WEBVTT -  Oracle's Force Majeure

0:00:02.529 --> 0:00:04.210
<v Speaker 1>Bloomberg Audio Studios.

0:00:04.670 --> 0:00:05.370
<v Speaker 2>Podcasts.

0:00:05.690 --> 0:00:11.550
<v Speaker 1>Radio. News. You're listening to Bloomberg Business Week with Carol

0:00:11.590 --> 0:00:14.790
<v Speaker 1>Masser and Tim Stenevek on Bloomberg Radio.

0:00:15.950 --> 0:00:19.950
<v Speaker 3>Hey, keep an eye on shares of Akamai because it

0:00:20.020 --> 0:00:23.620
<v Speaker 3>is up about 17.5% here in the African market. Yeah,

0:00:23.680 --> 0:00:25.720
<v Speaker 3>that's why we need to just talk about it a

0:00:25.739 --> 0:00:30.840
<v Speaker 3>little bit because Anthropic has signed an $ 11. 6 billion seven-year

0:00:30.860 --> 0:00:34.960
<v Speaker 3>contract with Akamai Technologies for computing power, adding to the

0:00:35.000 --> 0:00:37.220
<v Speaker 3>AI developer's growing list of data center deals. I think

0:00:37.240 --> 0:00:40.570
<v Speaker 3>what's interesting is Akamai is going to supply Anthropic access

0:00:40.590 --> 0:00:44.290
<v Speaker 3>to central processing units. We're talking about CPUs. Has issued

0:00:44.330 --> 0:00:46.710
<v Speaker 3>a warrant that gives Anthropic the right to buy Series

0:00:46.750 --> 0:00:50.070
<v Speaker 3>B shares that converts it to common stock. Talk about

0:00:50.610 --> 0:00:53.290
<v Speaker 3>quite a deal. And the contract represents the largest deal

0:00:53.800 --> 0:00:57.470
<v Speaker 3>in Akamai's history, with the company estimating that its CapEx spending,

0:00:57.510 --> 0:01:00.990
<v Speaker 3>its capital spending, tied to this one agreement, will total

0:01:01.490 --> 0:01:02.730
<v Speaker 3>about $ 5. 5 billion.

0:01:03.530 --> 0:01:04.340
<v Speaker 1>Fortunately... Talk.

0:01:04.380 --> 0:01:05.119
<v Speaker 3>About something that might.

0:01:05.140 --> 0:01:10.500
<v Speaker 1>Fortunately, Anurag Rana is here. He's Bloomberg Intelligence Senior Technology

0:01:10.540 --> 0:01:14.630
<v Speaker 1>Analyst for... Bloomberg Intelligence, as I mentioned. Check out his

0:01:14.670 --> 0:01:19.149
<v Speaker 1>research on the terminal. He joins us from our Chicago bureau. Anurag,

0:01:19.890 --> 0:01:26.840
<v Speaker 1>remind everybody where Akamai technology plays in the AI infrastructure world.

0:01:28.410 --> 0:01:30.310
<v Speaker 2>Actually, I wasn't even sure they were playing in that

0:01:30.350 --> 0:01:33.110
<v Speaker 2>world because before that, I thought they were doing something else.

0:01:33.410 --> 0:01:36.080
<v Speaker 1>Well, I know them from the Netflix days many years ago.

0:01:36.090 --> 0:01:40.400
<v Speaker 2>I know. But it seems like if you have free

0:01:40.440 --> 0:01:43.420
<v Speaker 2>data center capacity, you can go out and sell it

0:01:43.459 --> 0:01:45.980
<v Speaker 2>to others. Maybe I'll go see if I have computing

0:01:46.000 --> 0:01:48.500
<v Speaker 2>capacity at home. I can sell it to any of

0:01:48.520 --> 0:01:49.700
<v Speaker 2>these frontier labs as well.

0:01:50.600 --> 0:01:55.020
<v Speaker 1>So the story is about all of these companies going

0:01:55.040 --> 0:01:58.190
<v Speaker 1>and trying to find capacity wherever they can. And Akamai

0:01:58.230 --> 0:01:59.130
<v Speaker 1>has that capacity.

0:02:00.330 --> 0:02:02.580
<v Speaker 2>That's what I see. To be honest with you, before

0:02:02.670 --> 0:02:04.520
<v Speaker 2>I was coming here, I didn't even know this thing

0:02:04.560 --> 0:02:07.480
<v Speaker 2>was existing, frankly, that these guys do this business. But

0:02:07.680 --> 0:02:10.120
<v Speaker 2>that's not what it looks like right now, that if

0:02:10.160 --> 0:02:13.620
<v Speaker 2>you are looking for any way to make money, GPU

0:02:13.660 --> 0:02:16.460
<v Speaker 2>pricing is so high that this may be a good

0:02:16.500 --> 0:02:18.720
<v Speaker 2>business for everybody to get in, not just GPU, but

0:02:19.260 --> 0:02:20.740
<v Speaker 2>CPU pricing as well, obviously.

0:02:20.940 --> 0:02:23.150
<v Speaker 3>You know, I'm looking at the FA page for Akamai,

0:02:23.210 --> 0:02:32.460
<v Speaker 3>and so they had $ 4. 2 billion revenue in 2025. $ 11. 6

0:02:28.800 --> 0:02:32.500
<v Speaker 3>billion seven-year contract with Akamai, signed by Anthropic. They're going

0:02:32.520 --> 0:02:36.030
<v Speaker 3>to spend some $ 5. 5 billion, it says, in CapEx. How

0:02:36.040 --> 0:02:38.310
<v Speaker 3>does this kind of change? I know it's early. We're

0:02:38.330 --> 0:02:40.330
<v Speaker 3>throwing this at you just as it's been crossing.

0:02:40.770 --> 0:02:41.210
<v Speaker 1>But does it.

0:02:41.250 --> 0:02:43.930
<v Speaker 3>Change the balance sheet a lot for this company?

0:02:45.150 --> 0:02:47.570
<v Speaker 2>A lot depends on it. So the question is, when

0:02:47.610 --> 0:02:50.350
<v Speaker 2>you look at something like this, we are here to

0:02:50.389 --> 0:02:53.810
<v Speaker 2>talk originally about Oracle, but these deals, when you look

0:02:53.889 --> 0:02:57.760
<v Speaker 2>at it, are structured on a particular contract that they got.

0:02:58.160 --> 0:03:01.500
<v Speaker 2>So it's possible when you are looking at a specific

0:03:01.560 --> 0:03:05.520
<v Speaker 2>contract with a specific provider, you will have some backstops

0:03:05.600 --> 0:03:09.079
<v Speaker 2>as well, whether from the chip provider, whether it's NVIDIA

0:03:09.120 --> 0:03:11.760
<v Speaker 2>or somebody else. So the question really is, when they

0:03:11.800 --> 0:03:14.600
<v Speaker 2>go out in the market to look at debt for that,

0:03:14.940 --> 0:03:17.200
<v Speaker 2>what kind of covenants would be, what kind of backing

0:03:17.220 --> 0:03:19.960
<v Speaker 2>there is. So there is a lot of ifs and buts,

0:03:20.650 --> 0:03:24.120
<v Speaker 2>but all of these deals need to be financed one

0:03:24.160 --> 0:03:24.750
<v Speaker 2>way or the other.

0:03:25.210 --> 0:03:27.090
<v Speaker 3>Can I just ask you, though, is it another sign

0:03:27.110 --> 0:03:31.710
<v Speaker 3>of circular financing? Because this story, the whole idea, you've

0:03:31.750 --> 0:03:35.890
<v Speaker 3>also got Akamai said, Part of the warrant issued to

0:03:35.950 --> 0:03:39.210
<v Speaker 3>Anthropic represents about 2% of Akamai's common stock, but it

0:03:39.250 --> 0:03:42.850
<v Speaker 3>gives an opportunity for Anthropic to invest in this company. Again,

0:03:42.900 --> 0:03:43.840
<v Speaker 3>it's very close.

0:03:44.940 --> 0:03:46.760
<v Speaker 2>Yeah, but the thing is, think about it this way.

0:03:46.800 --> 0:03:50.060
<v Speaker 2>When you are a frontier lab and you're doing some

0:03:50.160 --> 0:03:53.520
<v Speaker 2>really high-end work and you need more capacity, you are

0:03:53.580 --> 0:03:55.840
<v Speaker 2>not getting it anywhere. I mean, you are tapped out

0:03:55.920 --> 0:03:59.020
<v Speaker 2>of any of the larger hyperscalers. You've already done what

0:03:59.060 --> 0:04:01.400
<v Speaker 2>you can with them. You know, you go and find

0:04:01.440 --> 0:04:05.570
<v Speaker 2>it for the neoclouds. After that, you know, You just

0:04:05.590 --> 0:04:07.940
<v Speaker 2>keep on going and trying to find wherever you can.

0:04:08.000 --> 0:04:11.040
<v Speaker 2>And one of the biggest complaints I think Sam Altman

0:04:11.060 --> 0:04:13.900
<v Speaker 2>has had against Anthropic has been that they haven't had

0:04:14.260 --> 0:04:19.160
<v Speaker 2>enough computing capacity. Now, the Anthropic side would be, well,

0:04:19.200 --> 0:04:21.600
<v Speaker 2>we go out and get capacity when we need it,

0:04:21.880 --> 0:04:24.440
<v Speaker 2>depending on the demand. So I think there's a lot

0:04:24.480 --> 0:04:27.120
<v Speaker 2>of that happening here. So I personally am not in

0:04:27.180 --> 0:04:30.920
<v Speaker 2>that circular financing camp. Only because the demand is coming

0:04:30.960 --> 0:04:34.620
<v Speaker 2>from customers, you're not manufacturing demand on the fly here.

0:04:35.480 --> 0:04:37.640
<v Speaker 1>Okay, should we talk about Oracle and a force majeure?

0:04:37.960 --> 0:04:38.180
<v Speaker 3>Yes.

0:04:38.339 --> 0:04:40.960
<v Speaker 1>Okay, this was one of the most notable decliners, I

0:04:41.000 --> 0:04:42.640
<v Speaker 1>would argue, and there's so much interest around it. If

0:04:42.980 --> 0:04:45.760
<v Speaker 1>you go to the last eight hours in the Bloomberg Terminal,

0:04:45.779 --> 0:04:47.979
<v Speaker 1>this is the most read story. Oracle sending a notice

0:04:48.020 --> 0:04:53.170
<v Speaker 1>citing force majeure to the developer of a massive data

0:04:53.190 --> 0:04:55.710
<v Speaker 1>center being built in New Mexico. This developer is a

0:04:56.370 --> 0:04:58.270
<v Speaker 1>unit of Blue Owl, and its attempt to put off

0:04:58.310 --> 0:05:01.609
<v Speaker 1>payments if the data center gets derailed and fails to

0:05:01.630 --> 0:05:04.510
<v Speaker 1>come online is as planned, this according to people familiar

0:05:05.010 --> 0:05:07.510
<v Speaker 1>with the matter. Is this a story about private credit?

0:05:07.710 --> 0:05:09.650
<v Speaker 1>Is it a story about circular financing? Is this story

0:05:09.670 --> 0:05:14.560
<v Speaker 1>about Oracle? Is it a story about the AI infrastructure spend?

0:05:14.570 --> 0:05:14.960
<v Speaker 3>What is it?

0:05:16.040 --> 0:05:19.480
<v Speaker 2>So Oracle addressed this issue very clearly on their conference call.

0:05:19.500 --> 0:05:22.680
<v Speaker 2>They basically talked about and said, yeah, they are building

0:05:22.760 --> 0:05:26.380
<v Speaker 2>multiple data centers in multiple locations. you would always get

0:05:26.420 --> 0:05:28.860
<v Speaker 2>into a situation where some of them are running on

0:05:28.880 --> 0:05:31.880
<v Speaker 2>schedule and some of them may be slightly behind. And

0:05:31.920 --> 0:05:34.760
<v Speaker 2>this New Mexico data center were asked specifically on the

0:05:34.800 --> 0:05:37.860
<v Speaker 2>conference call and said, well, what happens if this thing

0:05:37.900 --> 0:05:40.880
<v Speaker 2>doesn't come online as scheduled? Would you have an impact

0:05:40.960 --> 0:05:45.020
<v Speaker 2>on your sales growth? And the CEO was very categorically

0:05:45.760 --> 0:05:47.839
<v Speaker 2>straight up and said, It's not going to have any

0:05:47.920 --> 0:05:52.120
<v Speaker 2>impact on this year's number because these kinds of redundancies,

0:05:52.190 --> 0:05:55.110
<v Speaker 2>they keep on plan. So all of these cloud vendors,

0:05:55.270 --> 0:05:58.630
<v Speaker 2>whether it's a new cloud or hyperscalers, are building multiple

0:05:58.670 --> 0:06:02.030
<v Speaker 2>data centers around the world. Just because of the opposition

0:06:02.070 --> 0:06:04.330
<v Speaker 2>that you're seeing right now, you're going to face some

0:06:04.350 --> 0:06:07.310
<v Speaker 2>of this stuff. So in our view, this is a

0:06:07.440 --> 0:06:10.979
<v Speaker 2>bigger story that there could be some delays out there.

0:06:11.220 --> 0:06:12.960
<v Speaker 2>But the big question is, are you going to see

0:06:12.980 --> 0:06:15.299
<v Speaker 2>an acceleration in delays or this is a one-off?

0:06:18.040 --> 0:06:19.940
<v Speaker 3>So don't read it as a problem. Like you said,

0:06:19.980 --> 0:06:21.530
<v Speaker 3>if we start to see other companies, because it just

0:06:21.550 --> 0:06:25.880
<v Speaker 3>feels like, could they possibly walk away from this completely?

0:06:26.300 --> 0:06:29.130
<v Speaker 2>Would they? Oh, somebody will come and get it. That's

0:06:29.180 --> 0:06:32.720
<v Speaker 2>not an issue. I mean, if they don't have, I mean,

0:06:32.760 --> 0:06:36.640
<v Speaker 2>you just heard Anthropic, you know, looking anywhere for capacity.

0:06:36.680 --> 0:06:39.400
<v Speaker 2>So I'm not worried about finding a tenant for this.

0:06:39.460 --> 0:06:42.780
<v Speaker 2>The question is, will the local regulators let it go through?

0:06:42.800 --> 0:06:45.210
<v Speaker 2>I mean, the biggest issue now that we are seeing

0:06:45.250 --> 0:06:48.770
<v Speaker 2>is the backlash from local communities. That's really the bigger problem.

0:06:49.670 --> 0:06:54.650
<v Speaker 1>Oracle's down about 57%. From last year's highs, it was

0:06:54.810 --> 0:06:56.630
<v Speaker 1>sort of the poster child or has been the poster

0:06:56.650 --> 0:07:01.230
<v Speaker 1>child about the debt story when it comes to this

0:07:01.270 --> 0:07:03.859
<v Speaker 1>type of spend. Does this add to that story?

0:07:05.000 --> 0:07:07.680
<v Speaker 2>So it is definitely one of them is because they

0:07:07.700 --> 0:07:09.400
<v Speaker 2>already have a lot of debt on their balance sheet.

0:07:09.720 --> 0:07:12.840
<v Speaker 2>The second story is also from OpenAI because OpenAI has

0:07:12.880 --> 0:07:15.060
<v Speaker 2>made a lot of commitments to the, you know, you

0:07:15.080 --> 0:07:18.429
<v Speaker 2>could say computing commitments that we discussed just now. And

0:07:18.530 --> 0:07:20.830
<v Speaker 2>one of their biggest one is with Oracle, you know,

0:07:20.890 --> 0:07:23.470
<v Speaker 2>north of $ 300 billion or so when they first signed

0:07:23.510 --> 0:07:26.230
<v Speaker 2>it as part of the Stargate project. Now, what happens

0:07:26.290 --> 0:07:29.340
<v Speaker 2>over there is what if, you know, they see some

0:07:29.380 --> 0:07:34.460
<v Speaker 2>slowdown in frontier model training? Oracle gets impacted with that stuff. So, again,

0:07:34.700 --> 0:07:37.280
<v Speaker 2>not so much a near-term problem, but it could be

0:07:37.320 --> 0:07:39.250
<v Speaker 2>something that happens down the road. So there are two

0:07:39.320 --> 0:07:42.930
<v Speaker 2>things that are, you could say, two questions people have

0:07:43.030 --> 0:07:45.950
<v Speaker 2>around Oracle right now. Okay.

0:07:46.130 --> 0:07:47.890
<v Speaker 3>And what's interesting, too, is kind of I feel like

0:07:47.930 --> 0:07:51.150
<v Speaker 3>the power element. You know, they talk about the access

0:07:51.190 --> 0:07:54.450
<v Speaker 3>to meeting power commitments as we continue to see strains

0:07:54.510 --> 0:07:56.120
<v Speaker 3>on grades and so on and so forth. And it'll

0:07:56.260 --> 0:07:59.760
<v Speaker 3>be interesting. Interesting to see if for any reason this

0:07:59.800 --> 0:08:03.100
<v Speaker 3>kind of broadens out even more. What?

0:08:03.380 --> 0:08:07.190
<v Speaker 1>No, we have a lot to... I want to thank Anurag,

0:08:07.200 --> 0:08:11.210
<v Speaker 1>Senior Technology Analyst for Bloomberg Intelligence. We were planning to

0:08:11.250 --> 0:08:14.090
<v Speaker 1>talk Oracle. We got to talk about Akamai and Anthropic, too.

0:08:14.110 --> 0:08:14.670
<v Speaker 3>Two for one.

0:08:15.010 --> 0:08:15.510
<v Speaker 1>Yeah, we do.