# AI Builds Itself as the Spending Boom Accelerates

An episode of [Bloomberg Tech](https://invisiblenetwork.ai/podcast/1611767434), 18 Sep 2026, 44 min.

Audio: https://podtrac.com/pts/redirect.mp3/traffic.omny.fm/d/clips/e73c998e-6e60-432f-8610-ae210140c5b1/41764a4f-fc64-4e11-89ba-ae7c0030ab5e/2ccdf5c5-a092-43b3-8b25-b4c9010fba5a/audio.mp3?utm_source=Podcast&in_playlist=9caafc41-289c-4115-995d-ae7c0030ab75

Bloomberg’s Ed Ludlow breaks down Anthropic's finding that its Claude chatbot is leading 26% of its AI research and development, even as the global debate around AI heats up. Plus, SoftBank is closing out the week with nearly $21 billion in potential fresh borrowings as it builds out its AI financing capacity, and Crusoe CEO Chase Lochmiller discusses the race to build the infrastructure powering AI as the company closes nearly $4 billion in fresh funding. See omnystudio.com/listener for privacy information.

## Transcript

[0:02] Speaker 1: Bloomberg Audio Studios.
[0:04] Speaker 2: Podcasts.
[0:05] Speaker 1: Radio. News.
[0:13] Speaker 3: Bloomberg Tech is live from the heart of Silicon Valley
[0:17] Speaker 3: with Ed Ludlow in San Francisco.
[0:23] Speaker 4: This is Bloomberg Tech coming up as the global debate
[0:26] Speaker 4: around AI heats up and Fropic finds its claw chatbots
[0:29] Speaker 4: leading 26% of its AI research and development, plus SoftBank's
[0:34] Speaker 4: closing out the week with nearly $ 21 billion in potential
[0:37] Speaker 4: fresh borrowing as it builds out its AI financing capacity.
[0:40] Speaker 4: and Crusoe CEO Chase Lockmiller joins us on the race
[0:43] Speaker 4: to build the infrastructure powering AI. The company closes nearly
[0:47] Speaker 4: $ 4 billion in fresh funding. We're going to end the
[0:50] Speaker 4: week much as we started it, talking about Anthropic. Anthropic
[0:54] Speaker 4: says its Claude Chatbot drives more than a quarter of
[0:57] Speaker 4: the company's research and development work for AI and collaborates
[1:01] Speaker 4: with staffers for about 90% of their work. It's the
[1:04] Speaker 4: clearest indication yet that AI can help speed up development
[1:09] Speaker 4: of future models. Bloomberg's Shireen Ghaffari has the data. And Shireen,
[1:12] Speaker 4: just start by explaining the data that Anthropic disclosed.
[1:17] Speaker 5: That's right.
[1:17] Speaker 6: So Anthropic is basically measuring how much its own researchers,
[1:23] Speaker 6: its own R & D department is using AI in
[1:26] Speaker 6: its work to create new AI models. And what they
[1:29] Speaker 6: found is that right now, Claude is leading, Claude meaning
[1:33] Speaker 6: Anthropic's AI, 26% of that AI.
[1:38] Speaker 7: R & D work.
[1:40] Speaker 6: Now, that's not the same as AI doing AI. Everything
[1:43] Speaker 6: by itself, but that is significant. It's one of the
[1:47] Speaker 6: few kind of numbers that we have quantifying what many
[1:51] Speaker 6: labs like Anthropic Open AI are now moving toward, which
[1:55] Speaker 6: is the AI essentially feeding the AI to keep improving itself.
[2:00] Speaker 4: We're at the end of a seven-day week, even though
[2:02] Speaker 4: it's Friday, because on Saturday... seven days ago, Dario Amodei
[2:07] Speaker 4: published his essay. And one of the things he highlighted
[2:10] Speaker 4: at the top for the rationale for doing it now
[2:12] Speaker 4: is recursive self-improvement, the idea that AI is making the
[2:17] Speaker 4: training of future models better. But do you have an
[2:20] Speaker 4: answer to that, Shireen, why the data that was published
[2:23] Speaker 4: yesterday came out when it did?
[2:26] Speaker 6: You know, I think we don't know if there was
[2:28] Speaker 6: any one specific reason, but I will say the backdrop
[2:30] Speaker 6: of all of this is while there is obviously huge
[2:35] Speaker 6: business opportunity in reaching AI that can improve itself with RSI,
[2:39] Speaker 6: there's also huge concern, anxiety, risk. And we're seeing AI
[2:44] Speaker 6: lab employees at Anthropic, at OpenAI, starting to raise concerns
[2:48] Speaker 6: publicly about this. We saw the viral note of Jacob Coxon,
[2:52] Speaker 6: a former Anthropic and employee researcher saying, saying he doesn't
[2:56] Speaker 6: think any of the labs are handling this responsibly right now.
[3:01] Speaker 6: And I think that there's a real call for these
[3:04] Speaker 6: companies to start quantifying and potentially slowing down by Dario
[3:09] Speaker 6: Amadei's own essays recognition the pace of this, even if
[3:14] Speaker 6: that means losing some business momentum.
[3:18] Speaker 4: Author of the Bloomberg Q & AI newsletter, to which
[3:21] Speaker 4: you should subscribe, Bloomberg's Shireen Ghaffari. Thank you very much indeed.
[3:24] Speaker 4: As AI starts helping build the next generation of AI,
[3:27] Speaker 4: warnings about losing control are growing louder. But critics say
[3:31] Speaker 4: the focus on existential risk is overshadowing problems that are
[3:35] Speaker 4: already here. It's the subject of Bloomberg, Rachel Metz and
[3:38] Speaker 4: Davey Alba, latest reporting and, frankly, great body of work
[3:44] Speaker 4: trying to cut through the noise of the week. Devi,
[3:47] Speaker 4: I'm going to start with you. Just summarize what our
[3:50] Speaker 4: story outlines and take us to the origin, which is
[3:54] Speaker 4: the Hugging Face incident. Yeah, absolutely.
[3:57] Speaker 8: So we did a deep dive into sort of the
[4:00] Speaker 8: Hugging Face incident as an example of how a lot
[4:06] Speaker 8: of the rhetoric that's been going around right now is
[4:09] Speaker 8: talking about how a failure in safety and in the
[4:14] Speaker 8: sandbox not being properly cordoned off, that there wasn't warnings.
[4:21] Speaker 8: There were warnings that the human employees themselves did not
[4:27] Speaker 8: take seriously. That is being framed as, look, we can't
[4:31] Speaker 8: contain this AI.
[4:32] Speaker 1: It's so dangerous.
[4:33] Speaker 7: It's going to escape.
[4:34] Speaker 8: And that being an inevitability and that just, you know,
[4:38] Speaker 8: sort of works to sidestep the responsibility that these companies
[4:42] Speaker 8: have to avoid safety failures.
[4:46] Speaker 4: Hey, Rachel, you know what's weird about AI? What's weird
[4:49] Speaker 4: about AI? What's weird? Humans, peeps, design the experiment and
[4:56] Speaker 4: design the safeguards. Like that's the takeaway I got reading
[4:59] Speaker 4: your piece. Remind everyone that AI came from humans. Run
[5:04] Speaker 4: with it.
[5:05] Speaker 9: Yes.
[5:07] Speaker 10: Yeah, I think that's something that I've been thinking about
[5:09] Speaker 10: for a while. Since we heard about the hugging face
[5:12] Speaker 10: inadvertent hack and talking to people about it, I mean,
[5:15] Speaker 10: this comes up quite a lot, that humans were very
[5:18] Speaker 10: much involved with this.
[5:19] Speaker 1: I mean, humans set goals for AI.
[5:22] Speaker 10: Humans set up the situation here in the sandbox where
[5:27] Speaker 10: the AI is working on things individually.
[5:30] Speaker 1: And humans are training. I mean, there are humans up
[5:32] Speaker 1: and down.
[5:33] Speaker 10: Humans also noticed some situations where they could have stepped
[5:36] Speaker 10: in during evaluations when the AI wasn't doing quite what
[5:39] Speaker 10: it should have been doing.
[5:40] Speaker 1: And did not.
[5:41] Speaker 10: And OpenAI has acknowledged all of these things. And to
[5:44] Speaker 10: its credit, it said quite a lot about this incident
[5:47] Speaker 10: and also what it's doing to change things over there.
[5:49] Speaker 10: But I think that that came up a lot with
[5:52] Speaker 10: this story, especially as we're talking to people ongoing, that
[5:58] Speaker 10: people have a lot to do with this situation.
[6:01] Speaker 1: It's not as simple as saying, oh, well, AI went
[6:03] Speaker 1: and did something.
[6:05] Speaker 4: Rachel, Davey, we have some breaking news. Gavin Newsom, the
[6:08] Speaker 4: governor of California, has issued an executive order on the
[6:11] Speaker 4: creation of an AI kill switch. The order accelerates what
[6:17] Speaker 4: he says will be independent oversight of AI. That story
[6:21] Speaker 4: just hitting the terminal right now. Davey, this is difficult
[6:24] Speaker 4: because it's fluid, right? It's just hitting the newswire. You
[6:29] Speaker 4: talk as well in your parallel newsletter about this idea
[6:32] Speaker 4: that there is... plenty of elements of human control. Hugging
[6:36] Speaker 4: Face was the case study for that. Take us there.
[6:40] Speaker 1: Yeah, absolutely.
[6:41] Speaker 8: And, you know, with regard to this AI kill switch,
[6:43] Speaker 8: we also cover in our story the language that federal
[6:48] Speaker 8: lawmakers have started to use to create policy that would affect,
[6:54] Speaker 8: you And one thing that we point out is a
[6:58] Speaker 8: lot of this sort of rhetoric of doom is migrating
[7:01] Speaker 8: from the AI industry into the rhetoric of these lawmakers.
[7:06] Speaker 8: So I would want to take a look at how
[7:09] Speaker 8: Newsom is framing this kill switch. One thing that the
[7:14] Speaker 8: researchers that we spoke to said was that when you
[7:19] Speaker 8: start from a place of doom, sort of this assumption
[7:23] Speaker 8: that AI is going to threaten all humanity, the policies
[7:28] Speaker 8: may not match the actual harms that AI is causing
[7:31] Speaker 8: right now. So when we're talking about misinformation, surveillance, these
[7:37] Speaker 8: things that are happening right now, that is what is
[7:42] Speaker 8: sort of getting lost in the focus on doom.
[7:47] Speaker 4: Rachel, on the show in the last 24 hours, there
[7:49] Speaker 4: have been plenty of voices. Let's take researcher Andrew Ng
[7:52] Speaker 4: or Databricksio Aligotzi saying this is way overstated. The risk
[7:57] Speaker 4: to human life, to their mind, is near zero. You're
[8:00] Speaker 4: up in Montreal right now, right? Who have you been
[8:02] Speaker 4: speaking to? What are you hearing on the kind of counter-argument,
[8:06] Speaker 4: the anti-doomerism school of thought?
[8:11] Speaker 10: Yeah, I mean, I think up here I'm hearing a
[8:13] Speaker 10: little bit of both. I mean, I spoke to Yoshua Bengio,
[8:18] Speaker 10: who is a legendary figure in AI and has long
[8:21] Speaker 10: talked about what he sees as very real harms that
[8:26] Speaker 10: may come from AI. And so he sees this situation,
[8:29] Speaker 10: the Hugging Face incident in particular, as real. When I
[8:32] Speaker 10: spoke to him yesterday, it was sort of a mix
[8:34] Speaker 10: of a human and machine problem, and he's very, very
[8:36] Speaker 10: concerned about what may happen going forward. On the other hand,
[8:39] Speaker 10: I've spoken to other people, such as the CEO of Cohere,
[8:43] Speaker 10: Aidan Gomez, also a very key figure in AI, and
[8:47] Speaker 10: he had.
[8:49] Speaker 1: A thought that was a little bit more like, look.
[8:51] Speaker 10: I don't think that this is going to kill people
[8:54] Speaker 10: or kill us en masse, but there are some really
[8:58] Speaker 10: real issues here, and those do need to be worked on.
[9:01] Speaker 10: And maybe slowing down certain things could be helpful, but
[9:05] Speaker 10: it's not a situation where we're at imminent risk of
[9:11] Speaker 10: all being extinct.
[9:14] Speaker 4: Bloomberg's Rachel Metz and Davey Alba, authors of the report
[9:17] Speaker 4: Anthropics Warning.
[9:20] Speaker 4: Of existential risk hijacks large AI debate. It's a must
[9:24] Speaker 4: read on Bloomberg. Thank you both very much. We are
[9:26] Speaker 4: going to keep this conversation going. Sara Arahi is the
[9:28] Speaker 4: Franklin Templeton portfolio manager and partner with Franklin Ventures and
[9:33] Speaker 4: joins us now. And Franklin Ventures, also an investor in Anthropic. Okay, Friday,
[9:39] Speaker 4: net net, where have we ended this week?
[9:42] Speaker 9: Nice ending, what Shireen just talked about, giving us data
[9:46] Speaker 9: about the 26%.
[9:49] Speaker 7: Agents in the R & D work.
[9:51] Speaker 9: I think that's a good ending versus the doomerism started
[9:54] Speaker 9: at the beginning of the week. I'm definitely in the
[9:56] Speaker 9: anti-doomerism camp. I think that I'm appreciative that these founders
[10:03] Speaker 9: are talking about guardrails. They're talking about making sure that
[10:06] Speaker 9: everything is safely done. But I also think we.
[10:11] Speaker 7: Shouldn't be slowing down our pace here.
[10:13] Speaker 9: And even if we did on the pre-training side or
[10:16] Speaker 9: the training side where there's a lot of expenses on
[10:18] Speaker 9: that front.
[10:19] Speaker 7: Inference is going to continue. And that means the spend
[10:23] Speaker 7: will continue.
[10:23] Speaker 9: And as a function of that, you're going to see,
[10:26] Speaker 9: you know, a number of companies that pulled back, you know,
[10:29] Speaker 9: stocks in the equity market pulled back on Monday with
[10:31] Speaker 9: fears of, oh, slow down spending.
[10:33] Speaker 7: That spending can't slow, especially.
[10:35] Speaker 4: This is why I'm grateful to have you here. You
[10:37] Speaker 4: are a public market investor and you are a private
[10:40] Speaker 4: market investor. And so, like, let's go to public markets
[10:42] Speaker 4: for a second. The idea was if the labs are
[10:45] Speaker 4: pacing at the frontier, and that pre-training or training work
[10:48] Speaker 4: is impacted, the question posed, does that mean a pullback
[10:51] Speaker 4: on the spend for the compute? The net conclusion on
[10:55] Speaker 4: that is.
[10:56] Speaker 9: No, because inference requires a lot of compute. You've seen
[11:00] Speaker 9: a number of deals that these companies have done the
[11:02] Speaker 9: last few months to secure compute deals. Do you think
[11:05] Speaker 9: they would have done that if they thought they were
[11:06] Speaker 9: going to pull back, if pacing was going to pull
[11:08] Speaker 9: back spend? We don't think that would have happened. And
[11:11] Speaker 9: then it's interesting because, yes, that narrative was being talked
[11:15] Speaker 9: about on Monday. I was at an industrials conference on Tuesday,
[11:19] Speaker 9: and at that conference, a number of these infrastructure companies
[11:22] Speaker 9: that are exposed to data center were saying, our businesses
[11:25] Speaker 9: are diversified.
[11:26] Speaker 7: We have non-data center.
[11:27] Speaker 9: Exposure as well because they saw the reaction in the market.
[11:30] Speaker 7: But on Wednesday, some of that had already cleared up.
[11:33] Speaker 7: Both businesses are going to continue to grow.
[11:36] Speaker 4: It's really hard to frame. There are lots of different
[11:37] Speaker 4: camps on the right course of action. All of them
[11:41] Speaker 4: accept that the capabilities of the models are improving and
[11:46] Speaker 4: outpacing the rate at which alignment and safety keep up.
[11:49] Speaker 4: That seems to be a point of agreement. What Dario
[11:53] Speaker 4: Amodei advocated for in his essay on Saturday was independent evaluation,
[11:57] Speaker 4: but also the idea that antitrust waivers are appropriate so
[12:01] Speaker 4: that that collection of frontier labs can work together. How
[12:05] Speaker 4: do you feel about that?
[12:06] Speaker 9: Yeah, look, I don't know if they really want widespread
[12:09] Speaker 9: regulation and federal regulation.
[12:11] Speaker 4: You don't think that they want that?
[12:12] Speaker 7: I don't think they want that.
[12:13] Speaker 9: I do think that they want good actors here and
[12:18] Speaker 9: every company to put guardrails in place and make sure
[12:21] Speaker 9: that everything is done in a safe way, in a sandbox.
[12:24] Speaker 9: And we've seen examples of that, but that hasn't happened, unfortunately.
[12:27] Speaker 4: Which is the school of thought, right? You're the frontier labs,
[12:29] Speaker 4: you're the experts, you should have a handle on this.
[12:32] Speaker 9: Yes.
[12:32] Speaker 4: And don't release the model if it's not safe.
[12:34] Speaker 7: Exactly.
[12:35] Speaker 9: On top of that, though, some are talking about the
[12:38] Speaker 9: fact that there are a number of security products that
[12:41] Speaker 9: are going to be launched very soon. And so we
[12:44] Speaker 9: can also protect you. You're going to see some of
[12:47] Speaker 9: those product launches come again. A couple of weeks ago
[12:50] Speaker 9: at a tech conference, the entire theme at that tech
[12:52] Speaker 9: conference was cybersecurity. Right. Whether it's the public companies exposed
[12:56] Speaker 9: to this or the venture community was talking about what
[13:00] Speaker 9: some of the frontier labs are going to start to
[13:01] Speaker 9: offer and a number.
[13:02] Speaker 4: Of private companies. So can you help us understand this?
[13:04] Speaker 4: I think you heard Ali Ghodsi, the Dedebrick CEO, on
[13:06] Speaker 4: the show yesterday. He went cybersecurity straight away. Andrew Ng
[13:09] Speaker 4: also brought it up. Loads of people brought it up.
[13:11] Speaker 4: But the idea is that it means very strong cybersecurity
[13:15] Speaker 4: platforms are more important than ever.
[13:17] Speaker 7: They're incredibly important.
[13:18] Speaker 9: Not only are they important, but also just pacing these models.
[13:22] Speaker 9: You haven't seen widespread adoption of all these models right
[13:26] Speaker 9: now and outside of Anthropic talking about it right here,
[13:30] Speaker 9: right now, right?
[13:30] Speaker 4: Yes.
[13:31] Speaker 7: So corporations are.
[13:32] Speaker 9: Going to start, you know, you have GPT-5, GPT-6, you know,
[13:36] Speaker 9: Fable coming to market. Companies haven't, knowledge workers are not
[13:39] Speaker 9: even used to using these right now. Just with the
[13:42] Speaker 9: launch of Muse, Instinct, some of these agents that consumers
[13:45] Speaker 9: are starting to work with, I think that will bring
[13:48] Speaker 9: some adoption on the consumer side, and consumers are going
[13:51] Speaker 9: to start asking for that in the workplace, just like
[13:54] Speaker 9: you saw with the iPhone. It took a while for
[13:57] Speaker 9: consumers to move from BlackBerrys to iPhones, right? But at
[14:01] Speaker 9: the end of the day, it was they started seeing
[14:03] Speaker 9: what kind of benefit they get from this, and then
[14:05] Speaker 9: they started asking for it in their workplace.
[14:08] Speaker 4: Sarah, I appreciate there's a limit to what you can
[14:11] Speaker 4: say as an existing investor in Anthropic, but a question
[14:14] Speaker 4: posed by the market is, Are these safety concerns the
[14:18] Speaker 4: real reason why OpenAI will not do an IPO until 2027,
[14:24] Speaker 4: which Sam Altman said in an interview? Actually, there was
[14:27] Speaker 4: plenty of reporting that the IPO would be 2027 anyway.
[14:31] Speaker 4: But are the safety concerns a real reason for the
[14:33] Speaker 4: market to slow down and say, like, we need to
[14:36] Speaker 4: stop funding this in private or public markets?
[14:39] Speaker 7: Again, I don't think so.
[14:42] Speaker 9: We're starting to see some of these benefits, the productivity
[14:45] Speaker 9: gains from these models, the benefits that society will see,
[14:49] Speaker 9: not the... 10% or wherever that number comes from in
[14:53] Speaker 9: terms of doomerism.
[14:54] Speaker 4: The economic upside.
[14:56] Speaker 7: The upside.
[14:57] Speaker 9: We're starting to see that diffuse across a variety of industries.
[15:00] Speaker 9: You've seen the market start to broaden as a result
[15:03] Speaker 9: of that. So we shouldn't stop funding. And I think
[15:06] Speaker 9: IPO timelines, I think some of them were already being
[15:09] Speaker 9: pushed out to 2027. And I think we're going to
[15:11] Speaker 9: start to see some of these come through.
[15:13] Speaker 4: Sarah O'Rourke here, Franklin Templeton. It's really great to have
[15:15] Speaker 4: you back on the show.
[15:16] Speaker 7: Thank you.
[15:16] Speaker 4: Happy Friday, by the way.
[15:18] Speaker 7: Happy Friday.
[15:18] Speaker 4: Coming up, SoftBank is closing out the week with fresh
[15:21] Speaker 4: firepower to back its AI bets and maybe on the
[15:24] Speaker 4: road to more. We're going to go talk about the
[15:26] Speaker 4: Japanese conglomerate's massive new debt hole. That's next. This is
[15:29] Speaker 4: Bloomberg Tech. Let's take a look at today's big number,
[15:41] Speaker 4: $ 21 billion. That's how much fresh debt SoftBank's potentially closing
[15:45] Speaker 4: the week out with. Bloomberg's Peter Elstrom, who leads our
[15:48] Speaker 4: coverage of Asian technology companies, joins us now. It is
[15:51] Speaker 4: an ongoing story. It's an almost daily story. What's new
[15:54] Speaker 4: about this SoftBank borrowing?
[15:57] Speaker 13: Yeah, it's worth slowing down here to just pause and
[16:01] Speaker 13: understand exactly what Masayoshi Son is doing with SoftBank at
[16:04] Speaker 13: this point. As you pointed out, he's got more than
[16:06] Speaker 13: $ 20 billion in new loans that he's securing in order
[16:10] Speaker 13: to make this investment in OpenAI. He said he's going
[16:13] Speaker 13: to invest about $ 65 billion. So this is three different
[16:16] Speaker 13: slices of money that he's coming up with. One, he's
[16:19] Speaker 13: increasing the amount of money that he's borrowing against his
[16:22] Speaker 13: arm holding. He holds about 90%. of Arm, the chip maker.
[16:26] Speaker 13: He's also going to borrow money from Apollo. He's increasing
[16:29] Speaker 13: the size there. And he's also getting a loan from
[16:32] Speaker 13: 20 banks that he's securing all together. Add it up,
[16:35] Speaker 13: and he's borrowing about 45 to 50 billion dollars out
[16:38] Speaker 13: of the 65 billion dollars that he's gonna put into OpenAI.
[16:41] Speaker 13: And that's not the end of the leverage. He's actually
[16:43] Speaker 13: borrowing money against that OpenAI stake too. That's another 10
[16:46] Speaker 13: billion dollars or so. So this is sort of Masa
[16:49] Speaker 13: doing Masa things, building a lot of leverage on top
[16:52] Speaker 13: of these startup bets that he's making.
[16:55] Speaker 4: So I think a lot of people might pose the
[16:58] Speaker 4: reasonable question, what is SoftBank? Like, okay, we know that
[17:02] Speaker 4: there is Masayoshi San, there's SoftBank, they have exposure to OpenAI,
[17:06] Speaker 4: just explain the basics.
[17:08] Speaker 13: Well, SoftBank has been through many iterations since Masayoshi Son
[17:12] Speaker 13: founded the company. It's done everything from distribute software in
[17:16] Speaker 13: the past. It ran tech trade shows. It had a
[17:19] Speaker 13: bunch of tech magazines in the past. But right now,
[17:21] Speaker 13: it's primarily an investment vehicle sitting on top of a
[17:24] Speaker 13: telecom operation that has thrown off a bunch of cash.
[17:27] Speaker 4: In the past.
[17:28] Speaker 13: SoftBank Corp, he spun off a few years ago. So
[17:30] Speaker 13: it's an independent company, but it still generates a lot
[17:33] Speaker 13: of the profit. And he's been using that money to
[17:35] Speaker 13: make a bunch of different bets. He, of course, started
[17:37] Speaker 13: up the Vision Fund that fatally put money into WeWork.
[17:41] Speaker 13: But it's also now making other investments. And really, the
[17:44] Speaker 13: big bet that he's making is OpenAI. He's talked about
[17:47] Speaker 13: how important he thinks AI is going to be in
[17:49] Speaker 13: the future. And he believes that Sam Altman is going
[17:51] Speaker 13: to help lead that. So he's Put so much money
[17:53] Speaker 13: into the company, even people inside a soft bank have
[17:56] Speaker 13: raised questions about how risky that is and if he's
[17:59] Speaker 13: concentrated his bets too much on a company that now,
[18:01] Speaker 13: of course, as we've talked about on this show, they're
[18:03] Speaker 13: pushing back their IPO. So you've got him borrowing more
[18:06] Speaker 13: money as interest rates are rising. for an investment in
[18:10] Speaker 13: a company who's pushing back its IPO, pushing back that
[18:13] Speaker 13: day when he's going to be able to have a
[18:14] Speaker 13: liquid publicly traded stock to use as collateral.
[18:18] Speaker 6: Right.
[18:19] Speaker 4: Bloomberg's Peter Elstrom, happy Friday to you. Thank you very
[18:21] Speaker 4: much indeed. Now coming up, one of the year's biggest
[18:24] Speaker 4: AI trades is reversing as software stocks stage a comeback
[18:29] Speaker 4: and chips slide. Breakdown what's driving the turnaround. This is
[18:32] Speaker 4: Bloomberg Tech. Wall Street Week goes to India to see
[18:46] Speaker 4: what the fastest growing large economy offers to investors from
[18:50] Speaker 4: the tech center of Bangalore to Genome Valley in Hyderabad.
[18:54] Speaker 4: The economy is growing on the back of its tech sector,
[18:56] Speaker 4: but its future hinges on its ability to keep up
[18:59] Speaker 4: in the global AI race.
[19:02] Speaker 14: So I think it's both a challenge and an opportunity.
[19:05] Speaker 14: I think the challenge lies in the fact that India
[19:08] Speaker 14: has had great strength in IT services and providing a
[19:14] Speaker 14: lot of functions, white collar services at very attractive costs.
[19:20] Speaker 14: Some of that will get replaced by AI. And so
[19:24] Speaker 14: I think the India bears say, ah, that's going to happen.
[19:29] Speaker 14: I think what they fail to recognize is that These
[19:33] Speaker 14: companies will evolve. We own a large stake in a
[19:36] Speaker 14: business called Emphasis, an IT services company, and they have
[19:39] Speaker 14: pivoted their business to focus on AI deployment because changing
[19:45] Speaker 14: companies utilizing AI is also going to require an enormous
[19:49] Speaker 14: amount of manpower. I also think they underestimate the willingness
[19:54] Speaker 14: of people in India to utilize the AI and what
[19:57] Speaker 14: it will do there. The creation of new companies there,
[20:00] Speaker 14: the creation of AI infrastructure.
[20:04] Speaker 15: If I look back two decades ago, we were like
[20:05] Speaker 15: a call center. That call center has gone to a
[20:08] Speaker 15: back office. The back office went to a mid office.
[20:12] Speaker 15: And the mid office now is literally a front office
[20:15] Speaker 15: because tech is really part of the business, right? And
[20:19] Speaker 15: that's why I feel, despite AI, we're doing very well.
[20:23] Speaker 15: The demand's very robust. And we look forward to a
[20:26] Speaker 15: really strong rest of the decade.
[20:29] Speaker 4: You can watch the full episode of Wall Street Week
[20:31] Speaker 4: today at 6 p.m. Eastern, 3 p.m. Pacific. Okay, software
[20:35] Speaker 4: stocks are making a comeback. Strong earnings and warnings about
[20:39] Speaker 4: the pace of AI development are reversing one of this
[20:42] Speaker 4: year's biggest trades, long chips, short software. Okay, this Friday's
[20:46] Speaker 4: session might not reflect that necessarily, but Bloomberg's Ryan Veselica
[20:49] Speaker 4: has one of the best read stories on the Bloomberg.
[20:52] Speaker 4: This happened in June and... Today, it's the same, right?
[20:56] Speaker 4: The software doomerism is also saspocalypse fading.
[21:01] Speaker 3: Yeah, there's been a real reversal of sentiment. The first
[21:04] Speaker 3: half of the year was really marked by this concern
[21:07] Speaker 3: that AI was going to come in and decimate the
[21:10] Speaker 3: legacy software industry, which is why we saw the chip
[21:12] Speaker 3: sector do so well and the software sector really just collapsed.
[21:16] Speaker 3: Since then, we have seen a bit of a reversal.
[21:19] Speaker 3: We've seen some narrowing of this performance gap. Most recently,
[21:23] Speaker 3: we had a very strong earnings season for the software group,
[21:26] Speaker 3: including a number of companies that were really seen as
[21:29] Speaker 3: being at risk of disruption. So companies like Salesforce, ServiceNow,
[21:35] Speaker 3: we saw some mixed results, mixed to positive results out
[21:38] Speaker 3: of Oracle.
[21:38] Speaker 4: Microsoft had a huge quarter.
[21:40] Speaker 3: So all of these things just helped reinforce the idea
[21:42] Speaker 3: that maybe the selling had gone too far. Maybe the
[21:46] Speaker 3: valuations had gotten washed out. Everything was being sold indiscriminately since,
[21:49] Speaker 3: and we've seen people show some signs of interest of
[21:52] Speaker 3: getting back into the sector.
[21:54] Speaker 4: Ryan, real quick, Netflix, its first sell rating in a
[21:57] Speaker 4: long time, and the stock's on track for its worst
[21:59] Speaker 4: year since 2022.
[22:02] Speaker 3: Yeah, Wells Fargo Securities downgraded the stock underweight. It's the
[22:05] Speaker 3: first sell rating the stock has seen since I think
[22:06] Speaker 3: late February. A lot of concern there about engagement trends.
[22:11] Speaker 3: The analyst said that there's no big water cooler hits,
[22:13] Speaker 3: no huge success for people to really be driving engagement
[22:17] Speaker 3: there as Netflix is instead focusing on other kinds of
[22:20] Speaker 3: content like gaming, like podcasts. It's viewing that not as
[22:23] Speaker 3: a great strategy, at least in the near term. So
[22:26] Speaker 3: we are seeing some more weakness in that stock there.
[22:28] Speaker 3: That's been a big underperformer this year.
[22:31] Speaker 4: Bloomberg's Ryan for Celica on software and on Netflix. Thank
[22:35] Speaker 4: you very much. Coming up, as Frontier Labs push to
[22:37] Speaker 4: work together, the antitrust debate heats up. Neil Chilson from
[22:41] Speaker 4: the Abundance Institute, former FTC chief technologist. This is a
[22:44] Speaker 4: really important conversation. Halftime, live from San Francisco, this is
[22:50] Speaker 4: Bloomberg Tech. Welcome back to Bloomberg Tech. A lot has
[23:05] Speaker 4: happened this week. And in the end, the Nasdaq 100,
[23:08] Speaker 4: a very tech-heavy index, is modestly higher over five days,
[23:12] Speaker 4: three-tenths of a percent. Started the week with calls to
[23:15] Speaker 4: pace at the frontier from the frontier labs in AI.
[23:18] Speaker 4: In the middle of the week, we had the Fed.
[23:20] Speaker 4: Yields have risen. Basically, the expectation is further Fed rate
[23:24] Speaker 4: hikes because of energy and other inflationary pressures. We also
[23:29] Speaker 4: continue to have disclosures from the Frontier Labs about models misbehaving.
[23:34] Speaker 4: And then by the time we ended the week, we
[23:36] Speaker 4: had some kind of, OK, let's all calm down a
[23:38] Speaker 4: little bit. The risk on an existential basis probably isn't
[23:42] Speaker 4: what we thought it was on Monday. Chip stocks have underperformed, though.
[23:45] Speaker 4: On a five-day basis, the Philadelphia Semiconductor Index, or SOX,
[23:49] Speaker 4: is down half a percentage point. Nothing too severe. The
[23:52] Speaker 4: question public market investors had was, was all this warning
[23:55] Speaker 4: about AI going to result in the market pulling back
[23:58] Speaker 4: because they were concerned that spending on AI infrastructure would
[24:01] Speaker 4: be impacted? No, is the conclusion so far. Okay, so
[24:05] Speaker 4: the week started with anthropic CEO Dario Amadei arguing the
[24:08] Speaker 4: frontier labs will need some antitrust exemptions in the U.S.,
[24:11] Speaker 4: so they can coordinate in specific areas, as well as
[24:14] Speaker 4: have cooperation with China. David Sachs, who's co-chair of the
[24:18] Speaker 4: President's Council of Advisors on Science and Technology, pushed back
[24:21] Speaker 4: on that idea.
[24:23] Speaker 16: You should go ahead and make your products safe without
[24:27] Speaker 16: demanding any concession from the government. You know, we don't
[24:30] Speaker 16: believe that you should be free from competition. I'm sure
[24:34] Speaker 16: every company in America would love to have that kind
[24:36] Speaker 16: of deal. You have a basic obligation to make your
[24:39] Speaker 16: products safe. Do not demand antitrust waivers or liability waivers
[24:43] Speaker 16: in exchange for that.
[24:45] Speaker 5: Honestly, it feels like they're bartering with us.
[24:48] Speaker 4: For more on this debate, Neil Chilson, head of AI
[24:51] Speaker 4: policy at the Abundance Institute and a former FTC chief
[24:54] Speaker 4: technologist is with us. When we think antitrust, we think
[24:58] Speaker 4: monopolistic or anti-competitive practices. The Frontier Labs are saying, Neil,
[25:02] Speaker 4: we need this waiver so that we can work together.
[25:05] Speaker 4: Your response?
[25:08] Speaker 11: So there are so many ways that companies can work
[25:11] Speaker 11: together that are not anti-competitive. But obviously, antitrust has a
[25:16] Speaker 11: big role in making sure that when companies collaborate, they're
[25:20] Speaker 11: not doing things that ultimately harm the process of competition
[25:24] Speaker 11: or consumers themselves. and the results that they expect from
[25:27] Speaker 11: a competitive environment. And so I think that there are,
[25:32] Speaker 11: like I said, if you're building safe products and you
[25:34] Speaker 11: need to build an industry standard or best practices to
[25:37] Speaker 11: do that, there's lots of ways that these companies can collaborate.
[25:41] Speaker 11: Most of all, they can try internally to build their
[25:45] Speaker 11: products safely, and actually they're obligated in many ways to
[25:49] Speaker 11: do that.
[25:50] Speaker 4: You are, it seems, in a school of thought shared
[25:53] Speaker 4: by lots of leaders in AI that at the end
[25:56] Speaker 4: of the day, the frontier labs should work this out
[25:58] Speaker 4: for themselves if they have command of the technology. Regulatory
[26:03] Speaker 4: capture is a concern. It's the idea that those frontier
[26:06] Speaker 4: labs help set the rules that's to their benefit. Your interpretations.
[26:12] Speaker 11: So regulatory capture is a huge problem. But what we
[26:15] Speaker 11: know from history is that cartels are fragile. The only
[26:19] Speaker 11: cartels that are durable are ones that are endorsed by
[26:22] Speaker 11: the government. And so I think that is the risk
[26:25] Speaker 11: that David Sachs is pointing to, that the types of
[26:28] Speaker 11: exemptions that they're seeking here could make a durable cartel
[26:32] Speaker 11: that would ultimately harm competition and harm the U.S.
[26:36] Speaker 4: Ability to stay at the frontier. Are existing laws... particularly
[26:43] Speaker 4: those around illegal action, sufficient right now, Neil, to your mind?
[26:47] Speaker 4: In other words, if a model does something that constitutes
[26:50] Speaker 4: a criminal offense, there is a safety net of the
[26:54] Speaker 4: law anyway.
[26:57] Speaker 11: I think that's right, especially in the liability space. Our
[27:01] Speaker 11: existing law says that if you're doing something, if these
[27:06] Speaker 11: companies are training models and they're not doing it safely enough,
[27:10] Speaker 11: and those models cause harm to third parties, I think
[27:13] Speaker 11: the law is pretty clear, especially FTC consumer protection law,
[27:18] Speaker 11: and I think many tort laws, that those companies will
[27:21] Speaker 11: be responsible for those harms. And I think that is
[27:23] Speaker 11: part of what's driving uh dario's call for pacing and
[27:27] Speaker 11: it's part of what um is making the industry say
[27:30] Speaker 11: hey we we need to make sure that we're we
[27:33] Speaker 11: ourselves are being cautious that we're creating a product and
[27:37] Speaker 11: that we're creating it in a way that isn't causing
[27:39] Speaker 11: harm to third parties.
[27:41] Speaker 4: You were formerly the chief technologist at the ftc right
[27:45] Speaker 4: could you take me inside the organization a week like
[27:48] Speaker 4: this how would the FTC be looking at, thinking about,
[27:52] Speaker 4: responding to what has been a rollercoaster of events?
[27:56] Speaker 11: Well, I think for sure they would be looking at
[27:59] Speaker 11: the various authorities that they have, both on the competition
[28:02] Speaker 11: side and as well on the consumer protection side. And
[28:05] Speaker 11: they'd be looking for unfair methods of competition. They'd be
[28:08] Speaker 11: looking for unfair or deceptive acts or practices that might
[28:11] Speaker 11: be occurring here. And the FTC in particular has a
[28:14] Speaker 11: long history of data security enforcement practices under its consumer protection,
[28:20] Speaker 11: unfair and deceptive acts and practices. And I think here
[28:23] Speaker 11: they would be looking for some key indicators that what
[28:27] Speaker 11: the companies are doing is not reasonable, it does not
[28:30] Speaker 11: match best practices, and that it could be easily avoided
[28:33] Speaker 11: by the companies with some extra precaution. And they will
[28:38] Speaker 11: be looking to see if there's harm that's caused there.
[28:40] Speaker 11: and whether or not the harms that are caused are
[28:44] Speaker 11: outweighed by the benefits to the specific practices that the
[28:47] Speaker 11: companies are doing.
[28:48] Speaker 5: So I think there's a lot for them to think about,
[28:50] Speaker 5: and they're doing it.
[28:52] Speaker 4: Neil, the harm that's imagined by some, not all some,
[28:56] Speaker 4: is that AI poses an existential threat to human life, right?
[29:02] Speaker 4: Even if you go back 10 days from the anthropic
[29:05] Speaker 4: research of the resigns, I mean, how does the FTC
[29:08] Speaker 4: account for that harm? potential harm to a consumer.
[29:12] Speaker 5: It's a great question.
[29:14] Speaker 11: Obviously, those types of speculative harms, one would hope in
[29:20] Speaker 11: many ways that those would show up as much smaller
[29:23] Speaker 11: types of incidents early on.
[29:25] Speaker 5: And that does seem to be the case here.
[29:26] Speaker 11: And so we're seeing incidents where there is not a
[29:31] Speaker 11: lot of harm that's happening, but where it raises some
[29:34] Speaker 11: red flags. And I think what that means is there's
[29:37] Speaker 11: time to look at the laws, to look at the practices,
[29:40] Speaker 11: in the industry and say like, hey, we need to
[29:41] Speaker 11: change course here.
[29:43] Speaker 5: We need to correct things.
[29:44] Speaker 11: How do we do that? Do we need some extra laws?
[29:47] Speaker 11: We need to see if there's gaps in the law
[29:49] Speaker 11: to fill. But I think we are getting warning signs
[29:52] Speaker 11: here that are the exact types that start this iterative
[29:56] Speaker 11: process of saying, hey, how can we do better and
[29:58] Speaker 11: And how can we avoid these bigger harms in the future?
[30:02] Speaker 4: You're at the Abundance Institute now, an organization who wants
[30:05] Speaker 4: technology to grow and thrive, right? There's a lot of
[30:09] Speaker 4: positivity around AI.
[30:10] Speaker 12: It's useful.
[30:11] Speaker 4: There is some economic upside. Does the FTC ever take
[30:15] Speaker 4: that into account? Again, drawing on your past experience and
[30:18] Speaker 4: your current research on how it's good for people.
[30:22] Speaker 5: Oh, absolutely.
[30:23] Speaker 11: I mean, I think that's a big part of the analysis,
[30:26] Speaker 11: both on the competition side where the question is, are
[30:30] Speaker 11: these practices creating consumer welfare? Are they improving consumer welfare?
[30:34] Speaker 11: And then on the consumer protection side where the question
[30:38] Speaker 11: in the unfairness analysis in particular is whether any harms
[30:42] Speaker 11: caused by a practice are outweighed.
[30:44] Speaker 4: By benefits.
[30:44] Speaker 5: And I think in the.
[30:46] Speaker 11: AI space in particular, there are enormous benefits. We're seeing
[30:49] Speaker 11: that in stock values, we're seeing that investment, and we're
[30:52] Speaker 11: seeing it in the huge amount of adoption and usage
[30:55] Speaker 11: that individuals are doing. I think we're nowhere near the
[30:58] Speaker 11: top of of the types of problems that a general
[31:01] Speaker 11: purpose AI system can help us with. And so I
[31:06] Speaker 11: think there are enormous benefits here. We need to mitigate
[31:08] Speaker 11: the harms, but we need to not lose sight of
[31:11] Speaker 11: the fact that that mitigation, we want to make sure
[31:14] Speaker 11: it doesn't prevent the best possible world of the benefits
[31:18] Speaker 11: that could come.
[31:19] Speaker 4: Neil Chilson of the Abundance Institute, formerly chief technologist at
[31:23] Speaker 4: the FTC. Thank you so much for your time on
[31:25] Speaker 4: the show. Now, coming up, Crusoe CEO Chase Lockmiller joins
[31:29] Speaker 4: us on the race to build the infrastructure that's powering AI.
[31:32] Speaker 4: The company's just closed out almost $ 4 billion in fresh funding.
[31:37] Speaker 4: Fascinating list of names and strategic backers. Looking forward to
[31:41] Speaker 4: that conversation next. This is Bloomberg Tech.
[31:49] Speaker 8: Thank you.
[32:04] Speaker 4: Crusoe closed the nearly $ 4 billion Series F funding round,
[32:07] Speaker 4: valuing the AI infrastructure company at almost $ 31 billion. The
[32:11] Speaker 4: new capital comes as Crusoe says it has more than
[32:14] Speaker 4: $ 140 billion in contracted value across its platform. It's also
[32:19] Speaker 4: adding board members, has a lot of other momentum. Crusoe
[32:21] Speaker 4: CEO Chase Lockmiller back with us on Bloomberg Tech. This
[32:24] Speaker 4: round has been in the works a while. It's a
[32:27] Speaker 4: big round, a vote of confidence. And I think the
[32:30] Speaker 4: participants are super interesting. as well, Chase. Let's talk about
[32:34] Speaker 4: the capital need, you know, what it is you're going
[32:36] Speaker 4: to be able to get going on now.
[32:38] Speaker 5: Absolutely. So, you know, thanks for having me on the show.
[32:41] Speaker 5: First of all, Ed, always great to see you.
[32:43] Speaker 2: You know, this round was an exciting moment in the
[32:46] Speaker 2: culmination of just an incredible amount of hard work for
[32:50] Speaker 2: many years across, you know, the Crusoe team.
[32:52] Speaker 5: And, you know, it's really.
[32:54] Speaker 2: uh, been a, been, um, an onus on, on, on
[32:57] Speaker 2: the company and the whole industry to really dig deep, uh, and,
[33:00] Speaker 2: and really work hard to, uh, you know, deliver the
[33:03] Speaker 2: infrastructure layer of intelligence. Uh, you know, the, the, the
[33:06] Speaker 2: capital needs are, are, are really tremendous for, for, for
[33:09] Speaker 2: what's happening. And, and, and because of that, because of that,
[33:12] Speaker 2: you know, and, and the, uh, the value creation potential is, is,
[33:16] Speaker 2: is really enormous in terms of you know, what AI
[33:19] Speaker 2: is going to do to transform the global economy. And
[33:22] Speaker 2: that's really why the capital is being driven to the sector.
[33:25] Speaker 4: Chase, people are familiar with Crusoe, probably because of the
[33:29] Speaker 4: relationship with Oracle, the relationship with OpenAI. But if we
[33:33] Speaker 4: really scale it back to simple terms, you help build
[33:36] Speaker 4: data centers, and you also have your own cloud computing
[33:39] Speaker 4: platform now. Could you just explain that split?
[33:42] Speaker 2: Yeah, so Crusoe is a vertically integrated AI infrastructure platform that,
[33:46] Speaker 2: you know, helps manage infrastructure from the electron to the token.
[33:49] Speaker 2: So this is both building, operating, and energizing these large-scale
[33:54] Speaker 2: AI factories, and then building software to operate these.
[33:57] Speaker 5: AI factories, which is the Crusoe Cloud platform.
[34:00] Speaker 2: So you really think about what we deliver to our
[34:02] Speaker 2: customers can be a data center, in the case of
[34:05] Speaker 2: how we've worked with Oracle and and Microsoft and a
[34:08] Speaker 2: number of other big customers. It can be a managed
[34:11] Speaker 2: compute cluster, which is what we're doing with folks like
[34:14] Speaker 2: Perplexity and Figure and a number of other key AI innovators.
[34:18] Speaker 2: And it can also be managed services and managed intelligence,
[34:22] Speaker 2: which is what we're doing with delivering our Crusoe managed
[34:24] Speaker 2: inference platform and some of our other services like serverless
[34:28] Speaker 2: fine tuning.
[34:30] Speaker 4: Chase, with those kind of three buckets you just outlined,
[34:33] Speaker 4: what's the kind of percentage of sales for each currently?
[34:36] Speaker 4: And how is Crusoe's business kind of shifting in terms
[34:40] Speaker 4: of the mainstay of what you do?
[34:42] Speaker 2: Well, you know, each of those different businesses, each of
[34:46] Speaker 2: those different ways of sort of. serving customers, they typically
[34:50] Speaker 2: have different durations. So when folks are making data center
[34:53] Speaker 2: commitments to us, we're typically working with very large tech
[34:57] Speaker 2: companies that have investment grade credit ratings and they're typically
[35:00] Speaker 2: committing to us for somewhere between 15 and 20 years.
[35:04] Speaker 2: They're multi-decade long commitments. When you think about the life
[35:07] Speaker 2: cycle of a GPU and the managed compute clusters, those
[35:10] Speaker 2: can range anywhere from a few months to five to
[35:14] Speaker 2: six years, but they typically don't go beyond that just
[35:16] Speaker 2: because that's sort of the life cycle of the chips.
[35:20] Speaker 2: And then when you think about managed inference and managed
[35:22] Speaker 2: AI services, those are more, you know, they're developer tools
[35:26] Speaker 2: or it's building in the developer ecosystem. So they're more
[35:30] Speaker 2: on demand. You know, there are commitments that folks are making,
[35:32] Speaker 2: but they're typically shorter term in nature as they're scaling
[35:37] Speaker 2: their AI applications and the actual utilization process.
[35:40] Speaker 5: Of these services.
[35:41] Speaker 2: So it sort of offers a broad spectrum of services
[35:45] Speaker 2: that we can offer the industry, as well as a
[35:48] Speaker 2: broad spectrum of contracts and revenue resources that we're receiving
[35:53] Speaker 2: as a business.
[35:54] Speaker 4: You've added three important names as board members, and looking
[35:58] Speaker 4: at the round, the size, the participants, a lot of
[36:02] Speaker 4: people asking if this is kind of a precursor to you, Chase,
[36:05] Speaker 4: taking Crusoe on a path to the public markets and
[36:08] Speaker 4: an IPO.
[36:10] Speaker 2: Yeah, so we've been thrilled to add, when we think
[36:12] Speaker 2: about Crusoe, really the three core pillars of the business
[36:15] Speaker 2: are energy, data centers, and AI cloud computing. The three
[36:20] Speaker 2: board members we added, JB Strobel, who was the co-founder
[36:24] Speaker 2: and CTO of Tesla, who's one of the most brilliant
[36:28] Speaker 2: energy minds of our generation. He's joined our board on
[36:32] Speaker 2: the energy side. Bill Stein is widely considered to be
[36:36] Speaker 2: sort of the godfather of the data center industry, and
[36:39] Speaker 2: he's become you know, a dear friend and a, and
[36:41] Speaker 2: a great mentor for me and, and, uh, has helped
[36:44] Speaker 2: Crusoe grow quite a bit. He's joined our board that
[36:46] Speaker 2: really represents that data center pillar. And then Thomas Seifer
[36:50] Speaker 2: has been the longtime CFO of one of the most
[36:53] Speaker 2: successful cloud companies of all time, which is CloudFlare. And
[36:57] Speaker 2: Thomas brings a tremendous amount of financial knowledge and expertise
[37:01] Speaker 2: and has joined us as our audit chair and board member.
[37:04] Speaker 2: So as we think about the public markets, look, it's
[37:07] Speaker 2: something we're always considering what the best ways are to
[37:11] Speaker 2: capitalize Crusoe's growth. We are seeing tremendous demand from, you know,
[37:15] Speaker 2: across our suite of customers. And we do have, you know,
[37:18] Speaker 2: massive ambitions to continue to grow this vertically integrated AI
[37:22] Speaker 2: infrastructure platform. So, you know, I think, you know, at
[37:25] Speaker 2: some point it probably does make sense for us to
[37:27] Speaker 2: enter the public market, but it is kind of something
[37:29] Speaker 2: that we're constantly evaluating.
[37:32] Speaker 4: Chase, I respect that Crusoe's business is not just in
[37:36] Speaker 4: the United States, right? You are looking at the Gulf,
[37:38] Speaker 4: for example, and other regions. But right now in America,
[37:42] Speaker 4: data center is not necessarily a good word, a good phrase.
[37:46] Speaker 4: We have the midterms approaching where there are a number
[37:48] Speaker 4: of races where data centers are an issue for voters.
[37:52] Speaker 4: In Texas in particular, you have a big footprint in Texas,
[37:55] Speaker 4: and Governor Abbott has this sort of moratorium. Abilene's the
[37:59] Speaker 4: big one for you. It doesn't seem like that's impacted.
[38:02] Speaker 4: But talk about the general attitude towards data centers in
[38:05] Speaker 4: this country and if you are seeing any negative impact
[38:08] Speaker 4: because of the political environment right now.
[38:11] Speaker 2: Yeah, so we've been sort of working with the local
[38:17] Speaker 2: communities where we are partnered. I think more than anything,
[38:20] Speaker 2: the data center industry has sort of a marketing issue.
[38:24] Speaker 2: We are seeing a lot of information that's just, frankly,
[38:28] Speaker 2: not accurate. I think that the benefits that we provide
[38:31] Speaker 2: the communities and the partnership that we take when we're
[38:34] Speaker 2: building in those communities is not always being properly communicated.
[38:38] Speaker 5: So, you know, I think the water narrative, for instance,
[38:40] Speaker 5: is a.
[38:40] Speaker 2: Great example of this, where, you know, there's this narrative
[38:44] Speaker 2: out there that data centers are consuming and using tons
[38:47] Speaker 2: and tons of water. And, you know, the way we've
[38:49] Speaker 2: designed and engineered these next-gen purpose-built AI factories is really
[38:53] Speaker 2: to not use any water, use hardly any water. Each
[38:57] Speaker 2: of our big buildings in Abilene, Texas uses roughly the
[39:01] Speaker 2: same water footprint as about 10 single family homes. So
[39:04] Speaker 2: you're talking about basically water usage from people going to
[39:07] Speaker 2: the bathroom, watering our plants, just general purpose usage. But
[39:11] Speaker 2: the water to cool the chips, that's actually operating on
[39:13] Speaker 2: a closed loop system.
[39:15] Speaker 5: So you have cold water that goes into the racks.
[39:17] Speaker 2: And it gets cooled off in these air-cold chillers outside
[39:21] Speaker 2: of the data center. And from a job creation perspective,
[39:25] Speaker 2: these can be transformative for these communities. And from a
[39:28] Speaker 2: tax revenue perspective, they can also be transformative as well.
[39:31] Speaker 2: You look at a market like Abilene, we'll be about
[39:34] Speaker 2: a third of the total tax base in Abilene and
[39:37] Speaker 2: Taylor County. And we're more than doubling the annual receipts
[39:44] Speaker 2: for the local school system. And another area where we're
[39:49] Speaker 2: building in Texas, a large campus is Armstrong County, where
[39:53] Speaker 2: we're increasing the tax revenue by over 100%. So these
[39:59] Speaker 2: are massive transformational investments that really lead to these larger
[40:04] Speaker 2: budgets that the communities can invest in long-term infrastructure and
[40:08] Speaker 2: services like the police, the fire department, the roads, and
[40:12] Speaker 2: then obviously the education for the future generations of children.
[40:15] Speaker 4: But Chase, there's no specific projects that have been delayed
[40:20] Speaker 4: or impacted or paused because of local policy or opposition?
[40:25] Speaker 5: Um, not yet.
[40:27] Speaker 2: Um, but you know, we're definitely seeing it happen, uh,
[40:30] Speaker 2: and unfold in real time. So, um, you know, it's
[40:33] Speaker 2: something that we're trying to get ahead of and, you know,
[40:36] Speaker 2: really we think that the truth is on our side. And, um,
[40:38] Speaker 2: as people, you know, really get to understand the type
[40:41] Speaker 2: of partner that Crusoe is, uh, you know, we, we
[40:43] Speaker 2: think that, uh, you know, we, we can be, uh,
[40:46] Speaker 2: really great partners to local communities where we're investing for,
[40:49] Speaker 2: you know, multiple decades, uh, to come.
[40:51] Speaker 4: Chase Lockmiller, CEO of Crusoe, uh, almost $ 4 billion in funding,
[40:56] Speaker 4: a lot of momentum, new board members. Thank you very
[40:58] Speaker 4: much indeed. Now coming up, more warnings from big names
[41:01] Speaker 4: in AI. We're going to discuss what Microsoft's AI chief
[41:05] Speaker 4: and Amazon have had to say this morning. This is
[41:07] Speaker 4: Bloomberg Tech. Let's stick with AI risks because we've been
[41:26] Speaker 4: hearing more warnings and calls for rigorous testing. Blue Most
[41:29] Speaker 4: Matte Day is with us. And Matt, there are different
[41:31] Speaker 4: camps emerging, but we've heard from Microsoft's AI CEO.
[41:35] Speaker 12: What's the thesis there? So the thesis, it's Microsoft's Mustafa Suleiman.
[41:40] Speaker 12: He's a longtime AI developer, co-founded Google's DeepMind. He says
[41:43] Speaker 12: part of the reason that AI might appear like it's
[41:46] Speaker 12: got some kind of emotions or humanity is because it's
[41:48] Speaker 12: been trained to believe that. They're taking a shot at
[41:51] Speaker 12: Anthropic and Claude in particular, pointing to a bunch of
[41:54] Speaker 12: instances in the training data where, listen, if you say
[41:57] Speaker 12: Claude might have some sort of proto-emotions or an identity
[42:00] Speaker 12: or a sense of self, it's going to behave that way.
[42:02] Speaker 12: It doesn't mean the thing is conscious. It means you
[42:04] Speaker 12: taught it to do that, and Microsoft is suggesting we
[42:06] Speaker 12: all knock it off.
[42:08] Speaker 4: So I should say, we're talking to Matt Day about
[42:11] Speaker 4: this sort of Seattleite section of the AI market. We'll
[42:14] Speaker 4: go to Amazon next. Amazon is a massive hyperscaler, right?
[42:18] Speaker 4: A lot of workloads run on AWS, but also does
[42:21] Speaker 4: a big body of work in its own right. And
[42:23] Speaker 4: they just are kind of in line with Jensen Wang
[42:25] Speaker 4: about don't ship the model if it's not ready. Pretty simple, right?
[42:30] Speaker 12: Yeah, pretty common sense. You know, they said don't ship
[42:32] Speaker 12: it until it's safe, don't ship it until it's been tested.
[42:34] Speaker 12: You know, they're not really coming out one way or
[42:36] Speaker 12: another on the sort of broader should we slow down debate,
[42:38] Speaker 12: but sort of a pragmatic stance, which kind of fits
[42:41] Speaker 12: with Amazon. They've been sort of a neutral party in
[42:42] Speaker 12: a lot of this. They're hosting some open AI, they're
[42:44] Speaker 12: hosting some anthropic. So this one kind of fits with
[42:47] Speaker 12: their kind of reasonable position here.
[42:50] Speaker 4: Matt, just very, very quickly, Microsoft has some exposure to
[42:55] Speaker 4: open AI. Can we just explain that real quick?
[42:59] Speaker 12: Exposure to open AI? Oh, big old investor, big old backer,
[43:03] Speaker 12: big user of their models. Really, really tied up with them.
[43:06] Speaker 12: And increasingly anthropic, too. They're all kind of intertwined.
[43:09] Speaker 4: Bloomberg's Matt Day with the kind of Microsoft-Amazon viewpoint, which
[43:13] Speaker 4: came right at the end of this week. Thank you
[43:15] Speaker 4: very much. One final story as we head into the weekend.
[43:18] Speaker 4: Apple's iPhone 18 Pro is officially on sale around the world.
[43:22] Speaker 4: It's the first major product launch under new CEO John Ternes,
[43:26] Speaker 4: who marked the occasion at Apple's flagship store in New
[43:30] Speaker 4: York City, opening up the doors, mingling. talking to people
[43:35] Speaker 4: lined up to get that generation of handset. Of course,
[43:37] Speaker 4: the foldable iPhone Duo does not go on sale until
[43:41] Speaker 4: later next month. Apple shares softened by 7 tenths of 1%. Okay,
[43:46] Speaker 4: that does it for this edition of Bloomberg Tech. There
[43:48] Speaker 4: was so much in this show today about the events
[43:51] Speaker 4: of the last seven days. a summary, a lot of
[43:54] Speaker 4: different points of view weighing in. Recap all of that
[43:57] Speaker 4: on the podcast. You can find it on the Bloomberg
[43:59] Speaker 4: Terminal as well as online on Apple, Spotify, and iHeart.
[44:03] Speaker 4: I would wish you all a very happy and productive weekend.
[44:07] Speaker 4: Thank you for watching the show this week. This is
[44:09] Speaker 4: Bloomberg Tech.

Shows are listed from Apple Podcasts' charts and directory, and Podcast Index where it is used. Every show and episode belongs to its publisher and plays from the publisher's own feed; this site hosts none of them.