Bloomberg Audio Studios, podcasts, radio news. Bloomberg Tech is alive from coast to coast with Caroline Hide in New York and ev Loow in San Francisco.
This is Bloomberg Tech coming up.
Amazon kicks off what is likely to be one of the biggest corporate bond offerings ever.
In a bid to pay for the AI boom.
Plus, Google introduces AI agents across the Pentagon's workforce to automate routine jobs. This is the Defense Department's fight with anthropics drags.
On, and Hewlett Packard Enterprise projects sales topping estimates on AI hardware demand. We'll break it all down with the CEO, but now we break.
Down volatility in the market.
We once more focus in on the Middle East, the conflict with Iran and the fact that it has been upped in terms of aggressive force coming from the US to Iran today, but also aggressive moves in oil markets.
Just drag down Brent.
Down ten percent, significant volatility in the oil contract, but it means we're actually seeing some push higher in the stock market, a regain on the NASAQ one hundred. In fact, that a second straight day of games ed all of this as we think about really the end goal and whether this will be a short term or long term conflict.
I'm looking at the latest and what's been a wave of AI related bond sales and it's Amazon. So this is what Bloomberg's reporting, both US dollars denominated and Euro denominated bonds in US eleven tranches of high grade the longest duration I think twenty fifty six with a one point five to five percent yield over treasuries.
But the goal is to raise up to forty.
Two billion dollars, maybe up to ten billion euros. And we know the story, right, capital expenditures this year six hundred and fifty billion dollars or so across the hyperscalers, and many of them have looked to debt markets, looked to raise money through bond saleales to finance that capex.
There's a lot to dig into here, there is, and.
We're going to continue the conversation. Therefore, Edward Robert Schiffman and bretting Bag Intelligence, he's out with the reaction to this morning, saying that the bond cell is quote likely to draw a considerable demand despite a growing absolute debt load that's poised to increase again next year as spending rises. That the stock had been punished when they told us that they were going to be spending up to two hundred billion on capital expenditures.
But the bond market is going to lap this up.
It feels like, yeah, you know, My first thought was, Wow, they must be sewing.
A lot of books.
My second thought is, this is so wildly bullish for the credit markets and in particular for tech, impossibly the worst potential macro environment. Bondholders are coming in saying we want to own you. We want to own you in size, and we're confident that your future growth plans are going to play out.
And even though.
There's so much spending and we know more bonds are likely to becoming, people are going to be lining up for this deal.
I actually just misspoke at the board. I said twenty fifty six, Actually twenty seventy six, the longest maturity that Bloomberg's reporting Amazon's thinking about, the one that would yield one point five five percent over treasuries. How does this kind of stack up in the history of corporate debt markets, recent and historic? Right, we talked a little bit about what Alphabet's done most recently. Oracle is a slightly different credit profile, but like a lot of focus on that too.
Well, this is going to be the biggest bond deal of the year, it looks like by far, and I'll tell you, I think people just need to get used to this. These hyperscalers are effectively becoming serial issuers. The one hundred billion dollar plus debt club is growing. So not only do we have new current larger needs, but just think about it, over the next handful of years, as bonds mature, there's always going to be a constant
need to refinance and push out the curve. Amazon itself is twenty billion dollars in debt maturing over the next three years, so this is going to be a constant and look for more of this action. Again, I think bondholders can take huge advantage of this. The type of concessions that are being offered here seem wildly wide to me. I actually think the demand is going to be so strong that they're going to come in dramatically, maybe twenty thirty plus basis points, and.
They're tapping just every type of part of the market. You've got like nineteen.
Different tranches, so different inspiration dates. Ultimately of when the debt is going to becoming jew but also different currencies as well. How are they managing to satiate all the type of investor out there?
You know, we talked about this when Alphabet came that they were doing a small tranches of Swiss francs and British pounds and why they weren't tapping the Euro market. You know, the r market is widely I think underinvested in technology, and they're somewhat probably price agnostic. They're willing to take down any Amazon bonds across the curve almost
at any price because they're so underweighted. So I think there's a huge on tap market there that we're going to see other companies come and grab as well.
That's the Bloomberg Intelligence react from Robert Schiffman. I would say the details of Amazon's bond sale are according to Bloomberg sources, and Bloomberg saying that the US side at least mike price. Today a separate story, a court ruled that Perplexity must stop using its Comet web browser agent to make purchases on Amazon's marketplace. Amazon sued, accusing Perplexity of computer fraud for not disclosing when Comet was shopping
for a user and refusing to stop when asked. The order also bars access to password protected parts of Amazon systems and requires destruction of Amazon data, with enforcement pause for a week to allow for an appeal. Perplexity has yet to respond to a messages seeking comment.
Cara, it's avas company. Then, of course it's about e commerce and about cloud. Let's talk about Oracle now, because it's about cloud and software and its earnings are after the bell. It comes after recent Bloomberg reporting that Oracle and OpenAI have ended plans to expand that flagship data center project in Texas. While will investors be looking for in the company's results, We want to talk it all through with Gabrielle Borge's software sector analysts at Goldman Sachs Research.
How much of this is a test of Oracles all bread and butter software or a test of how much investors are willing to well stom at the capex of getting into the cloud and furthering the AI focus.
Yeah, Hi, it's great to be back. Good morning. I think you touched on it perfectly there where the capex makes sense if you can justify the unit economics that are attached to the AI infrastructure projects, and so Oracle's given us a couple of nuggets in this regard. They've taught us that they expect to be able to spend less than one hundred billion in debt raised in order to be able to fulfill north of five hundred billion
in cloud backlog that's now on their balance sheet. Those are commitments from costomers for the next five six years. They've also taught us that the gross margin on these projects is thirty to forty percent over the lifetime of a contract, so that's typically five to six years. And so if you put those two data points together, whatever incremental information we can get from the management team tonight that helps make this less of a black box, I think we'll go along way.
Because we are worried about turning cash flow negative, We are worried about maybe some of the ability to drive revenue growth to substantiate the capital expenditure spend. What do you need to see in terms of growth and pointing to growth because the stock is actually pretty cheap now.
Yeah, So when you actually break down the phases of the different facilities coming online, so Abilene, for example in Texas, that's a one point to big our capacity facility where Oracle has already said that the four point five gigawats that will be coming online to fulfill their existing commitments. They already have all the leasing agreements for that already
signed and ready to go. What happens next is as the revenue comes online, we expect to see is really nice gross profit acceleration in Oracle's four co business that includes the big part of OCI that's tied to some of these contracts, that's Oracle Cloud. We've spect to see
that as we go through the year. And so what we've noticed over the last three or four quarters with our core is that it's less about the specific numbers and some of the short term dynamics in the given quarter and more about what they say about their ability to fulfill some of these contracts over a longer time frame, which is why just taking a step back, I do think you'll see BREST profit acceleration and US as we'll be looking for that as we go towards the end of this year.
Gabriella, I hear you on that, but you know, the streets expecting the infrastructure unit to grow almost eighty percent in the quarter, you know what if they're off by a few percentage points, because you know, in general that's been the formula, right, a very high bar for tangible AI related growth in those infrastructure divisions.
Yeah, I think you're making a really good point here, because we have seen for a couple of quarters now that the OCI estimate intra quarter can move around, and part of that is the timing of any one piece of facility coming online. I do think though, what we're seeing from a land signal standpoint is still very strong. And coming back to Abiline, Abiline has been rapping really
nice over the last three months. The one other thing I'll say is ourcle stock does react to apros our potential delays for example, so as we've gotten more headlines out out of the new flow over the last two three months now, I think expectations do move around. All of which is to say what we've seen during software earning season has got to have a clean print for the stuff to work on the print, and so and I hear you there, it's got to be a clean print for the stock to work on the print.
What I reported with the team in the case of Abilene was that simply Oracle and Crusoe decided not to proceed from with an expansion from one point two gig watts to two Google watts, and that Crusoe was negotiating with Meta to come in and take that additional capacity, which many of your colleagues on the cell side pointed
out might not have been reflected in RPO anyway. But what the story does highlight, and you made the point a moment ago on the four point five gigawatts agreed, is concentration risk one customer, which is open AI.
Is that something you model for, Yeah, we.
Do actually in a couple of different ways. So we take a discount to our open AI forecast and to what ourcle has said their open AI backlog is to account for dynamics like the one you're talking about, the extreme customer concentration risk. The other thing I'll mention here though, is Oracle will tell you that they can have fungibility of CAPEX. What does that mean? That means you can actually reconfigure a training cluster to be able to serve
inference use cases. It means you can reconfigure for one customer to serve another customer, so there is some fungibility there. What that means is that if AI demand remains strong and that internalizes on enterprise AI option, which we've spoken about before. Then someone should be able to come in and pick up that capex in the event that it needs to be moved around and it comes back to the same air of fungibility.
Gary Ellavorg is Golm and Zach's research analyst looking ahead to Oracle, which is after the bell.
Thank you very much.
Like coming up, we can to have the latest on each expanding use of AI by the Pentagon.
That's next. This is Bloomberg Tech.
As President Trump declared yesterday, we're crushing the enemy in an overwhelming display of technical skill and military force. We will not relent until the enemy is totally and decisively defeated.
That was US Defense Secretary Pete Hegseef weighing in on the war in Iran during a briefing. The Pentagon says it's conducting the most intense day of attacks yet. This is the Islamic Republic has been firing drones and missiles at targets across the Middle East. In return, let's bring in Bloomberg's Washington correspondent, Tyler kendo I saw a lot of the headlines on the terminal this morning, for example,
around drone activity in the Gulf the UAE. We continue to hear from this administration on or at least field questions on the timeline for this war.
What's the latest we need to know?
Yeah, hey, and good morning.
Well, as you heard, their secretary Hexcess said that the US will quote not relent until all of its objectives are achieved. After yesterday, President Trump appeared to suggest that the US was ahead of schedule when it comes to
maintaining its goals in the region. But then we heard from Iran this morning and a new statement from the country's parliament speaker saying that Iran is quote not at all interested in a potential ceasefire deal, and the IRGC released a statement overnight vowing to continue to block oil exports.
Now.
President Trump yesterday had threatened to widen the list of US targets to electricity infrastructure if attacks in the Strait of her moves and disruptions do continue. I will point out that the President reiterated that he is prepared to send a naval escort into the Strait to help those tankers through. But when the Joint chiefs of Staffed Chair Dan Kane was asked about this earlier this morning. It does not appear that any of those plans have been
put into motion. Bloomberg News is also reporting that electronic warfare is now jamming signals in the Critical Waterway, which at this point really does remain effectively closed, and in Caroline. Within the last hour, we got the headline that G seven nations are now asking the International Energy Agency to prepare scenarios for the potential release of emergency stockpiles as these oil flows continue to be impacted in the region.
And the oil market certainly moves and despite the newsflow, Tyler Kendall, we so appreciate it. Let's stick with Defense News because Bloomberg has learned that the Pentagon plans to add Google's AI agents to handle unclassified work and as in talks to bring those tools to the classified and top secret cloud. Bloomber's Katrina Manson has been importing on this and in fact continues to report at the very
edge of AI adoption within the Pentagon. Look, what is Emil Michael talking about in terms of Google.
Is this because they're having to.
Force out anthropic This I think is fair to see as separate where Emil Michael came in in this senior role very focused on AI. Back in August, he told me he was shocked that the Pentagon had adopted so few AI tools. It's brought in chatbots, that's thanks to Gemini. It's had those for the last three months. One point two million people in the Pentagon have used those or
in the Defense Department writ large. Now they're adding agents, so that's going to automate some tasks that corporations, of course across the country are already trying.
But it is.
Interesting timing because at this very moment they're bringing Gemini agents onto unclassified networks and really quite pleased that they have this partner in Google at a time that clearly Anthropic has been cut.
Twenty four hours ago on this program, we broke the news of Anthropic suit against the Defense Department. Late yesterday evening we got more of a detailed response. What is the Pentagon's official position and what does it see in this suit from Anthropic against it.
Emil Michael told me that he was expecting this. He doesn't think that bringing this to the courts is a way to resolve this problem, and he said, call me quite simply.
We're moving on.
So I do think that in this news today the focus on Google. Of course, they've recently struck deals with Xai and Open Ai to start operating on the classified cloud. The Pentagon is really saying, we have three partners now that we feel comfortable with and we are not expecting to go back to anthropic.
Can you give us the context of just how much AI is playing a role within the very here and now geopolitical issue of Iran and the conflict there.
From our reporting, we understand the scent the central US central commander conducting the operations in Iran is relying on a variety of AI tools, and from separate reporting we understand that means mayben smart system. This comes out of Project Maven, an AI effort started in twenty seventeen.
You have a book coming out on I Do Yes, I Do Yeah.
Thanks, and that helps narrow down time targets. Of course, SKom is hitting I think they've hit five thousand targets in ten days now, so an extraordinary number for the US to be churning through. And at the same time AI is helping with processors that enable fast movement.
Bloomers Katrina Manson with the latest reporting deep on what the Pentagon is doing with AI.
Thank you very much.
Now coming up, AI start up Legora raises five hundred and fifty million dollars, boosting its valuation to five point five to five billion. We speak with the CEO, Max Unistrand. Next, this is Bloomberg tech.
Logora, an AI platform built for lawyers, has just raised a massive five hundred and fifty million dollars in a series defunding round, pushing its valuation to an excess of five and a half billion. Joining us now to talk about the raise, the company's growth in the future of AI and laws Lagora CEO Max Unistrand, Max, this is only a few months after you just raised your previous series. I mean, what are you going to be doing this using this money for?
We are investing in the United States. We landed in New York a year ago. This was the first week in twenty twenty five New York Legal Week when Lagoire launched in the United States. Now we've opened in New York, We've opened in Denver, Chicago, Houston. We're investing in the product and infrastructure, in customer support teams to support the astronomic demand that we're seeing from the US, and that is.
What the vcs are responding to here, the fact that you are signing more and more legal professionals to come and join your platform. But yesterday we just had Harvey AI on and what has it that they're a competitive force out there? How are you seeing your space in the world? How are you unique versus a Harvey for example.
So I think legal tech has for a long time been sort of stuck in the pre digital era.
With llms. We're seeing a wave of.
Companies come and automated tasks, build new systems, and it's a Renner songs for the industry. So there's lots of companies going very quickly. There's a lot of green field opportunity. We started in Stockholm, Sweden, just three years ago, and now we've landed here. We're expanding with big law firms, with big enterprises, and I think the reality is that
this market has only done zero to one. Now we are on the journey of one to ten and ten to one hundred, and there's so much left to build and that's our opportunity and our challenge.
Max.
What's interesting about what you just outlined is different jurisdictions. Right, You're talking about American law, Nordic law, European law.
Take us inside building.
Legora and the necessary data sets to be able to service the legal profession in those different markets where laws and fightings are completely different.
So in order to build systems and agents that perform real work, there's really two different types of data that you need to work with. As you said, it's the local jurisdictional data, the case law, the legislation, and then it's the firms and the enterprises own data. And it's the combination of the two within a platform like Ligora that becomes incredibly valuable because I think these systems used
to kind of feel like co pilots. You were working together with them on small tasks, but now you're actually able to offload entire end to end I mean agentic workflows, and the systems are performing and to end work for you.
So when we started.
Out in Stockholm, Sweden, I mean that market is smaller than one of the big law firms here in New York. So we really practice that muscle. And now that we're here, I'm an engineer, we really lid with a product.
Max I suffered through law school and I'm the son of two lawyers, and I brought that up yesterday, so I'll be consistent. And I know from that context the lawyers build by the minute, sometimes scanning barcodes for five minutes of work fifteen minutes work. How do you charge and structure your revenues for the platform?
So which charge on a set basis? Today?
I think that that is the easiest way for some of these customers to buy. But as we move towards this world where leguore is performing more end to end work, I think both consumption and outcome based pricing models will make a lot of sense for our clients and also in turn their clients. Right the way that law firms are charging for tasks that AI can perform together with them will have to move from a buildable hour model to more of a fix thier outcome based system.
Can you just set the scene here, because you're building upon in many ways the innovations of Anthropic, but Anthropic is also doing its own claw plug in for law. How does this competitive environment unfold for the future businesses you want to sign on? How do they decide between you and Thropic harvey what they already have as point solutions?
Yeah?
So I mean Legorre wouldn't exist without the llms. We were founded in late spring of twenty twenty three off the back of the release of GPT three point five. That was the first time the models became good enough to build any real software. In the beginning, there were so many issues around the models. You had to reduce the hallucinations, you had to guardrail them, We had to build a lot of systems around them. Today we are very focused on bringing the capabilities of those models into
an enterprise legal setting. So when we work with the law firm like White and Case or Clearly Gottlib, they are deploying Laguera and they're turning over, you know, billions of dollars using our system. And the interesting thing for us is every quarter the ground shakes a little bit. The new models are capable of doing new things, which
needs to be reflected in our product roadmap. We cannot plan for a year ahead because three months, in six months the capabilities might have changed, so our product has to change.
Max.
We just had twenty seconds. But I think what Caroline's question is, what is your modes? Do you have a mote?
I think we have a real mold.
Of course, I mean the way that we're deployed within this enterprise is connected to their data, the way we've set up the permissions, whether we work with delivering work. For instance, Deba Boys just launched their client portal on the World where they are connecting their work to their external clients. So there's a real network effect rates in our product tool.
Max and Strand of Legora. Thank you very much for your time here on Bloomberg Tech. Now coming up, Apple's latest product is delayed because it waits the debut of its new AI seri. We have more on that Bloomberg reporting and some supply chain coverage as well with what's going on with Apple. It's halftime. Don't go anywhere, We'll be right back.
This is Bloomberg Tech. Welcome back to Bloomberg Tech.
Technology stocks actually in particular now pushing higher having seen then as that one hundred lower earlier in the session. Actually that chart tells the story pretty well, but a lot of the story is still what's happening with the war in Iran, the movement in oil, general risk sentiment around geo politics. After the bell we get Oracle earnings, and so once again we kind of focused on the AI story. And capital expenditures. Amazon is up by almost
half percentage point. The big story of the day that we hit earlier in the program is them looking at a bond sale US dollar denominated and Euro denominated of up to forty two billion dollars. And we had the details on the program earlier. Oracles treading water of course ahead of its print later today. But that is the AI story once again, the credit conditions, the capital expenditures, and whether this build out is intact Carrott.
We've got plenty more on the AI story and supply chain stories now because Apple's manufacturing pivot to India and away from China is picking up momentum. I've phoned production in the country SAG fifty twenty five and now accounts roughly a quarter of all.
iPhones made globally. For more.
Blomberg's Consumer Tech and Apple Managing editor Mark German joins us now, and Mark, let's start with the supply chain element of things. They've done this swiftly and effectively.
Yeah, our colleagues in India reporting that now about one in four iPhones are produced in India. This is a bit of an acceleration. About a year ago it was one in five iPhones and a lot of the phone production in India. That pivot happened because of tariffs about a year ago. Apple looking for ways to produce the phones without that potential tariff hit for devices built in China and then imported.
Into the United States.
And they're going to continue doing more in India. They've got AirPod development happening there at some point, more accessory development happening there, more underline component development happening there. So India not only big for Apple in terms of growth in retail, but also for underlying manufacturing, not only in India but in other parts of the world as well.
The other piece of reporting you have out this morning, Mark is about the Apple Smart Home Display. Now in my house, I right now have a lot of Amazon ex devices, and so my understanding is that the Apples Smart Home Display is like most akin to Amazon's Echo Show. What are you learning? Why delayed? What's behind it?
Yeah, this device very much based on the new SERI and the new AI features Apple has been working on. Those features have yet to come to market, those continue to be delayed, and as those features get delayed, the hardware gets delayed. Because, as you know, Apple's very much a hardware and software and services integrated company. It's all one kit. It all comes together, so if one piece is not working, you can't release the entire thing. It's
a pretty nifty device. You can walk up to it and it can know who you are and give you personalized content. And that very much relies on those new serie features that the company announced a couple of years ago. They're now aiming to roll them out before the end of the year. And so this device, instead of launching this month, likely to launch around September.
The famous Mark German and all things Apple, the Apple summary, Thank you very much. There's a lot more news out there, Carrot.
There is ed and it's time now for talking tech and first st up by d is exploring options to enter the world of F one racing. According to sources, it's a bid to boost the ev makers global appeal. Now the move would mark a rare attempt by a Chinese manufacturer to take on a sport dominated by European and US teams.
Plus China's enthusiasm.
For all things open, Claw is igniting a stock rally among local tech firms moving swiftly to embrace the open source AI program, and I remember Openklaw is an agent that leverages llms including anthropics Claude to perform daily functions and the tool has garnered a cult like status in China. And Polymarket is enlisting firms including Pan Andeer to help police its force contracts. This is a prediction market faces
really intense scrutiny over inside of training. It's all according to the sources, who say that the firms will prevent and report suspicious activity with a monitoring system. The Polymarket is building.
Out ed Okay, coming up, a three month old startup founded by AI pioneer Yan Lacoon has raised over a billion dollars in seed funding. I'll tell you about the company called Advanced Machine Intelligence as next. This is Bloomberg Tech, a startup founded by French AI pioneer Yan Lacun has raised over a billion dollars in its initial funding round. Advanced Machine Intelligence or AMI, is focused on building world
models that use visual and spatial data. Bloomberg's venture reporter Natasha Mascarinus broke the story and joins us, Now the Internet is kind of reacted by another one and what is it that they do? But you know, there's a history with Yandlercun and his work with Meta in particular. Let's start with the vision for AMI. What is it they're seeking to break through on?
Sure? I mean Yan Lacoon has been a long critic of LM's being the best way to advance AI. So a ME is a new company that'socused on world models. The idea is that they are going to be able to integrate with enterprises like robotics companies or car manufacturers and help build AI that can operate in the real
world versus being something that consumers everyday. Consumers interact with via chatbots, and so let's focus on text and think more around spatial data, visual data and trying to operate again in real world and real industries.
And maybe even the future of defense. There were so many interesting parts to this story, Natasha that I want to dig in on, but won't call my eye as someone who sat in New York, they're not going to Silicon Valley. This is because they really feel that there's what did he call it, that they've been LLLM pilled there. This is a European story at it's hot, but it's also a non Silicon Valley one.
You're exactly right.
I mean, this is one of the largest venture rounds raised ever in Europe and it is actively one that has not included a lot of the traditional Silicon Valley investors.
When I asked you about that decision, he said that Silicon Valley investors were looking for really high ownership in the business, and he believed that it was important to try and to check talent that wasn't just focused on building within the large language models, which means not going to where people are working for Anthropic Open AI, even
his former employer Meta, and trying to go to different geographies. Now, I did ask if he will have to raise from Silicon Valley eventually, and he said he's definitely open to that for his next round.
Lakuma is at Meta for almost fifteen years, more than ten years at least. Is there any evidence of reporting that that that's the way they go, some kind of relationship with Meta, some collaborative work something like that.
Yeah.
So at this point, Meta was not listed as an investor in the new business, but we did ask him about possible partnerships, which he did confirm. So while the average consumer may not interact with his new company through a chatbot, there's not a too far of a world where we could see it showing up in the meta ray bands or other AI ware of all the meta's working on. So we're definitely tracking the possible partnership and you can birm that there are ongoing discussions there.
Natasha Mascarinis just scoop off to scoot. We so appreciate coming on without reporting. Let's not talk about another funding or indeed relationship.
In Video is investing in Thinking Machines Lab.
It's an AI startup founded by opening Ey executive mirror Marati now and Video is set to provide it's Vera Ruben systems to power Thinking Machines AI models as part of a multi year agreement. Blomberg AI reporter Rachel Metz joins us some more so, what is Thinking Machines Lab doing?
Yeah, so this company, they are building their own AI models and they have a deal that they just announced with in Video. They're getting investment from the Video and they are also going to be using quite a lot of in Vidia's chip power to train and run their AI models.
I posted your story on X and you know noting that up to a gig or what ver Rubin or a minimum of a giga what sorry, ver Rubin systems like, that's a lot of compute.
But the unknowns are.
Is in video making this investment in cash or chips in lieu of cash.
We just don't know.
And the response that everyone gets back is, so, what does thinking Machines do? What is thinking machines other than AI?
Okay, so so far, Thinking Machines has released one product, and it is called Tinker, and it came out last year, and that is a software that companies can use to fine tune AI models. What they're working on as well is some AI models. It's not quite known yet what
the details of those are. It sounds like they will be things that will be useful to a range of enterprises, but it's not entirely clear yet if they're going to be focused on specific uses or if it's going to be a more general kind of thing like the models that underpin say chat GPT or.
To have just issued or developed a model called Tinker and already the talked of as in the realm of a fifty billion dollar market capitalization, it's on us to remind ourselves why Mira Marati is taken so seriously by this market for a hot second, she led open Ai, but really she was the CTO, right, Yes, she was.
The CTO Opening Eye. She's worked add a number of tech companies over the years, so she's been in the industry for quite a long time. I think the challenge that she's facing here with this company is can she build large language models? Not just can she build them? A handful of companies can build them. They require obviously a lot of money, a lot of expertise, and a lot of data, but also can she then convince companies or some other form of customer to pay for them.
Tinker is an effort to help people fine tune models, so it's not quite up to having It's not the same as having a model and then having people use it. But that was sort of their first step into business, and they do have customers for that, so it'll be interesting to see over the next year or so what happens with their business.
We'd reported at the end of last year the Thinking Machines was going out doing an early round at like a fifty billion dollar valuation when it was basically a handful of people, and then what followed was all kinds of media reports about internally things not being well right, people leaving. We never really got to the bottom of it. I mean, this is a big piece of news for them, but we have a sense of how it's going otherwise beyond getting the backing of Nvidia.
Yeah, that is a bit of a question mark. We know that the company has about one hundred employees and it has, as you mentioned, it recently lost some employees who went back to Opening I. Initially, when Mira Murradi started this company, she started it with a quite a number of people that came from Opening I, so that it was certainly a signal if people were going back
to that company that they had come from. But I mean, I think this is still a very new company, and sometimes these things take time to get going, get off the ground, and we'll see what happens throughout the rest of this year.
Bloomberg's Retro mets top Reporting. Thank you very much.
Coming up on the show, HPE exceeds expectations on its forecast for the current quarter. It's all about AI demand and all about AI hardware demand. We can discuss what's driving that outlet next with HPE CEO Antonio Neeri.
This is Blomberg Tech.
Shares of HPE modestly higher, up about six tens of one percent. Been chopping around in the session, but the company gave a strong sales and revenue projection for the current quarter.
What's driving it?
Demand demand for AI hardware in particular, has discussed with HPE CEO Antonio Nery, and let's start with that. You know, clearly there is some staying power in that demand at rack scale, you know, crystallize it for us.
What are you seeing in real time?
Yeah, good morning, Ed, and thanks for having me. We saw a tremendous demand throughout the quarter. I think the momentum in I continues in the build out of the data center. But what I'm really pleased is the momentum we see an enterprise. In fact, most of our revenue in AI that we're recognize in Q one came from the enterprise segment, which shows you two key elements. One is the adoption of agentic AI into the company's workflow,
and second is the growing of influencing. A lot of focuses always around the training and training these new models, but for us, our focus being both the sovereign space and the influencing and enterprise space. But in our portfolio, we saw tremendous momentum networking. Our strategy where journeys has paid off, and that's why we see double digits year over your growth in the order intake.
You know post Juniper networking is is upside for you, right. I think you have a lot of confidence that growth will come in networking. For the audience that doesn't understand what it is you do in that space. Why is networking on the rise in parallel with just kind of the broader AI hardware demand that you're seeing.
Well, I do believe ed and this was the coresis of the Juniper acquisation that the next inflection point in terms of disruption will come from the networking connectivity layer. Think about when you build a data center, you have to connect this data center to other data centers through the Internet, and that's what we call data center interconnect. Juniper has one of the most innovative products in the route in space, by the way, this passport that we
grew mid twenty percent in the order intake. And then once you are inside the data center, you need to be able to connect large amount of GPU use and CPUs together and that's why you need a data't set. Data doesn't resewarching portfolio in addition to some of the other core tenants of the technology and there you know, the Juniper portfolio group mid forty percent in the order intake. So I will say for us, that has been a
core tenant. That's why you know now they represent almost thirty percent of the company revenues but more than half of the profit. And for us, that's also an ability to raise the outlook for the remainder of the year. And also they're outlooking free cash flow because the structural margins of that business and the working capital efficiency are very, very different than the rest of the portfolio.
I mean you actually said, look, you're not done raising prices. You talk about discipline and what is a very dynamic environment, Antonio talk.
To us about well, how you're weathering.
The supply chain issue, in particularly the cost the memory costs at the moment.
How do you navigate that?
Yeah, Carol, I mean, obviously the demand and supply has a huge mismatch, and you actually need to go back all the twenty two to twenty three timeframe to understand how we got here. Accelerated by the hypercycle we see in the AI space, but fundamentally in our guidance, the new guidance, the new outlook that we raise for twenty twenty six. We included our ability to see the supply that we need to convert that into a revenue profit.
But the reality is that we do not have enough supply against the order intake and the backlog, because otherwise we'll have even a higher outlook. So what we're doing we have taken three very unique steps. Number one is securing as much supply as we can, but again we don't have all the supply that we would like to have. Number two is taking a very agile posture when it comes down to pricing, and as I said yesterday, we
are not done raising pricing. I think that the cycle will continue well into twenty twenty seven, although we will hope that we will reach some sort of elevated price in stability in twenty twenty six. And number three, you know, my experience has thought me that ultimately you need to have direct, transparent conversation both with customers and our partners.
Last week I was in Europe. I met more than twenty customers the Mobile Congress in London, and they appreciate that level of transparency, understanding the environment we're in.
I mean, you're actually forgoing supplying some equipment to mobile service providers, So you're being choosy with who you work with at this time of geopolitical angst and when we question how much the Middle East is going to be able to build out the data center capacity as once thought. Has the Iran conflict impacted your view on the world or indeed the supply of the world and chips.
Not on the supply side, Carol, And obviously our first priority because we have businesses in that region, whether ue Qatar, bar In, sauda Arabia, Israel, right is the well being the safety of our employees. We have approximately one thousand employees in that region and the good news all are doing well at this point in time. But from a supply perspective, no, But obviously as the conflict may gets elongated, they may have other implications, particularly on the logistics side
of the equation. You know, look, air freight routes are a little bit more complicated. We don't use boats of ships for transporting our equipments. But over time we have to assess if there is a revenue impact. What I can tell you right now there is even more demand then we saw particularly you know, three four months ago, so we have to navigate that together. We have to be smart about it.
Antonio.
You have never let me tear apart an HPE server, and that's okay. But if I got my hands in the compute trade, you'd have the high bandwidth memory around the GPU wherever it comes from. You'd have the ddrs around the CPU, and somewhere you'd have the SSDs.
And flash memory. Be very very specific.
Of those three, what's the biggest impact to you right now in constraining your sales and your outlook, very simple.
Is DDR and nand or the SSDs. Hbm's less constrained, but obviously a lot of the allocation on the capacity has moved there. And as you know, there is also a transition from the prior generation of hbms to the new generation of HBM because you need more membody channels. But it's really the DDR four and five which now nobody almost use it, but DDR five and the nand part which is the SSDs.
You said just earlier.
Look, one of your the three key steps you took was to get as much supply as you could and to ensure that that was locked in. You wish you could have more Who is it that you need to push more? Who is it do you need to have more security for? How can you perhaps get hands on further supply going forward.
Look, we have a long standard relationship with the three core suppliers that provide DDR memory and hpms, and there is a little bit a larger ecosystem for the land space, but we're having weekly engagements with them. We are looking to swap components driving different configurations, but the reality all of them are significantly constrained.
Karl Antonio just twenty seconds, what's the dollar figure dollar terms, and of what was left on the table, how much better the outlook could have been.
Yeah, it's pretty sizable. I will say, you know, we guided five pennies on the midpoint of the range and then we raise our outlook for free cash flow by almost two hundred million at the midpoint. But the backlog is very very large at this point in time. If you think about the AI system backlog is over five billion. We raise again the networking for networks for AI goal for the end of the year. So it's a very
sizeable backlog. So we hope we make progress as we go along, but we categorize, we leveled it as a proved guidance at this point in time.
Anthony Nari, CEO of HPE, thanks for the candid conversation.
We appreciate it. Now that does it with this addition of Bloomberg
Tech ed, great interviews, great Bloomberg reporting, recap it on the podcast new Where to find it on the terminal and online on Apple, Spotify, and iHeart this is Bloomberg Tech