# OpenAI’s Revenue Surge as AI Infrastructure Race Intensifies

An episode of [Bloomberg Tech](https://invisiblenetwork.ai/podcast/1611767434), 9 Oct 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/9315a4a3-4d7f-4af8-a2c4-b4de01126ad8/audio.mp3?utm_source=Podcast&in_playlist=9caafc41-289c-4115-995d-ae7c0030ab75

Bloomberg’s Ed Ludlow breaks down OpenAI's latest annualized revenue, expected to reach or exceed $70 billion by the end of the year. Plus, Oracle keeps several data centers on track by relying on truck deliveries of natural gas, and a conversation with Kleiner Perkins partner Ilya Fushman on the mega AI IPO pipeline and the future of physical AI. See omnystudio.com/listener for privacy information.

## Transcript

[0:02] Speaker 1: Bloomberg Audio Studios.
[0:04] Speaker 2: Podcasts. Radio.
[0:06] Speaker 3: News.
[0:13] Speaker 1: Bloomberg Tech is live from the heart of Silicon Valley
[0:17] Speaker 1: with Ed Ludlow in San Francisco. This is Bloomberg Tech.
[0:24] Speaker 4: Coming up, OpenAI expects to reach or exceed $ 70 billion
[0:27] Speaker 4: in annualized revenue by the end of the year.
[0:30] Speaker 1: Plus, Oracle keeps several.
[0:32] Speaker 4: Data centers on track by relying on truck deliveries of
[0:35] Speaker 4: natural gas. And we sit down with Kleiner Perkins' partner,
[0:38] Speaker 4: Ilya Fushman, to get his take on the mega IPO pipeline.
[0:42] Speaker 1: And a lot more. Really big conversation coming up. You
[0:46] Speaker 1: made it. Happy Friday. In a week that started.
[0:49] Speaker 4: With the NASDAQ 100 building on and posting consecutive record highs,
[0:52] Speaker 4: and let's be honest, then we had a bit of
[0:54] Speaker 4: a wobble in the markets. We're going to talk about
[0:56] Speaker 4: why in just a second. Net-net, the NASDAQ 100 is
[0:58] Speaker 4: flat on the week. The pain really felt in semiconductors.
[1:01] Speaker 4: What's so interesting is the declines in chip stocks. Look
[1:04] Speaker 4: at the SOX, the Philadelphia Semiconductor Index. puts this index
[1:08] Speaker 4: on track for a pretty steep weekly drop, its first
[1:10] Speaker 4: weekly decline going back to the week that ended August 28th,
[1:13] Speaker 4: the picks and shovels. Again, we'll get to that, but
[1:16] Speaker 4: it's all because of some confusion around the top story.
[1:19] Speaker 4: And today's top story and big number, $ 70 billion. That's
[1:22] Speaker 4: OpenAI's projected annualized revenue by the end of the year.
[1:26] Speaker 4: According to sources who say the company shared that number
[1:30] Speaker 4: with investors as part of discussion for its latest fundraising efforts,
[1:34] Speaker 4: Bloomberg's Rebecca Torrance. broke the story last night with Shireen Ghaffari,
[1:38] Speaker 4: I think gave a lot of clarity on what's been
[1:40] Speaker 4: a confusing few days. Just explain the basics of what
[1:43] Speaker 4: we're reporting here.
[1:44] Speaker 1: Absolutely.
[1:45] Speaker 2: So it's really rattled the markets out, as you mentioned.
[1:49] Speaker 2: This particular story shares that OpenAI's annualized revenue, which is
[1:53] Speaker 2: a projection of its sales over a longer period, is
[1:55] Speaker 2: set to hit $ 70 billion by the end of this year.
[1:59] Speaker 2: We reported earlier yesterday that right now it's at about
[2:02] Speaker 2: $ 50 billion. When that number was reported by a bunch
[2:05] Speaker 2: of outlets, there was a bit of confusion because there
[2:08] Speaker 2: was a $ 70 billion number out there previously.
[2:11] Speaker 1: Right.
[2:12] Speaker 2: And the understanding had been that that is where OpenAI's
[2:15] Speaker 2: annualized revenue is now.
[2:17] Speaker 1: So it threw the.
[2:18] Speaker 2: Markets for a bit of a loop to say, oh,
[2:19] Speaker 2: this number is now $ 20 billion less than we had expected.
[2:22] Speaker 5: Right.
[2:23] Speaker 2: What that's coming from is a difference in how OpenAI
[2:26] Speaker 2: and Anthropic calculate their revenue, which we can talk a
[2:29] Speaker 2: bit about.
[2:29] Speaker 6: Yeah.
[2:30] Speaker 1: Well, let's talk about it.
[2:31] Speaker 4: I'll point out first, what is annualized revenue? A lot
[2:34] Speaker 4: of people miss the basics, and that is they're not
[2:36] Speaker 4: posting in calendar 26, $ 70 billion. OpenAI basically expects to
[2:41] Speaker 4: end this year making money at a pace that if
[2:44] Speaker 4: they kept going at that pace over the next 12 months,
[2:47] Speaker 4: they'd make $ 70 billion of revenue. Audience, that's how we
[2:51] Speaker 4: understand it. But what they categorize in revenue accounts for
[2:55] Speaker 4: some of the confusion. So now explain that part.
[2:57] Speaker 6: Right.
[2:57] Speaker 2: So Anthropic, when it's calculating its annualized revenue, and our
[3:01] Speaker 2: understanding is that the most recent figure there that's public
[3:04] Speaker 2: has been around $ 65 billion for them, that that includes
[3:07] Speaker 2: gross sales from its cloud partners. OpenAI strips that out.
[3:11] Speaker 2: And so the figure that is being shared here, the
[3:14] Speaker 2: $ 50 billion is its net revenue. annualized revenue just from
[3:18] Speaker 2: its share of the cloud revenue.
[3:19] Speaker 1: As opposed to gross. Right, exactly.
[3:22] Speaker 2: But because it's not a one-to-one apples-to-apples comparison with Anthropic,
[3:27] Speaker 2: it's really hard to compare these two businesses, these direct
[3:31] Speaker 2: competitors against each other.
[3:33] Speaker 4: So, look, there are lots of people who have sort
[3:35] Speaker 4: of maybe a more limited understanding of private markets and
[3:38] Speaker 4: the companies. And they're thinking, like, why do we have
[3:39] Speaker 4: all these financial metrics? These companies aren't even public yet, right?
[3:43] Speaker 4: In our reporting, the whole point, go back a week,
[3:46] Speaker 4: we broke the story on this private funding round. And
[3:49] Speaker 4: as part of that mechanism, as I understand it, and
[3:52] Speaker 4: in your reporting too, OpenAI is basically explaining to investors
[3:56] Speaker 4: what's going on.
[3:56] Speaker 1: As part of an effort to raise funds.
[3:58] Speaker 2: Mm-hmm.
[3:59] Speaker 1: That's correct.
[4:00] Speaker 2: And this is happening in AI a lot because these
[4:02] Speaker 2: companies are growing so quickly to share just the revenue
[4:06] Speaker 2: as far as it's come this year. It doesn't feel
[4:08] Speaker 2: like quite enough. And so OpenAI, in its conversations with investors,
[4:12] Speaker 2: we've reported that it's in talks to raise $ 30 billion
[4:14] Speaker 2: at a $ 1. 4 trillion valuation, finds it more apt to
[4:17] Speaker 2: use the $ 50 billion number to explain, look how much
[4:20] Speaker 2: we're growing, $ 70 billion in annualized revenue by the end
[4:23] Speaker 2: of the year. We have much more on the way.
[4:24] Speaker 4: There's just so many people on the socials that are like,
[4:26] Speaker 4: can you imagine if these companies were public in every
[4:28] Speaker 4: quarter when they file earnings? We'll save that for another time.
[4:32] Speaker 4: Rebecca Torrance, great job. Thank you very much for being
[4:34] Speaker 4: with us. Yesterday, a debate over one AI company's revenue
[4:39] Speaker 4: run rate helped knock the NASDAQ 100 down 1.4% and
[4:42] Speaker 4: chips down 3.4%.
[4:42] Speaker 1: Today, you just heard the reporting.
[4:47] Speaker 4: The open AI revenue picture looks stronger. At least there's
[4:50] Speaker 4: some clarity. We're ending Friday with a little bit of
[4:53] Speaker 4: a bounce back. That tells you how sensitive the market's
[4:56] Speaker 4: become to the question, does the revenue ultimately justify all
[5:00] Speaker 4: this AI spending? For more, Michelle Weaver, Morgan Stanley, US
[5:04] Speaker 4: Thematic Research Strategist and Executive Director as well as now.
[5:07] Speaker 4: And I read your body of work. You think we're
[5:09] Speaker 4: kind of already in this next phase of AI anyway.
[5:12] Speaker 4: It's just so amazing to see financial markets around the
[5:15] Speaker 4: world so focused on ARR numbers.
[5:20] Speaker 1: What does that tell you?
[5:22] Speaker 7: So I think it's super important, first off, to split
[5:24] Speaker 7: the universe. So this theme is the most dominant in
[5:27] Speaker 7: the market. And you have companies that are AI enablers
[5:31] Speaker 7: and you have companies that are AI adopters. And the
[5:33] Speaker 7: way we make sense of all of that is through
[5:36] Speaker 7: our proprietary stock mapping, which we do with all of
[5:39] Speaker 7: our analysts. We tag all 3,600 stocks that we cover
[5:42] Speaker 7: on their exposure to AI. And And what we're finding
[5:45] Speaker 7: now is incrementally positive data points around AI adoption. And
[5:49] Speaker 7: that's what really matters for the build-out. You need the
[5:52] Speaker 7: diffusion part of the story to play out. And while
[5:55] Speaker 7: it's still early, we are seeing positive data points there.
[5:57] Speaker 7: We found from this mapping 460 basis points of EBIT
[6:01] Speaker 7: margin expansion for stocks that our analysts tagged as high
[6:05] Speaker 7: materiality adopters.
[6:06] Speaker 1: That's just ahead of.
[6:07] Speaker 7: The S & P 500 and well ahead of the
[6:09] Speaker 7: broader MSCI world index. We're also seeing upside in earnings
[6:14] Speaker 7: increasingly go into that adopter bucket as well.
[6:18] Speaker 1: I've really enjoyed passing through your.
[6:22] Speaker 2: Notes.
[6:22] Speaker 1: You know, I reflected on the week that was.
[6:25] Speaker 4: And one of the reasons I really enjoy speaking to
[6:26] Speaker 4: you is like one session a market does not make.
[6:28] Speaker 1: I say that all the time.
[6:30] Speaker 4: But we started with the NASDAQ 100 posting consecutive record highs.
[6:34] Speaker 4: Everyone already has a lens, an eye to earnings. about
[6:38] Speaker 4: a lot of what you're talking about, it's still picks
[6:41] Speaker 4: and shovels. It's still the AI build out. We're still
[6:44] Speaker 4: monitoring hyperscaler capex. Nothing's really changed as far as I
[6:47] Speaker 4: can tell.
[6:49] Speaker 7: So we think we're really in a transition period right now.
[6:51] Speaker 7: And so what we're calling for is a barbell. Typically,
[6:54] Speaker 7: in the tech cycle, you have a pretty linear handoff.
[6:57] Speaker 7: It goes semiconductors, infrastructure, software and services, and the applications
[7:01] Speaker 7: that are built on top of that tech cycle. And
[7:04] Speaker 7: I think the cycle is different. The majority of the
[7:06] Speaker 7: returns have really been accruing on the enabler side, but
[7:09] Speaker 7: we have some really serious bottlenecks to contend with right now,
[7:12] Speaker 7: namely power and politics. And so that's leaving us to
[7:15] Speaker 7: be selective on some of the enabler names, but still
[7:18] Speaker 7: broaden out into those early cycle adopters.
[7:21] Speaker 3: Right.
[7:22] Speaker 4: So Michelle, actually ahead of the midterms, I want to
[7:24] Speaker 4: put the polling data to one side because you at
[7:27] Speaker 4: Morgan Stanley, you have your own survey data. You're looking
[7:30] Speaker 4: at data centers through that lens of the consumer. What
[7:33] Speaker 4: are your findings?
[7:35] Speaker 7: So we found that when we run our surveys, this
[7:37] Speaker 7: is a broad sample, 2,000 U.S. consumers diversified by gender, region,
[7:43] Speaker 7: and race. We found that net sentiment is overall positive
[7:46] Speaker 7: for AI. So 14% net positive, more people supporting AI
[7:51] Speaker 7: than disfavoring it. But when we make the issue local,
[7:55] Speaker 7: when we ask, would you support a data center within
[7:57] Speaker 7: 10 miles of your home, the issue becomes sharply negative
[8:01] Speaker 7: and you have net sentiment at negative 20%. So in
[8:04] Speaker 7: the abstract, people are still more supportive of AI. But
[8:07] Speaker 7: when you make it local and you make it about
[8:09] Speaker 7: the build out itself, that's where you see it really turn.
[8:13] Speaker 4: So, you know, I appreciate that you have a longer
[8:16] Speaker 4: time horizon in your research where you look. But in
[8:19] Speaker 4: the midterms, there are specific races where candidates make that
[8:24] Speaker 4: data centers more from the standpoint of proximity to community
[8:27] Speaker 4: and issue. What are the sort of short term effects
[8:30] Speaker 4: of the outcomes in those races?
[8:33] Speaker 7: So I think what's most important to watch as we
[8:35] Speaker 7: head into midterms is the outcome at the governor level.
[8:38] Speaker 7: So you've seen a lot of the moratoriums and a
[8:40] Speaker 7: lot of the restrictions that have been put on so
[8:43] Speaker 7: far being done via executive order by governor. So that's
[8:47] Speaker 7: where I would expect the most change to happen coming
[8:49] Speaker 7: out of midterms. While it's something that's impacting and been
[8:52] Speaker 7: a big issue in the congressional races, I think more
[8:54] Speaker 7: of the change is going to happen at that state level.
[8:57] Speaker 4: So the next question, do we later get federal regulation
[9:02] Speaker 4: on data center moratorium or is something that pauses or
[9:06] Speaker 4: halts the activity?
[9:09] Speaker 7: So the federal government, by and large, still looks very,
[9:12] Speaker 7: very supportive of the broad data center build-out. The one
[9:15] Speaker 7: issue that you're seeing more momentum behind at the federal
[9:18] Speaker 7: level is on protecting rate payers and protecting consumer power bills.
[9:23] Speaker 7: So if anything's to change, I would expect it to
[9:26] Speaker 7: be more centered around that versus a broader slowdown or
[9:30] Speaker 7: a broader halt on the data center build-out.
[9:33] Speaker 4: Just real quick, how much of a headwind is power
[9:35] Speaker 4: as a bottleneck right now in your research?
[9:38] Speaker 7: Power, I would say, is still the number one bottleneck.
[9:41] Speaker 7: And the way we think about the power shortfall is
[9:43] Speaker 7: we take our semiconductor team's chip sales estimates. We look
[9:47] Speaker 7: at how much power would be needed for that. And
[9:49] Speaker 7: then we look at available grid capacity. And so when
[9:51] Speaker 7: we do that math, we find a 57 gigawatt shortfall
[9:55] Speaker 7: in the power needed through 2028. For context, New York
[9:58] Speaker 7: City uses around 6 to 10. 10 gigawatts of power.
[10:01] Speaker 7: So we're several New York City short here. When we
[10:04] Speaker 7: layer on time to power solutions, so things like Bitcoin
[10:07] Speaker 7: miner site conversion, natural gas turbines, fuel cells, all those
[10:12] Speaker 7: behind the meter solutions, we still end up with a
[10:14] Speaker 7: 30 to 40% shortfall in the power needed.
[10:15] Speaker 4: So We have a really interesting story about this exact
[10:19] Speaker 4: issue coming up in the program next. Michelle Weaver of
[10:21] Speaker 4: Morgan Stanley, really great to have you back with us.
[10:23] Speaker 4: Happy Friday. Thank you very much. By the way, an
[10:26] Speaker 4: update overnight. NVIDIA-backed data center operator Firmus is exploring a
[10:30] Speaker 4: private funding round of as much as $ 3 billion after
[10:34] Speaker 4: pulling its planned IPO in Australia.
[10:36] Speaker 1: That's according to sources.
[10:37] Speaker 4: The company had been targeting about $ 5 billion from the listing,
[10:41] Speaker 4: but investor concerns over valuation, higher borrowing costs, the capital needed...
[10:45] Speaker 4: to build out its data center pipeline, basically weighed on
[10:47] Speaker 4: the deal. Firmus says it will consider other funding options,
[10:50] Speaker 4: which would include basically.
[10:51] Speaker 1: A mix of debt and equity.
[10:54] Speaker 4: All right, coming up, Oracle's taking on a new approach
[10:57] Speaker 4: to kind of avoid delays in data center bills. This
[10:59] Speaker 4: is a super interesting one from Bloomberg's Brody Ford.
[11:02] Speaker 1: We have that next. This is Bloomberg Tech.
[11:16] Speaker 4: The AI build-outs reshaping corporate credit. Hyperscalers are on track
[11:20] Speaker 4: to raise around $ 300 billion in the investment-grade bond market
[11:23] Speaker 4: this year alone. Investors are still buying, but as the
[11:26] Speaker 4: number grows, so do concerns about how much debt the
[11:30] Speaker 4: market can absorb. Bloomberg's Wall Street Week takes a closer
[11:33] Speaker 4: look at the risks behind AI's growing debt boom.
[11:36] Speaker 6: By now, we all know that AI has an insatiable
[11:40] Speaker 6: appetite for compute and that means more data centers, chips,
[11:43] Speaker 6: and other expensive infrastructure. Equity investors have already benefited, seeing
[11:48] Speaker 6: shares of the six hyperscalers rise by an average of 15% year-to-date.
[11:53] Speaker 6: Now it's the debt investors' turn.
[11:56] Speaker 8: So the numbers keep growing, and I think that's the
[11:59] Speaker 8: challenging piece of it.
[12:01] Speaker 6: Debt investors like Lauren Moran, a fixed-income portfolio manager at
[12:05] Speaker 6: Wellington Management.
[12:07] Speaker 8: When we look back to 2025, where really this borrowing
[12:11] Speaker 8: from hyperscalers had really only begun, through the course of 2026,
[12:16] Speaker 8: we're probably looking at around $ 300 billion in the investment-grade
[12:21] Speaker 8: market borrowings from hyperscalers themselves.
[12:25] Speaker 6: By some estimates, the tech sector accounted for just 5%
[12:27] Speaker 6: of investment-grade corporate bond issuance in 2024. That number is
[12:33] Speaker 6: now close to 20%.
[12:35] Speaker 8: When we think about over the course of this year,
[12:37] Speaker 8: likely we'll see two and a quarter trillion of total
[12:40] Speaker 8: issuance in the U.S. investment grade market. So it's a
[12:43] Speaker 8: pretty substantial portion of the debt that we're seeing in
[12:47] Speaker 8: the market. And importantly, when we think about the actual
[12:51] Speaker 8: growth in our debt markets is largely coming from the
[12:54] Speaker 8: hyperscalers at this point. And we look at year over
[12:57] Speaker 8: year investment grade issuance being up around 35 percent.
[13:01] Speaker 6: Scott Schulte is the global co-head of investment-grade debt syndicate
[13:05] Speaker 6: at Barclays.
[13:07] Speaker 3: I think where you're seeing more differentiation is in some
[13:09] Speaker 3: of the off-balance sheet structures, whether it's the GPU financings,
[13:13] Speaker 3: the chip financings, the actual data centers themselves. You've seen
[13:17] Speaker 3: in recent days there's been force majeure, at least implied
[13:21] Speaker 3: in a couple instances. And so you have to start
[13:23] Speaker 3: thinking about the AI ecosystem holistically and tangentially what impacts it,
[13:28] Speaker 3: meaning can I get the power? Can I get the transmission?
[13:32] Speaker 3: Can I get the heating and cooling? Can I get
[13:34] Speaker 3: the thing built in two years when it's supposed to
[13:36] Speaker 3: be built? And if it's not, what are the outs
[13:39] Speaker 3: that the hyperscaler might have in terms of the contractual
[13:42] Speaker 3: agreement on that data center?
[13:44] Speaker 4: You can get the full story on Wall Street Week tonight,
[13:47] Speaker 4: 6 p.m.
[13:48] Speaker 1: Eastern Time with Bloomberg's David Weston.
[13:51] Speaker 4: Oracle is using a new power strategy to keep many
[13:54] Speaker 4: of its data centers on track, trucking natural gas directly
[13:59] Speaker 4: to the server farms. The company is considering using this
[14:01] Speaker 4: approach to speed up a build in New Mexico where
[14:04] Speaker 4: a gas pipeline needed to help power one of its
[14:06] Speaker 4: crucial data centers has been delayed. So shares of Oracle
[14:09] Speaker 4: are up a lot. They're up like 5% in the
[14:11] Speaker 4: session now. After this story broke yesterday, they dropped a lot.
[14:15] Speaker 4: There were other factors in that, including the OpenAI story
[14:17] Speaker 4: that Rebecca hit earlier today. Here to explain, and the
[14:21] Speaker 4: person that broke the story, Bimbo's Brody Ford. We'll get
[14:25] Speaker 4: to the reaction later. The basics, they're trucking in natural gas.
[14:30] Speaker 4: That seems reasonable. I mean, what's that about?
[14:33] Speaker 5: In an ideal world, you wouldn't be anywhere near this.
[14:36] Speaker 5: You would have a data center that you plug into
[14:38] Speaker 5: the electrical grid. Failing that, you have an on-site power
[14:42] Speaker 5: plant that you have a pipeline to.
[14:45] Speaker 1: And hence the delay on this pipeline being a problem.
[14:48] Speaker 5: Yes, but then if a pipeline is delayed six months out,
[14:51] Speaker 5: then you start saying, wow, we've borrowed a lot of
[14:54] Speaker 5: money to build these data centers. Our entire business is
[14:56] Speaker 5: premised on getting them online quickly. We need to find
[15:00] Speaker 5: a stopgap. And so trucking out large amounts of gas
[15:03] Speaker 5: is something you would do for, say, mining in the
[15:06] Speaker 5: outback or temporary industrial actions in very remote places. And
[15:12] Speaker 5: now you have 30 trucks a day going to data
[15:14] Speaker 5: centers outside of Salt Lake City. And so we're quite
[15:18] Speaker 5: far from the ideal power scenario, but Oracle is finding
[15:22] Speaker 5: a way to keep these things moving.
[15:23] Speaker 4: So Oracle did not respond to our request to comment
[15:26] Speaker 4: on the story, but I saw posting on the socials
[15:29] Speaker 4: something related to this. And then the stock reaction in
[15:32] Speaker 4: the moment, there was a lot going on.
[15:34] Speaker 1: Right, yeah.
[15:35] Speaker 5: Oracle didn't respond to our request for comment, but then
[15:37] Speaker 5: they tweeted pretty much saying, hey, we love our partner
[15:39] Speaker 5: that helps us do this thing.
[15:41] Speaker 4: I see, like getting the gas onto the trucks and
[15:44] Speaker 4: to the site. And then, yeah, I think the stock
[15:46] Speaker 4: was under, it's up again, up almost 5% today, but
[15:48] Speaker 4: it was under pressure in part because of the confusion
[15:51] Speaker 4: around OpenAI's ARR number. And then when this report broke,
[15:54] Speaker 4: there was continued sort of downward slide.
[15:56] Speaker 1: Is that right?
[15:57] Speaker 2: Right.
[15:57] Speaker 5: Our understanding there is that the market for this has
[16:00] Speaker 5: a bit, hey, the delays are more significant than maybe
[16:03] Speaker 5: we expected, right? I mean, this is quite a, this
[16:07] Speaker 5: is a situation where you only do this if you're
[16:10] Speaker 5: really under the gun. And so our reporting shows that
[16:13] Speaker 5: Oracle's finding a way to make it work, but their
[16:16] Speaker 5: back is up against the wall here.
[16:18] Speaker 1: Bloomberg's Brady Ford. Terrific reporting. Thank you very much indeed.
[16:22] Speaker 4: Okay, coming up, Hiba Amber from the Ericsson Immigration Group
[16:24] Speaker 4: is going to join us to discuss the suspension of
[16:27] Speaker 4: some companies and universities, which includes Microsoft, from a worker
[16:31] Speaker 4: visa program. What impact could that have on tech, but
[16:34] Speaker 4: also the wider economy?
[16:36] Speaker 1: That's next. This is Bloomberg Tech.
[16:46] Speaker 4: Microsoft, Adobe, along with half a dozen other companies and
[16:50] Speaker 4: nine universities have been accused of abusing a U.S.
[16:53] Speaker 1: Worker visa program.
[16:55] Speaker 4: The Trump administration has suspended them from the Permanent Labor
[16:59] Speaker 4: Certification Program, or PERM. The program allows companies to sponsor
[17:03] Speaker 4: foreign workers for permanent residency, green cards, in other words.
[17:07] Speaker 4: It's a small part of the U.S. immigration ecosystem and
[17:10] Speaker 4: a part of the debate. But let's talk about the
[17:12] Speaker 4: impact it could have on tech and the economy today.
[17:15] Speaker 4: with Hibamba, a partner of the Ericsson Immigration Group. This
[17:18] Speaker 4: broke just before we came on air yesterday, and we
[17:21] Speaker 4: heard from the vice president, J.D.
[17:23] Speaker 1: Vance.
[17:23] Speaker 4: Just let's start with your reaction to the action taken
[17:26] Speaker 4: by this administration.
[17:29] Speaker 9: So it seems that the administration's concern is that companies
[17:33] Speaker 9: laid off Americans while sponsoring foreign workers, which is a
[17:37] Speaker 9: reasonable question to ask, but I think that it helps
[17:41] Speaker 9: to understand how the immigration process actually works here. So
[17:46] Speaker 9: the program that is being suspended is for existing employees
[17:50] Speaker 9: that companies have already deemed to be a value add
[17:53] Speaker 9: to the company. There is no new employee that's being
[17:57] Speaker 9: hired through the program that's been suspended. It's the same
[18:00] Speaker 9: person that's been working for the company for, in some instances,
[18:04] Speaker 9: many years. The other thing I think that is really
[18:07] Speaker 9: important for us to understand is that the rules involved
[18:10] Speaker 9: in this program actually already account for certain layoff protections.
[18:15] Speaker 9: So if a company has had layoffs in that same
[18:17] Speaker 9: position in the past six months, there is a very
[18:21] Speaker 9: high likelihood that the application that is part and parcel
[18:24] Speaker 9: of the PERM program would actually not be permitted to
[18:26] Speaker 9: even proceed, like would not be permitted to be filed.
[18:30] Speaker 9: I think that there are a lot of unanswered questions
[18:32] Speaker 9: as to what the administration's motivation is, because I do
[18:36] Speaker 9: question whether or not suspending a program for people who
[18:40] Speaker 9: are already here in the U.S. gets us to where
[18:42] Speaker 9: they want to go in terms of focusing on whether U.S.
[18:46] Speaker 9: citizens are being replaced.
[18:48] Speaker 4: I think Microsoft is not the only company or case
[18:51] Speaker 4: study here, but it was the name singled out by
[18:56] Speaker 4: the vice president. I actually think we have some sound
[18:59] Speaker 4: of what J.D. Vance said.
[19:02] Speaker 3: There has been no company in the United States, unfortunately,
[19:05] Speaker 3: that has abused this system more than Microsoft.
[19:07] Speaker 1: Thanks to an ongoing.
[19:08] Speaker 10: Investigation, we have decided to suspend the.
[19:11] Speaker 3: PERM program for Microsoft, which means they will no longer
[19:15] Speaker 3: be able to take those H-1Bs and apply for permanent
[19:18] Speaker 3: resident status within.
[19:19] Speaker 1: The United States of America.
[19:25] Speaker 4: So in Microsoft's response, and the context is that the program,
[19:31] Speaker 4: not the response from the administration, but Microsoft's alleged activity
[19:35] Speaker 4: largely relates to H-1Bs. And what Microsoft said is of
[19:38] Speaker 4: those 6,000 H-1B applications last year, the majority were from
[19:43] Speaker 4: existing employees already.
[19:46] Speaker 1: In the United States. Why was that an important detail?
[19:50] Speaker 1: in the response?
[19:53] Speaker 9: I think that the more important thing to focus on here, again,
[19:56] Speaker 9: is the nuance. A layoff that is followed by the
[20:00] Speaker 9: immigration filing doesn't automatically mean that there was an American
[20:04] Speaker 9: worker that was replaced. I would imagine that a company
[20:07] Speaker 9: the size of Microsoft can cut jobs in one part
[20:09] Speaker 9: of the business and still need very different skills in another. So,
[20:13] Speaker 9: for example, if somebody was laid off in sales or gaming,
[20:15] Speaker 9: that isn't automatically interchangeable with a role that needs to
[20:19] Speaker 9: be filled for AI infrastructure or engineering or something like that.
[20:23] Speaker 9: What we're seeing is company-wide totals, but also very little
[20:27] Speaker 9: information about the details for why the administration is making
[20:30] Speaker 9: these allegations in terms of misuse of a program.
[20:36] Speaker 4: The big picture concern is that this impacts talent pipelines,
[20:40] Speaker 4: and there will be a negative economic impact to those
[20:43] Speaker 4: technology companies mentioned, but also scholars, right, in the nine
[20:48] Speaker 4: universities cases. Is that the concern that you hear from
[20:51] Speaker 4: clients that are relevant here?
[20:54] Speaker 9: It is because it's all about the talent pipeline. The
[20:57] Speaker 9: talent pipeline comes from a combination of U.S. workers, foreign workers,
[21:01] Speaker 9: and also recent foreign graduate students. And there is plenty
[21:06] Speaker 9: of data to support that a very significant number of
[21:09] Speaker 9: students that participate in higher education programs for STEM, engineering,
[21:14] Speaker 9: computer science, are actually themselves foreign students. So I think
[21:19] Speaker 9: that it's important to remember that with respect to the
[21:22] Speaker 9: talent pipeline, we're not going to be able to increase
[21:25] Speaker 9: a talent pipeline or develop it overnight. And we have,
[21:29] Speaker 9: to a certain degree, reliance on foreign workers and foreign students.
[21:33] Speaker 9: And so the perspective probably needs to be a bit
[21:37] Speaker 9: more nuanced because we're using this foreign talent, the best
[21:40] Speaker 9: and the brightest, to add to the existing talent that U.S.
[21:44] Speaker 9: workers possess. And adding talent is not necessarily the same
[21:48] Speaker 9: thing as taking away from a U.S.
[21:50] Speaker 11: Worker.
[21:50] Speaker 9: An additional skilled employee, an additional player can basically elevate
[21:55] Speaker 9: the entire team.
[21:57] Speaker 4: Very quickly, Hiba, the next step is a retaliation or
[22:01] Speaker 4: a response in the courts from the companies?
[22:03] Speaker 10: Yes.
[22:04] Speaker 9: I think we have to see how it plays out.
[22:06] Speaker 9: I think that there's a lot more detail and information
[22:10] Speaker 9: that's missing on the part of the administration. We do
[22:13] Speaker 9: have a statement and response from Microsoft themselves. Whether or
[22:16] Speaker 9: not this is going to be litigated remains to be seen.
[22:20] Speaker 4: Steve Anver, partner of Ericsson Immigration Group, thank you very much.
[22:24] Speaker 4: Coming up in the program, Ilya Fushman from Kind of
[22:26] Speaker 4: Perkins joins us. We're going to talk about AI revenues,
[22:30] Speaker 4: the IPO pipeline, but also the next big opportunities for
[22:34] Speaker 4: venture investors. What's around the corner that we're not talking
[22:37] Speaker 4: about right now? Looking forward to it. Halftime here in
[22:40] Speaker 4: the city of San Francisco. This is Bloomberg Tech. Welcome
[22:56] Speaker 4: back to Bloomberg Tech. Again, we made it to Friday
[22:59] Speaker 4: in a week where technology stocks started pushing fresh records.
[23:03] Speaker 4: And then the real story in the middle of the
[23:04] Speaker 4: week was confusion about annualized revenues from the frontier labs
[23:08] Speaker 4: and what that means for the.
[23:10] Speaker 1: Justification of spending.
[23:11] Speaker 4: And net net, the Nasdaq 100 is modestly higher this Friday.
[23:15] Speaker 4: The SOX is down and on track for its first
[23:17] Speaker 4: weekly decline since the week ending August 28th. And people
[23:20] Speaker 4: are still trying to make sense of where we are
[23:22] Speaker 4: in a cycle and where we are in the health
[23:25] Speaker 4: of those frontier labs. But so interesting to see private
[23:28] Speaker 4: companies with data and the reporting from newsrooms driving markets everywhere.
[23:34] Speaker 4: I want to take a step back and look at
[23:35] Speaker 4: where technology and the money backing it goes next. Kleiner
[23:39] Speaker 4: Perkins has been making those bets for more than 50 years.
[23:42] Speaker 4: From Amazon and Google to, more recently, Anthropic and Waymo.
[23:46] Speaker 4: And earlier this year, it raised another $ 3. 5 billion to
[23:49] Speaker 4: invest in the AI boom. Partner Ilya Fishman joins us now,
[23:53] Speaker 4: and it's really great to have you back with us.
[23:55] Speaker 10: Thanks for having me.
[23:57] Speaker 1: ARR, you.
[23:59] Speaker 4: Are an investor in Anthropic, and I appreciate it. Both
[24:02] Speaker 4: companies have filed confidentially. There's a limit here to what
[24:05] Speaker 4: we can discuss, right? But this ARR confusion, why is
[24:10] Speaker 4: that a metric that even matters if you're a venture
[24:12] Speaker 4: capitalist or a public market investor?
[24:16] Speaker 11: Taking a step back, I think if you look at
[24:17] Speaker 11: these companies, whether it's Anthropic, OpenAI, Waymo, Databricks, something that's
[24:24] Speaker 11: truly incredible that's happening today is just the scale of
[24:27] Speaker 11: revenue and the rate at which these companies are compounding
[24:30] Speaker 11: at scale. Certainly things like ARR do matter in terms
[24:33] Speaker 11: of the predictability of that revenue and the longevity of
[24:36] Speaker 11: that revenue and understanding how that revenue manifests itself over time.
[24:40] Speaker 11: But I think the more important picture here is just
[24:42] Speaker 11: we've never seen companies grow this big, this fast in
[24:45] Speaker 11: the history of technology. And that points to something truly
[24:48] Speaker 11: foundational in what they're solving.
[24:50] Speaker 4: You know, part of the confusion was also just an
[24:53] Speaker 4: accounting issue. discrepancy between how OpenAI does ARR and Anthropic
[24:57] Speaker 4: does ARR. And that was interesting. They are similar and
[25:01] Speaker 4: they are different. We have plenty of people on this
[25:03] Speaker 4: show that are on the cap table of both. When
[25:06] Speaker 4: you guys decided, okay, we will make an investment into Anthropic,
[25:10] Speaker 4: what was the thesis? What was the sort of guiding
[25:12] Speaker 4: principle of that?
[25:14] Speaker 11: Well, what we really see is across the board, whether
[25:16] Speaker 11: it's at the application layer, at the individual layer, this
[25:20] Speaker 11: impact of AI is still incredibly early and incredibly palpable, right?
[25:27] Speaker 11: It's like nothing in terms of technology, like nothing we've
[25:30] Speaker 11: seen in prior cycles. So for us, it's much less
[25:32] Speaker 11: about sort of those details that you're talking about. It's
[25:36] Speaker 11: really more about what does the future hold here? And
[25:38] Speaker 11: the future here is, from our perspective, really exciting. We're
[25:41] Speaker 11: here in San Francisco. The more AI-native, AI-centric companies look
[25:47] Speaker 11: and feel very differently inside than even the most forward-thinking
[25:51] Speaker 11: traditional enterprises. So just at the enterprise layer, getting this
[25:55] Speaker 11: technology into the market, whether it's directly through access of
[25:59] Speaker 11: anthropic and open AI APIs or it's through companies like Harvey, cognition,
[26:03] Speaker 11: open evidence that our companies are.
[26:06] Speaker 4: We've just been showing some of the portfolio and later
[26:09] Speaker 4: on in our conversation, I think we should talk a
[26:11] Speaker 4: little bit about defense technology and hardware. But if we
[26:15] Speaker 4: guys bring back up some of the port codes and
[26:18] Speaker 4: we can cycle through them, what you're talking about, about
[26:20] Speaker 4: the application layer and enterprise more specifically, how closely tied
[26:26] Speaker 4: are the fortunes of those companies to to the success
[26:29] Speaker 4: that OpenAI and Anthropic are having or hope to have?
[26:33] Speaker 11: Well, I think they're all interconnected. Ultimately, it's really the
[26:36] Speaker 11: ability to access intelligence. And some of that intelligence is
[26:39] Speaker 11: accessed through very specific domain applications like a Harvey for
[26:43] Speaker 11: legal and professional services. And some of that is going
[26:45] Speaker 11: to be accessed through developers in products like Cognition or
[26:49] Speaker 11: directly through Codex and Claude. But ultimately, it's really all
[26:54] Speaker 11: about What is the cost of intelligence that you can
[26:58] Speaker 11: access today? And that cost in terms of the dollars
[27:02] Speaker 11: per really, really high quality thinking has actually been kind
[27:06] Speaker 11: of going down on average because you're able to do
[27:09] Speaker 11: creative things like balance to older iterations of models with
[27:13] Speaker 11: newer ones. There's obviously a big movement in open source
[27:16] Speaker 11: right now and open weight models. So overall, this is
[27:19] Speaker 11: really for us about what is the opportunity of getting
[27:23] Speaker 11: this intelligence into the whole economy. And now with products
[27:26] Speaker 11: like Instinct, which we're also invested in, and obviously Muse
[27:29] Speaker 11: and things like that, now it's actually finally hitting the consumer.
[27:33] Speaker 10: And that's really the exciting part here.
[27:34] Speaker 1: It's interesting.
[27:35] Speaker 4: I've been using Dots a lot since launch. I was
[27:38] Speaker 4: at Dev Day last week. And I've been using Muse
[27:40] Speaker 4: a lot and Instinct less so. But what's so interesting
[27:45] Speaker 4: is that there is a market there of real products.
[27:50] Speaker 4: And people have choices, right? How do you know you've
[27:53] Speaker 4: picked the right one or backed the right one?
[27:56] Speaker 11: I think they're going to be tailored to different people
[27:58] Speaker 11: with different needs in some sense, but really what's more
[28:00] Speaker 11: important about that is I think this is the first
[28:03] Speaker 11: beyond the chat GPT moment. This is the first time
[28:07] Speaker 11: where consumers are deeply able to experience the magic of AI.
[28:10] Speaker 11: This agent that just does stuff for you, that hasn't
[28:13] Speaker 11: happened before, and I think it's a really foundational moment,
[28:15] Speaker 11: which is why you're seeing so much activity in this space.
[28:18] Speaker 1: Do you genuinely, day-to-day, use these agents for anything useful?
[28:22] Speaker 10: Yeah, I do.
[28:24] Speaker 11: Actually, my wife uses them a lot more than I do,
[28:26] Speaker 11: and I think it's a great product for families, but
[28:28] Speaker 11: everything from scheduling dinners, kids' activities... I buy pizza for
[28:32] Speaker 11: my kids, the standard order. I just tell it to
[28:34] Speaker 11: get it and it texts me when the pizza is done.
[28:37] Speaker 11: So it's really, it's like having, you know, a copy
[28:40] Speaker 11: of you that does stuff for you.
[28:41] Speaker 4: You know, with all of the focus on both the
[28:45] Speaker 4: Frontier Labs, but also these real agents, right? They are
[28:48] Speaker 4: agents in the sense that they can.
[28:49] Speaker 10: Take an action.
[28:50] Speaker 1: They are genuinely... the.
[28:52] Speaker 4: Creators of, say, Autonomous, I still find the most interesting
[28:55] Speaker 4: case study with Kleiner Perkins to be Waymo and the
[28:57] Speaker 4: patience that was required for everyone that went along with Waymo.
[29:02] Speaker 4: And now here we are in cities across America and
[29:05] Speaker 4: them generating, you know, revenue generating fares on rides. What
[29:10] Speaker 4: was that like being on the cap table? Well, the
[29:13] Speaker 4: patience element. So we got in the capital a little
[29:16] Speaker 4: more recently.
[29:16] Speaker 1: Okay.
[29:17] Speaker 11: But we've obviously been tracking the company for a very
[29:19] Speaker 11: long time. And it's magic, right? It's like teleportation.
[29:22] Speaker 10: You sit in the Waymo.
[29:23] Speaker 1: This was the big round they did right at the
[29:24] Speaker 1: start of this year.
[29:25] Speaker 4: Exactly.
[29:25] Speaker 10: Yeah.
[29:27] Speaker 11: But it really feels like you're in a teleportation machine
[29:30] Speaker 11: when you're in a Waymo. And by the way, you know,
[29:32] Speaker 11: Waymo is also still early, right? It's still, you know,
[29:34] Speaker 11: in the select two cities.
[29:36] Speaker 10: It's expanding.
[29:37] Speaker 4: It's kind of the parallel I'm trying to draw, you know, it,
[29:40] Speaker 4: It seems like it's been around for a long time,
[29:42] Speaker 4: but in terms of everyday person's access to the technology.
[29:45] Speaker 11: It's happening now, which is why, again, which is why
[29:47] Speaker 11: you're seeing so much excitement and exuberance from the venture community.
[29:49] Speaker 11: And you have Waymo, you have obviously autonomy on the
[29:52] Speaker 11: water with companies like Saronic. So these foundational technologies on autonomy, AI,
[29:58] Speaker 11: are finally starting to make their ways into, let's call
[30:01] Speaker 11: it the real world, the physical world. And that's, by
[30:04] Speaker 11: the way, a big place where we will spend more
[30:06] Speaker 11: time now.
[30:07] Speaker 4: It has been an important year for the firm, so
[30:10] Speaker 4: $ 3. 5 billion of new funds. I try to track where
[30:15] Speaker 4: the venture dollars are flowing, and it's still very highly concentrated.
[30:18] Speaker 4: I think 21 big deals account for the majority of
[30:23] Speaker 4: it year-to-date, and there's a bit to go. How does
[30:27] Speaker 4: you raising those new funds reflect that, that you need
[30:30] Speaker 4: to be in the game now?
[30:31] Speaker 1: On the bigger deal.
[30:32] Speaker 10: So we raised $ 3. 5 billion. A billion of that is
[30:34] Speaker 10: early stage.
[30:35] Speaker 1: $ 2.
[30:35] Speaker 10: 5 billion of that is late stage.
[30:37] Speaker 11: There's been about just over $ 400 billion invested through, I think, June,
[30:41] Speaker 11: end of June this year in venture. And over 80%
[30:44] Speaker 11: of that has gone to late stage deals, just to
[30:46] Speaker 11: give you a bit of a sense of just the
[30:48] Speaker 11: dollars going. And it's, you know, honestly, truly a reflection
[30:51] Speaker 11: of the fact that you see these companies, whether it's Anthropic,
[30:53] Speaker 11: whether it's Waymo, whether it's Databricks, whether it's Saronic and Harvey, Cognition,
[30:58] Speaker 11: Penev and Sierra. These companies are hitting revenues and compounding
[31:03] Speaker 11: at a scale truly we've never seen before.
[31:06] Speaker 10: And so if you.
[31:07] Speaker 11: Have the ability to access those and deploy capital, going
[31:12] Speaker 11: back to Waymo is still early and it's hugely transformational.
[31:16] Speaker 10: It's just an obvious opportunity set.
[31:18] Speaker 1: You know, I misspoke, right?
[31:20] Speaker 4: You got into Waymo in this January round, which was
[31:22] Speaker 4: actually notable at the time. Similarly, like with Sequoia, who
[31:25] Speaker 4: I think co-led that round. But this is a 50-year firm.
[31:30] Speaker 4: How do you go generation to generation at a firm
[31:34] Speaker 4: like KP and be like, okay, this year, this is
[31:38] Speaker 4: something we're doing differently, or it's still in line with
[31:40] Speaker 4: everything we've done historically?
[31:42] Speaker 11: Yeah, I mean, our goal at Kleiner Perkins is to
[31:46] Speaker 11: invest in history-making companies, actually companies that make history, and
[31:49] Speaker 11: it's on our website. And so really what our main
[31:52] Speaker 11: focus is is understanding when the time is right and
[31:56] Speaker 11: when the market is right for a company to break
[31:58] Speaker 11: out and really create... and define a category. And so
[32:03] Speaker 11: a lot of what we spend time on is really that,
[32:04] Speaker 11: understanding technology trends, but then really being informed by the
[32:07] Speaker 11: entrepreneurs who come to our door every day and tell
[32:10] Speaker 11: us about what's happening in the world. And so it's
[32:12] Speaker 11: very much an informed decision, but ultimately it comes down, excuse.
[32:16] Speaker 10: Me, to those founders.
[32:18] Speaker 4: We have a lot more to discuss. Leo Fishman from
[32:20] Speaker 4: Client Effectives is staying with us here on Bloomberg Tech.
[32:23] Speaker 4: And next, we'll probably talk a little bit more about
[32:26] Speaker 4: what people call the trillion dollar opportunity in AI. But
[32:28] Speaker 4: I kind of want to get to what people aren't
[32:30] Speaker 4: talking about right now. The next thing beyond the models,
[32:35] Speaker 4: physical AI. This is Bloomberg Tech. Bye.
[32:40] Speaker 1: Bye.
[32:54] Speaker 4: We're continuing our extended conversation with Kleiner Perkins' partner, Ilya Fushman.
[32:57] Speaker 4: We've talked about the enormous sums flowing into frontier AI.
[33:01] Speaker 4: Now we want to look at what comes next. Because
[33:03] Speaker 4: Ilya's view is that some of the next trillion dollar
[33:05] Speaker 4: AI companies will be built in the physical world. And
[33:09] Speaker 4: that's where I'd love to start. What drives the conviction
[33:12] Speaker 4: on that idea?
[33:13] Speaker 11: Well, at the technology layer, it's really the fact that
[33:17] Speaker 11: you do have companies like Waymo and that have developed
[33:19] Speaker 11: robust autonomy. They've developed models to understand and map out
[33:23] Speaker 11: the physical world. And those are finding their way into
[33:26] Speaker 11: robotics and adjacent domains. And there's these massive industries, whether
[33:30] Speaker 11: you think about transportation, supply chain, factory automation, building the
[33:36] Speaker 11: physical things. You know, I was just in Texas for
[33:39] Speaker 11: the Saronic Port Alpha unveiling last week.
[33:42] Speaker 1: These are massive. We had Dino on the show.
[33:43] Speaker 11: Yeah, these are massive, incredible opportunities. And they're becoming much
[33:48] Speaker 11: more viable in terms of technology permeation into that world
[33:51] Speaker 11: today because of all this work that's been done at
[33:53] Speaker 11: the frontier.
[33:54] Speaker 4: Yeah, at least with Saronic, right, it's tangible. You can say, okay,
[33:58] Speaker 4: they want to build advanced chip yards. domestically here in
[34:03] Speaker 4: the United States? Is it as simple as, and they
[34:05] Speaker 4: can do that more quickly now because of AI?
[34:08] Speaker 11: They can produce products that are much more interesting and
[34:11] Speaker 11: competitive because of AI and because of technology, the autonomy
[34:15] Speaker 11: piece of it. But if you just look at supply
[34:17] Speaker 11: chain more broadly, and we have companies in our portfolio
[34:19] Speaker 11: like Motive or Stored that have really taken a look
[34:22] Speaker 11: at how goods and services move around. How do you
[34:26] Speaker 11: actually inject vision, computer vision, robotics into safety, into performance,
[34:31] Speaker 11: into automating warehouses so that the cost of access to
[34:34] Speaker 11: things for consumers goes down. And that's becoming very tangible.
[34:38] Speaker 11: And then you have, obviously, the transportation sector with companies
[34:41] Speaker 11: like Waymo. So all of these massive industries that have
[34:46] Speaker 11: been around for ages that are truly trillion-dollar markets are
[34:49] Speaker 11: coming up for grabs. And then in automotive, again, in
[34:52] Speaker 11: another big sector, we have companies like Applied Intuition and
[34:55] Speaker 11: most recently Mind Robotics, which is a robotics spin-out from Rivian,
[34:59] Speaker 11: really going after injecting technology into the automotive industry today.
[35:05] Speaker 11: both domestically and globally. Again, all of this is possible
[35:08] Speaker 11: because of advances in these large language models and video
[35:11] Speaker 11: models and hardware.
[35:12] Speaker 4: Those things are happening now. We see evidence that venture
[35:15] Speaker 4: is supposed to be about seeing around corners.
[35:19] Speaker 1: What is the next thing.
[35:21] Speaker 4: That people aren't talking about right now that you are
[35:24] Speaker 4: just convinced? It is the next wave, the next act,
[35:28] Speaker 4: whatever you call it.
[35:29] Speaker 11: Well, I think the investments are being made now into
[35:31] Speaker 11: what will become impactful in the coming years. And part
[35:34] Speaker 11: of it is truly, I think the moment for robotics
[35:36] Speaker 11: is finally now, both because of the hardware costs coming
[35:39] Speaker 11: down and because of model capabilities. There's a domain around
[35:43] Speaker 11: science and kind of the frontier that's becoming very interesting.
[35:46] Speaker 11: Can you automate labs and do materials discovery, do drug discovery?
[35:51] Speaker 11: We have investors in chai discovery, which is helping find
[35:55] Speaker 11: antibodies and make drugs. We have investors in a company
[35:57] Speaker 11: called New Limit, which is going after overall cellular rejuvenation,
[36:01] Speaker 11: so helping us become younger in a sense. And these
[36:05] Speaker 11: things are only possible now and have been only possible
[36:08] Speaker 11: over the past.
[36:09] Speaker 10: Couple of years.
[36:10] Speaker 11: And so that's what I think you'll see more of
[36:12] Speaker 11: these kinds of things permeate into the economy. And then
[36:14] Speaker 11: you might have, you know, eventually robotics in the home.
[36:17] Speaker 11: You might have robotics pervasively through our lives and certainly
[36:22] Speaker 11: in the manufacturing sector in a really massive way in
[36:25] Speaker 11: the near to midterm. And then there's chips, there's the
[36:28] Speaker 11: foundations of AI, and there's AI making AI.
[36:32] Speaker 4: You are still accountable to LPs and your job is
[36:36] Speaker 4: to make returns, right? And we should therefore talk about
[36:39] Speaker 4: the IPO window or what you see. Anthropic is the
[36:44] Speaker 4: low-hanging fruit for that. You also had some M &
[36:46] Speaker 4: A success of late as well, which I find interesting.
[36:50] Speaker 1: What's happening.
[36:51] Speaker 10: Yeah, what's happening.
[36:51] Speaker 11: So first, the IPO market, SpaceX, a great outcome. You
[36:56] Speaker 11: obviously have a few interesting companies on the come. The
[37:00] Speaker 11: window is in some ways open. In some ways, you
[37:03] Speaker 11: have to be thoughtful about how you approach that window.
[37:06] Speaker 11: But overall, I'd say feeling much more optimistic today than
[37:10] Speaker 11: in the kind of historical look back window. The M &
[37:13] Speaker 11: A market is really quite interesting because If you think about,
[37:18] Speaker 11: and we had Salesforce acquire Finn. Oh, and McCabe and Finn, yeah.
[37:21] Speaker 11: Intercom became Finn. Incredible story, by the way. We had
[37:25] Speaker 11: ServiceNow acquire MoveWorks, also a great acquisition. What you're starting
[37:30] Speaker 11: to see is companies that have historically been grounded in
[37:33] Speaker 11: SaaS and traditional software really reinvent themselves through acquisition. And
[37:38] Speaker 11: I think that's a really important point for all these
[37:41] Speaker 11: companies to think about because the talent required... to become
[37:45] Speaker 11: an AI organization is probably sitting out there in a
[37:49] Speaker 11: startup somewhere. And it's really, really imperative for some of
[37:53] Speaker 11: these companies to bring that talent inside.
[37:55] Speaker 4: So let's end our conversation with that. I wanted to
[37:57] Speaker 4: ask you, what's the company you desperately want to see
[38:01] Speaker 4: somebody build, but that's.
[38:03] Speaker 1: Not being done yet?
[38:05] Speaker 11: Well, I think, I mean, I think by the way,
[38:06] Speaker 11: everything is getting built, which is pretty, pretty exciting, but
[38:08] Speaker 11: I do think back to kind of our conversation, you know,
[38:11] Speaker 11: I think the moment you have for the magical moment
[38:15] Speaker 11: you have for something like instinct or muse, I think
[38:17] Speaker 11: that's going to happen in business and enterprise.
[38:20] Speaker 2: Right.
[38:20] Speaker 10: And that's pretty obvious.
[38:21] Speaker 1: Just a tool that becomes ubiquitous.
[38:23] Speaker 10: Yeah.
[38:24] Speaker 11: Just because the thing about it, you know, if you
[38:26] Speaker 11: could clone yourself, right. And you could, Have somebody do
[38:31] Speaker 11: your work for you and save you that time so
[38:33] Speaker 11: you can spend it with family or just do more.
[38:35] Speaker 10: Right.
[38:36] Speaker 11: That would be incredible, incredible for you and incredible for
[38:39] Speaker 11: for society. I think that's an obvious set of things
[38:43] Speaker 11: that will happen. And then I think from that, you know,
[38:46] Speaker 11: the world world is your oyster.
[38:48] Speaker 10: You can do anything.
[38:49] Speaker 1: Yeah, Fishman. Connor Perkins.
[38:51] Speaker 4: It's great to have you back with us on Bloomberg Tech.
[38:52] Speaker 4: Thank you very much indeed. There is a lot going
[38:55] Speaker 4: on in the world. Let's get out to New York
[38:56] Speaker 4: where Bloomberg's Yohara Anand is standing by.
[38:58] Speaker 10: Hi, Yohara.
[38:59] Speaker 1: Hi, Ed.
[39:00] Speaker 12: It's time now for Talking Tech. First up, NVIDIA-backed Lumentum
[39:04] Speaker 12: says its optical components are completely sold out through early
[39:08] Speaker 12: 2029 with customers already locking in orders for 2030. The
[39:13] Speaker 12: company makes technology that helps move data faster inside AI
[39:17] Speaker 12: data centers and says... It's racing to add capacity. Plus,
[39:21] Speaker 12: a nuclear startup co-founded and backed by Vivek Ramaswamy has
[39:25] Speaker 12: raised about $ 400 million at a roughly $ 1. 9 billion level. valuation.
[39:31] Speaker 12: That's according to sources. Atlas Atomics is developing nuclear technology
[39:35] Speaker 12: that could offer a cheaper source of reliable power. General
[39:39] Speaker 12: Catalyst is leading the Series A round, with Andreessen Horowitz
[39:42] Speaker 12: also participating. And Revolut is setting its sights on Wall Street.
[39:48] Speaker 12: CEO Nick Stronsky says the fintech wants to take on
[39:51] Speaker 12: JP Morgan and American Express as it expands in the U.S.
[39:55] Speaker 12: And if Revolut eventually goes public, he says the U.S.
[39:59] Speaker 12: would be its primary listing.
[40:01] Speaker 11: There was always a target given deep liquidity in the U.S.
[40:04] Speaker 6: Market.
[40:05] Speaker 11: So I guess if we do IPO, we'll do obviously
[40:08] Speaker 11: in the U.S.
[40:09] Speaker 5: Yeah.
[40:10] Speaker 10: So dual listing, U.K., U.S., but primary would be in
[40:13] Speaker 10: the U.S.
[40:14] Speaker 5: Yes.
[40:17] Speaker 1: Happy Friday to you, Yahira.
[40:18] Speaker 4: And just a reminder, you can catch more of that
[40:20] Speaker 4: interview in the full episode of Bloomberg Tech Europe, which
[40:22] Speaker 4: is on the Bloomberg Terminal and on YouTube. Okay, coming up,
[40:25] Speaker 4: companies are no longer encouraging workers to maximize their token use,
[40:30] Speaker 4: but that doesn't mean concerns about the cost of AI
[40:33] Speaker 4: are gone for CFOs.
[40:35] Speaker 1: We got more on that debate next. This is Bloomberg Tech.
[40:45] Speaker 1: Thank you.
[40:50] Speaker 4: Workers running up their AI use or token maxing has
[40:53] Speaker 4: given way to value maxing, model maxing, or outcome maxing,
[40:58] Speaker 4: whatever you want to call it. But AI spend is
[41:00] Speaker 4: still keeping chief financial officers up at night. That's the
[41:04] Speaker 4: focus of this week's CFO briefing newsletter. And the author,
[41:07] Speaker 4: Bloomberg's Nita Trentman, is with us. And no pressure, but
[41:09] Speaker 4: like so many CFOs are watching right now. And yeah,
[41:14] Speaker 4: I know many of them are up at night. I mean,
[41:16] Speaker 4: how much of a concern is it token cost broadly.
[41:21] Speaker 13: Yeah, thank you, Ed, for having me. And as you said,
[41:23] Speaker 13: it's basically the single one thing that's keeping CFOs up
[41:26] Speaker 13: at night. It comes up in every interview that I do,
[41:29] Speaker 13: and I speak to CFOs pretty much every day. And
[41:32] Speaker 13: so this, of course, tells us something that even though
[41:35] Speaker 13: companies are embracing the technology more and more, we're seeing
[41:37] Speaker 13: more implementation across use across departments, across use cases. We're
[41:42] Speaker 13: also seeing that the main question around costs and how
[41:46] Speaker 13: does it compare to productivity and savings that you generate
[41:50] Speaker 13: as a business, that that's still an open question for CFOs.
[41:54] Speaker 4: I mean, spare a thought for CFOs everywhere. Like the
[41:56] Speaker 4: Jensen Wang view on this, the CEO and founder of NVIDIA,
[41:59] Speaker 4: is like he wants to see his engineers spending tokens
[42:03] Speaker 4: with no limits as fast as they possibly can. But
[42:05] Speaker 4: what are the strategies that the CFOs, but the companies
[42:08] Speaker 4: more broadly are using to budget, like to manage it?
[42:12] Speaker 13: Yeah, so they always talk about having an envelope for it.
[42:15] Speaker 13: So basically sort of that's their figure that they want
[42:18] Speaker 13: to start the year with. And then they're saying that
[42:21] Speaker 13: they are flexible in terms of like how they're allocating
[42:24] Speaker 13: that amount across teams. So engineers usually get the most
[42:27] Speaker 13: amount and then you can ask for more tokens if
[42:30] Speaker 13: you need to. At some companies they're saying, well, we
[42:32] Speaker 13: actually want to see how you use your existing budget
[42:35] Speaker 13: of tokens and whether that was a productive use. And
[42:39] Speaker 13: some companies are also just resorting to increasing the budget
[42:42] Speaker 13: as we go along. Match Group, for example, they started
[42:44] Speaker 13: the year with $ 5 million as a budget for their tokens,
[42:47] Speaker 13: and they doubled that throughout the course of the year
[42:49] Speaker 13: because they realized, okay, what we thought initially wasn't sufficient.
[42:53] Speaker 13: Of course, also model orchestration is a big focus for
[42:56] Speaker 13: companies to say, okay, what are the models and what's
[42:59] Speaker 13: the right.
[43:00] Speaker 1: Task to use this model for?
[43:03] Speaker 13: But it's still, I think companies are trying to figure
[43:05] Speaker 13: out how this works best and depending on how exposed
[43:08] Speaker 13: to AI they are, some are better at it than others.
[43:12] Speaker 1: Hey, we're in it right now.
[43:14] Speaker 4: Companies with a calendar fiscal year, they budget late summer
[43:18] Speaker 4: to fall. They must all be there right now, like 2027.
[43:20] Speaker 4: Token maxing no more. Bloomberg's Nita Trentman, thank you so much.
[43:25] Speaker 4: Check out the CFO briefing newsletter. It's coming out this
[43:27] Speaker 4: Sunday afternoon, lunchtime, Pacific time, early evening, New York time.
[43:34] Speaker 1: I'll put it out there on the socials. What a week, eh?
[43:38] Speaker 1: That does it for this edition of Bloomberg Tech. A
[43:40] Speaker 1: lot of you are.
[43:42] Speaker 4: posting at me on the socials about the top of
[43:44] Speaker 4: the show and the explanation of ARR. So thank you
[43:47] Speaker 4: for that recap, that conversation with Rebecca Torrance on the podcast.
[43:51] Speaker 4: You can find it on the terminal as well as
[43:52] Speaker 4: online on Apple, Spotify, and iHeart. Thank you so much
[43:56] Speaker 4: for being with us all week long here on Bloomberg Tech.
[43:59] Speaker 4: From the city of San Francisco, this is 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.