watch mid-market developers shift entire

111
00:06:17,070 --> 00:06:19,549
workflows to the cheaper provider for everything

112
00:06:19,549 --> 00:06:21,970
except the most complex enterprise agents.