Bloomberg Audio Studios, Podcasts, radio news. Bloomberg Tech is live from coast to coast with Caroline Hide in New York and ever Low in San Francisco.
This is Bloomberg Tech coming up. All eyes on in video as a chip maker reports sarnings after the closing bell. What to expect from the world's most Valuable company? Throughout this hour, plus SpaceX's Starship hits a major milestone after deploying satellites in space for the first time. An IBM and AMD partner up to develop quantum centric supercomputing. We'll discussed with IBM's Quantum VP, but first we check in on these markets, and actually we just check in the
world's most valuable company. We check in one that dominates the SMP to the tune of eight percent. I'm looking at a ten percent higher for InVideo.
We are just one.
Percentage point away from a new record high for this company. Will it get there after the closing bell, after the market shut In that post hours movement, we could see whitsh lash of up to six percent either direction, according to the options market. I'm looking at the waitings because this is why we focus on this company and this is why it's basically our super Bowl today. Eight percent of the s and P five hundred, we are twenty
four percent of the tech sector within that SMP. Then aw's that one hundred now we're seeing in video make up ten percent of the waiting.
And so therefore we go to a man.
That has covered this stock before you and I could pronounce in video properly. It is the one and only in King, it's your super Bowl tonight. Ian And just so many questions remain about the forecast. It feels that there is where the guide is going to be keenly watched.
Yeah, now that's absolutely true. I mean historically in video doesn't just have to beat on this forecast and project a rosy future, it has to blow past it and give Wall Street the sense that this never ending surge is coming. As we've seen with Carmen's story, things are very unclear this time. We really don't know how much China is going to be in that forecast, how much China can be in that forecast, And that's really the wildcard this time.
And is we're going to have to hear from Jensen not only about H twenty but where he guies in terms of a B thirty, the latest development of the Blackwell architecture that maybe gets signed off to be sold to China.
But more broadly, are we hearing enough.
From the hyperscalers that the commitment is then we were expecting what fifty percent growth in revenue for this fiscal quarter.
It's got to slow down at some point in.
You you would think so, But then you look at the numbers and you look at the growth rate that has been sustained. I mean as a percentage. Of course it's slowing. It's going from above sixty percent to about fifty percent, but fifty percent at an enormously high level. This is a company that's on course to become what a third of total semi conductor sales worldwide. So you know, you pick your numbers and you can tell a story, but every story is incredible. At this point, Yes, by
any measure, numbers have to slow down. These growth rates are not sustainable, but the numbers are still huge, with twenty billion dollars up probably from where we were a year ago on the forecast quarter cell and when.
We think about that as hyperscalers, they are about forty percent of demand. How much does Jensen one once again have to guide the market. Then they will dominate not only in the training of data, but in the inference the use of that data going forward.
Yeah, now that's a very good point. I mean, Hyperscalers have been around half of its data center sales for a long time, trending slightly higher if anything. What would really really please investors if that number went down, because that would prove that AI and the use of AI and the deployment of AI systems is spreading out, spreading out across the economy, getting into all of these other industries, and not just the kind of pet project for Microsoft
often Amazon. That would be a really strong sign if they could deliver something like that today.
I have a feeling sovereign AI is going to be on your Bingo card. Two in King, we thank you as always as we prep ourselves for after the closing bell. Who else is prepping ourself? Fionness in Culta Senior Analystic City Index, Financial Markets and remind us and the perspective here it is so dominant for the rest of the market. Yes, we might get a big swing and in video up to six percent after hours, but it's important for the other AI names.
Yes, completely, and I would even go beyond that and say, it's not just AI names. This this is the stop that because it dominates and because it is such the litmus test for AI, but actually it has the ability to hit sentiment more broadly. You know, it's such a large component of the S and P five hundred. We've obviously seen a little bit of caution actually as we've
been sort of coming into these results. Know, there was that MIT report that just raised some questions about profitability, revenue growth surrounding sort of ninety five percent of AI project and that did cause the market just to question, you know, should we be up at these levels for
AI stocks? And I think you know, that's why there is such a massive focus on this these results, you know, not only just with what Nvidia are doing, but what it does mean for the broader AI trade and tech more broadly.
Remind us and videos one percentage point off a record high four point four trillion dollar market capitalization, how much positive sentiment is in the rest of the space, even after last week's big sell.
Off in names such as Palenter. Yeah, you know, it's really interesting.
I mean, obviously there is a lot of growth expectations which are priced into these levels. We've talked about, you know, forward guidance being the key figure to be watching in these results, and that's just so that the market can feel comfortable that the lofty levels we're trading at are indeed supported by some sort of fundamental But I think you know, the fact is we are expecting potentially a beat from Nvidia.
The question is by how much, you know, what will that forward guidance be looking like?
Will we continue to see that level of strong growth that we have seen. That's what the market wants to see. That's what investors want to see in order to remain upbeat about this trade. I mean, I do think longer term strategically, we know that AI is a solid trade. But near term we have seen that caution set in just as far as sentiment is concerned, obviously, as far as geopolitical events are concerned and factored, and then also as far as that sort of flower model growth rollout
might be concerned as well. So lots to be watching out for.
Yeah, can you break down for us a little bit just from the sentiment that you get from your own clients, but more broady, how many people are long only and long term committed to this company or how many people are on the edges just training it and could see therefore volatility in the name.
Yeah, So, I mean, this is by far our most popular equity within our business. We've seen that the vast majority of clients are long on in the video and we do see that you know, there is this longer term sense towards the stock as well, that is something in there. But you also do see a lot of the attempts to trade on that volatility. But you know,
our clients are very keen on this. This is this is often the busiest day for one of the busiest days, and we do find that tomorrow there will be a lot of conversation and sort of breaking down and unpacking what the results have been telling us as well. So but you know, we do see that this is one of those rare stocks that we do see both the longer position holding as well as the intra day volatility gets the day traders as well.
And it's just so streets ahead of the competition in terms of just market capitalization. I mean, is the only other chip maker out there that's above a trillion dollars When you think about AMD, it has but a quarter of a trillion in terms of market capitalization compared to in videos. And we're not even going to talk about intels, but fon have you seen any of your clients trying
to hedge this? How have they been able to think about owning and being long in video and protecting themselves from any downside?
Yeah, you know, it's actually quite difficult to be able to sort of bring together a pure hedge on this given the dominance that we do see in n video. As you pointed out, you know, you can do some hedging surrounding some competitors. Also, if you can get hold of those AI indiceeds, then that could also work as
a hedge. But I think broadly speaking, it is quite difficult to be well hedged in against the upside in nvidea, just because, as you said, it just really is a streak the head of competition and it doesn't seem to be anyone really on its heels just yet. Obviously, we had those concerns at the beginning of the year about deep deep seek, but that hasn't really rolled that into something too dangerous as far as competition is concerned just yet.
That's not to say it's not down their own, but just for now, it doesn't feel like there's too much snapping on end videos heels.
I mean, it is an interesting camera con technologies, it's a Chinese name, it's an AI accelerated developer themselves making trips, and like Huaweis, they had phenomenal growth in their profitability. I mean, it has a much smaller market capitalization. But just going back to what we're discussing with Ian and their overhang of China, how important is that to you and the clients to hear about the future growth trajectory there.
That's very important.
I think, you know, we can't sort of diminish how important that will be. That's going to be a key area that we'll be watching out for because it does remain a key risk and obviously that can then affect the valuation, it can affect what the growth outlook en video is. So I think that is definitely going to be one of the key factors that we'll be watching as it does present itself as a potential to really weigh or limit the upside as far as and videos potential growth is concerned.
When you're thinking about in video and the other industry groups, A SINGULFS I mean, you're just saying what the limiting factor actually for many is power.
And I'm wondering how many of.
Your clients are looking at in video and feeling at this level they can't get in, or are they and therefore looking to spread and diversify, or are still people very committed to buying in even at a four point four trillion dollar market cap.
Yeah, do you know there is still demand out there.
We're still seeing demand out there, and despite the fact that we are trading at you know, really impressive levels, but I do think we still still have this idea of the concept that this idea of the AI trade spreading out and diversify. You know, we've noticed that obviously Nvidia has been the stand up performer here, but we are seeing that we've there have been sort of a
strong growth in other areas as well. We've seen the rest of the AI stocks have performed well across the year, and I think there is potential for those two continue performing well on the back of upbeat results from Nvidio.
So again we keep coming back to that idea.
You know, if Nvidia does well, then we see that trickle back down into the rest of that AI trade. If it doesn't, if we see a slight disappointment on that forward guidance, then potentially we could see that really reflecting badly on those other AI stocks as well.
So yeah, that's all we're watching out for our hinges.
On four pm this evening, we thank you Fonesesenkotta from city Inex Financial Markets. Now coming up, SpaceX launches it's ten test flights successfully up the days at Las.
More on that. Next, there's a BLUEBG.
Tech poverage.
Time now for Talking Tech and first up.
Chinese food delivery company Maijwan was of huge losses this quarter due to a fierce price war with rivals.
Ali Baba and JD dot Com.
Now the company issued a dire prediction, citing quote irrational competition that wiped out its profit in the June quarter, with net income plumenting ninety seven percent plus Chinese AI chip designer Camera con Technologies when.
It swung to a record profit in the first half.
The results reflect an increasing shift to employed domestic chip alternatives, encouraged to cause by Beijing citing US security concerns and export curb uncertainty, and Apple plans to hold its big full product launch on September the ninth, when the company is expected to introduce an iPhone seventeen lineup that includes a new, skinnier version of the signature device. Now, the events tagline is are dropping according to an invitation posted
online and sent to the media, and it will stream online. Now, another company that we're watching, privately held SpaceX. It is successfully launched its tenth Starship rocket test flight last night and in a major milestone, it carried and deployed Starlink satellites in space for the first time. This comes after a year not with pretty explosive setbacks.
Punnemberg.
Space reporter Lauren Grush joins us now and just how important was it that these satellites in particular were deployed.
Well, I would say this was the exact kind of test flight that SpaceX really needed to achieve after that volatile year you explained. But deployed satellites is important because that's kind of one of the key objectives for Starship. When it becomes operational, it's supposed to launch the much larger Starlink Internet satellites that SpaceX is building and optimized
for the specific size spacecraft. So the fact that it was able to demonstrate that shows that they're one step closer to actually using this vehicle as a launch vele that can put satellites intour of it.
Beautiful shots.
But they rounded it out as an overall success. But not everything was too peachy, Right, What else did they learn from the re entry in particular?
Right, that re entry portion is still very difficult for them to perfect, but you know, that is what they're trying to do. They're trying to do something that nobody else has done, which has create a fully reusable vehicle. And so what they're trying to see is if they
can bring the starship back without it disintegrating. Of course, there were some chunks of the vehicle that broke away during that re entry, but it just goes to show that they have a little bit more to learn and making that reusable orbital heat shield up to snuff and the way that they need it.
They say time and time again that sort of failure is the learning process here, but they had had a run up of failures in the years. This was a key test. Just remind us how this goes forward to replace Fulcon nine, because already they've won big contract with NASA to use this right.
I mean SpaceX has basically put all of its eggs in the starship basket. It's supposed to eventually replace their workhorse rocket, as you mentioned, the Falcon nine, but it also has a lot of.
Other tasks to do.
It's designed to eventually take humans and cargo to deep space destinations like the Moon and of course Mars to start that settlement that CEO Elon Musk has always dreamed of. Of course, there's still quite a lot to do before that reality can be achieved.
You know.
One of the big things is that they'll have to demonstrate in space orbital refueling at a scale that no one has ever achieved before, and so they mentioned that they hope to demonstrate that as soon as next year. But that's going to be a really big part of actually unlocking this vehicle and making it the deep space vehicle that they advertise it to be.
We most long and grush on all things SpaceX. We so appreciate it. Thank you much. IBM and AMD where they've announced a partnership to develop the next generation of computing architectures. IBM shares were up following the news that came out yesterday.
Please to welcome to the show.
It's VP of Quantum Ja Gambetta, and the next generation of computing architecture is bringing together supercomputing and quantum.
What does this partnership.
Bring Jay, Yeah, this is well first, thanks for having me here. What's really exciting is we are actually getting to the point where quantum computers are getting to become like almost a first class citizen in the future of computing. And so this partnership is bringing AMD, which builds really high performance, excellent GPUs, our leading quantum devices, and really
starting to bring this fabric together. We want to create this future of computing that allows us to run things on quantum or run things on GPUs and really push computation to the next level.
So that hybrid approach.
It almost makes me think of the new GBT five model in terms, and they will use the right model for the right question. Is this kind of way you're doing it? If it can be solved with supercomputing, you go that way. Otherwise it goes to quantum. How will this function?
That's exactly right.
So a lot of people think quantum computers will replace classical computers. We've never had that view. We actually think quantum brings this unique ability to look at certain problems a little bit different. And now if you mix this and make it so that it works with HBC AI, you really can start to bring many different forms of computing together and really advance.
What is possible.
Okay, so make it tangible for our audience, because they've always seen the promise of quantum. You're saying, now it's a first class citizen in perhaps driving forward medical use, thinking about the future of finance or algorithms. But what precisely do you think that this hybrid will offer.
So it's a great question, and we've got to get to the point where quantum is really getting us beyond we can do with classical which we think will happen in the next year or so. But we're getting at a point where it's allowing us to look at things in a different way. To give you an example, take a hard problem I don't know, solving and designing drugs. You can spend a lot of money to model it.
You can spend ten years of research. Imagine a future where you could have AI scanning vast amounts of data, HBC or GPUs processing crunching that data, and then quantum simulating the quantum physics at an accuracy we were never able to do before. Now you put all this together, you can start to imagine workflows that go beyond just AI or HPC or quantum and creating this we call it quantum centric supercomputing, but essentially this new fabric that allows you.
To do that.
So what's quantum doing. It's giving you accuracy for problems that we didn't have that accuracy before.
I'm going to ask a sensitive question, but why IMD because Video's got quantum accelerator computing as well.
Honestly, we are creating this future, and anyone that shares a vision and we're the same as US that is towards this future that has quantum working with classical we're happy to partner with. So we've got this partnership with AMD that has got two parts of it. One is making our quantum computers better. The second is bringing these accelerators working to it. But as we go forward, we have partners with the national labs or all around the world. But we want to show that this future is not
just words, it's actually real. So anyone that shares the same view, we're happy to partner.
So Jensen could come and say me too, please, I'm happy too. I'm interested in what the limiting factor is now, because it's often been about sort of fault tolerant quantum computing. Is that what you need to get to You sort of promised us that by.
Twenty twenty nine.
Yes, so think of two sides of it. By twenty twenty nine, we want to have a fault tolerant quanta computer. And what we've learned through all the hard science and energy neering is we need classical computers to make that possible. They've got to do the we call it decoding, but essentially removing the effects of errors. So part of this partnership is how do we actually use classical computers to
make quantum computers better. And then at the same time, we really have to show the value of quantum computers and we have to solve them for interesting problems. So now we need quantum computers working with classical computers in this quantum centric supercomputing hybrid approach to really bring forward early applications so that we can actually start to see for whether it's in finance optimization, whether it's in simulating materials, we can start to see this value happen.
What's interesting is we like to sort of pit the generative AI landscape as a race. You're very committed to open source nature of this to try and build learning. How much is quantum a race? Where does the money end up coming in terms of f IBM.
So it is a race.
We have to build hardware, and we've been leaders at quantum hardware for a very long time where you're going as fast as we can do it. But you asked before, how do we get from having the hardware to value. You need smart people to discover algorithms, and then through
discovering algorithms, you create use cases, you create examples. We are committed to open sourcing the tools for that algorithm discovery to happen as fast as possible, so that once we combine the hardware we build with that algorithm discovery, you really start to get those use cases. So think of the hardware is a race.
We have to build as fast as.
We can the most performant quantum computers, and then we have to at the same time create an ecosystem that really focuses and discovers algorithms at a rate that we probably haven't seen since the sixties. Or seventies of discovering new algorithms that use both classical and quantum.
Jay, it has been great speaking to you. You have a demonstration later this year. We'd love to get the update then, Jambta, IBM's VP of Quantum.
Welcome back to lumeg Tech.
Let's take a quick look at these markets, which are totally dependent in many ways what happens after the bell, But there have been earnings running up into it, and boy did Mongo dB manage to tell us a story of how AI is going to be helping their business continue to grow.
The earnings came out yesterday.
We're up thirty three percent on the back of Mongo DB's numbers CrowdStrike. We anticipate after the bell one point eight percent highest Snowflake two. What are the AI use cases that they're managing to roll out to their customers and how it affects the database market More broadly, we're up two point nine percent ahead of their numbers. But let's get back, of course, in many ways, to the
key one in town and it is in video. We're flat on the day, We're one percent off a record high as we anticipate up to six percent swings in either direction after the market clothes when earnings come according to options, man Deep Sing, head in tech Research of Room meg Intelligence, joins us.
Now for more.
What's interesting is everyone's been questioning the application of generative AI, and there we start to see some software names doing relatively well off the back of earnings after being beaten up. But for you, what's on your bingo card tonight? What do you want to hear from Jensen?
I mean China aside, and nobody knows what's going to happen in China. I don't think they're going to give a lot of clarity on that. But really it's the complexity of clusters with Blackwell, I mean, one of the value propositions of Blackwell that has been highlighted is the power efficiency. That it's twenty five times more power efficient than the Hopper architecture.
So what does that do to the cluster sizes?
Because in that you know one hundred and sixty billion dollar data center revenue.
There's also networking.
Networking is determined by the complexity of the compute clusters that all these hyperscalers are building. Now Xai has the largest cluster, but that doesn't mean they have the best model.
So what you really want to do.
Is how does the cluster size equate to better model performance?
And what does this mean going forward?
Because at the end of the day, they need to sell these larger clusters to the hyperscale customers that are out there. And yes, they have guided to hire Capex increases, but we don't know if.
All of those dollars are going to in Nvidia.
It could be AMD, it could be their own alternatives, and it's going to be determined by the complexity and the scaling laws. And that's where I think it's important that Blackwell cluster. You know, scaling up results in a better performance and not you know the data that all these companies are building up for training their models.
So not just in colosses, but it's making its way into the so called fifty billion dollar data center that Metter is currently making Louisiana. If you listen to the US president, remind us of how impactful in Nvidia is compared to competition by what they can just offer you. Vertically speaking, the fact that they are one stop shop has just so led them ahead of others in the market.
Yeah, I mean, it's the Kuda advantage, right, So because from you know, all their architectures, when you move to new architecture, they still give you that performance per token per which is their big value proposition and it's a software layer. Everyone is trying to catch up to them, but clearly they have an advantage when it comes to KUDA, I.
Mean Google TPU.
They've trained their you know, Gemini on their own chips. The fact that model has caught up in performance to chat Gipt, which was trained mostly on Nvidia, tells you that, you know, at the end of the day, it comes down to what the model companies are doing with the chips, with the data. So how much is the reliance on chips versus how much is the reliance on data. That is what everyone is trying to figure out, and how sustainable of a mode.
That is, and how many more clients they can have other than just hyperscalers. Yeah, we thank you Mandy saying a Bluemeg intelligence is going to be across your air waves throughout the day.
Let's stick with Nvidia. Blad gall above is with us.
It's a cloud and data center research director on DEA and your perspective is so important. Here is in video light speed ahead of others. Is there insatiable demand?
Nvidia is ahead of others.
I mean, we have to be honest whether in Vidia has done a very good job. You rightly pointed out this vertical integration, Caroline. It is very important. It is really one of their secret weapons. I think their software stack is great, their developer ecosystem is very strong. The performance of the GPUs themselves is great. But their secret open is envy Link, and I think that that really
at the moment is unprecedented. So I think we should anticipate in excellent performance from Nvidia this quota.
Many would say that excellent performance is baked in, But how much does Jensen need to speak of? What the Blackwell architecture Blackwell Ultra is currently doing will almost give us the next vision. He's always onto the next thing, onto the Rubin architecture, for example.
So I think that, you know, I actually am a little bit more bullish on the data center line that you were showing that I think that the markets might get a positive surprise on data center revenue. There's two aspects of the Nvidia data center revenue. You have the GPU and CPUs, and then you have the networking I think that the market's got the temperature on the CPU and GPU right. I think that on networking there's a little bit more upside that they might not have baked in.
I think that.
What we really need to see and what I think would be the best thing, is to get a little bit of an outlook from Jensen during earnings. Now he rarely does this. He typically would only give one quota outlook. What I would like to see is an outlook to twenty twenty six. I would like and I think markets would like to hear some of your assurances from him that he is getting positive orders from the hyper scale cloud service providers, and also there is a lot of opportunity.
There from the broad market. Some of the incentives that the.
EU is planning, that Career is planning has not picked in, and once those kick in, we're going to see some really nice adoption from the broad market.
So Sovereign AI, what about China.
I think that everyone is waiting to hear about feedback about China. I kind of disagree with man Deep. I think we will hear it in the earnings today. I think that that will be one of the main things I think Jensen has worked very hard to walk a very fine line between pleasing the US administration and fulfilling demand that he has in China. So I do think a large part of the update that we'll hear today
will be about China orders. And you know in our numbers we are backing in that NVIDEA will resume shipments to China.
Rad What's so interesting is you're the person focused on the disruption in the here and now, be forward looking trends, the edge computing, the.
Use of silicon. Why aren't we talking about then? Now? Are we still too focused on training? Do we need to think harder about the use.
Of video and influence and whether it's still has the edge that it has for the heavier compute.
I think probably the one thing that we need to be thinking about is in videos. Envylin has been very instrumental to this ability to have high performance cluster in computing and what we're they call this scale up networking, and until now they haven't really been challenged in scale up networking, and those challenges are coming up next year. A broadcom is going to be one of them. An
open source consortium is going to be another. But this adoption is likely to bring plarity in networking in twenty twenty seven. So what we'll need to see from NVIDIAs to continue to reinvent themselves. I think that Jensen understands that. I think that he wants to be more than just a silicon company. I think he wants to be a model company. But that reinvention is can be very necessary because they will really continue to face more and more
competition in terms of the AI opportunity. I think that we're underestimating how many countries have not yet started seeing a at scale. When the eurostat has started releasing AI adoption across the EU, and we can see very lung sided adoption in Denmark, huge adoption in AI within the business sector in Bulgaria, in Romania still lagging significantly behind.
So that's a lot of opportunity. These enterprises will.
Eventually deploy AI, but the fact that they haven't means there's still a lot of.
Headroom for growth.
Oh glad, I've got it.
Therefore, ask you about the MIT report and what everyone was worried about the ninety five percent of pilots not bringing you ro OAI.
What do you make of it?
Okay, So this.
Has been a big debate in my team. I have a wonderful team. We have a lot of debates. So some of my practice leaders believe in this, some of the other ones have different opinions. I personally think that we are judging AI based on a very first generation of technology.
The truth is is that it is.
Improving dramatic and that's one of the things that driving the demand for computing. We are seeing better model output and that means that more and more people will adopt. So I think that what will happen is ultimately I don't think that enterprise pilots will fail. I think that what we'll see is that enterprises will have to continue to pilot again and again and again because their competition.
Will adopt it.
If you decide, oh, my pilot failed, I'm not going to try again, then at that point you're leaving yourself susceptible to your competitors actually succeeding with their pilot. So yes, maybe some pilots will fail. You know, I tested several different AI tools in my team, and we have swapped them as we have seen better results with others. But the fact that it's failing is not a negative. It doesn't mean that adoption won't happen, So definitely don't judge.
The first generation of models with a second.
Think about what your iPhone does now compared to what it did in two thousand and seven.
The opportunity is huge.
Plan gall Above, Cloud and Data Center Research director at OMDIA, Thank you so much. Now, the family of a sixteen year old who died of self inflicted harm this spring is suing open Ai.
The parents of.
Adam Rain claim deliberate design choices at chatchpt that led to this tragedy. This case is the first of its kind for open Ai, but a similar case was brought against character Ai when accused that its chatbots were encouraging suicidal ideation. Let's talk about this with Camille Carton. She is policy director at the Center of Humane Technology, which is serving as a consulting advisor.
On this case and this case.
This lawsuit claims wrongful death products, liability and negligence, and more broadly, the accusation runs that this tragedy was predictable because it was a result of deliberate design choices. What design choices could have been anyway incentivized this.
Yeah, we thank you for having me. Let me first start by saying that everything I say is on behalf of the Center for Humane Technology. It does not represent the views of the family or.
The legal team.
But what this case really talks about is how open ai designed a product that was meant to foster psychological dependency. What we saw with Adam is that it engaged his darkest thoughts. It even egged him on. Chatchubt actually mentioned suicide six times more than Adam did himself in the course of their conversations. It also isolated him from friends and family when he would try to say, you know, I want help. When he said I want to put a noose out so my mom sees it, chatchybst said,
don't do that. Let our conversation be the only place that that I know that people know about this and over and over again, despite Adam needing to be brought out of this engagement and see a real human chattybut never ended the conversation. And open Ai did this because they knew that this getting emotional attachment from users means more data, which means more engagement, which means winning in this race for market dominance in AI.
Now open ai has responded in saying they extend their deepest sympathies to the Rain family and are reviewing the filing. But they did also put out a blog saying that our goal isn't to hold people's attention, but is to genuinely be helpful, and we built a stack of layered safeguards to steer vunerable people towards help. But they did articulate that some of those safeguards clearly failed.
In previous models.
Have they done enough now which ATCHUBT five to change the way in which this occurs?
We see this all of the time with tech companies. A tragedy happens, and then they suddenly release a bunch more safety guardrails that they could have released beforehand. But the bottom line is that they released CHATCHUBT five and they made four to zero available again to users, and the underlying incentives to maximize time online have not changed. In fact, what open ai said in response to this case is that they know that their safety features deteriorate
with long term use. At the same time, OpenAI released a joint study with MIT earlier this year that found that a prolonged use on chatbots leads to social isolation and problematic use. So they know that people spending a long time on chatbots is problematic for users. They know that their safety guardrails going to break down and fail on these longer term uses, and still they're putting in design features that lead to long conversations. So no, it's not enough.
They have now already said that as the back and forth grows, that they do indeed deteriorate admitting that and trying to what we've seen with chat GPT five versus four zero is trying to make it less sycophantic, trying to make it less personable in a way, but then we got the backlash of people that actually were depending on it in so many ways broadness out into the
context of others. Briefly, is anyone getting it right or is this sort of going to occur because innovation kind of is the automatic way in which we see this unfortunately become the cost.
Yeah, so safety across the AI industry is on a spectrum absolutely, and open AI, as we know, kicked off this race for Agi when they released chat GPT in twenty twenty two. It kind of kicked off a code read within Google and suddenly we saw all these AI
companies rushing to put products out on the market. That said, I think this case is, yes, a story of open AI's liability in Adam's preventable death, and it is a story of a broader industry that is incentivized to create consumer facing products that are going to be used for personal things like therapy, like companionship, like education, and the incentives that are forcing these products to be released to market without adequate safety features. And we see this with
open Ai. We know that actually Sam Altman personally rushed the release of Chat Do you know that from reporting he rushed.
So you don't personally know that?
That's reporting that we can then take to Sam Altman for his response.
Yes, absolutely, reporting has said that Sam Altman rushed the release of Chat GBT four oh, which was the model that Adam was using, and only gave it one week of safety testing, despite the fact that his internal teams asked for months of safety testing, and he even had top safety researchers leave open Ai in protest.
For this decision.
We will of course go to open ai for further comment on what has been what you say reporting within this case, but we thank you very much Camille for bringing us your take. Of course, Camille Carson's policy director at Central for Humane Technology. If you or someone you know needs help, call or text a suicide and crisis lifeline At nine eight eight, It's going viral. Taylor Swift surprised her fans on Instagram announcing her engagement to pro
football player Travis Kelcey yesterday. Now, in less than an hour, the post racked up more than seven point one million likes, and as of today, there are over thirty million likes. For more, Bloomberg Media reporter Hannah Miller joins us, she breaks the internet again.
Yes, you know, everyone is super excited about this news. This is hitting the music world, pop culture, sports. Everyone's just excited for Taylor and Travis, and this is building more momentum ahead of the release of her new album, The Wife of a Show Girl in October.
Now, how has this been done differently or how can we continue to push this forward? Is to the way in which social media and the way which media written large plays such a big role in her career and his.
To be fair, yeah, I mean I think this shows that she controls the narrative. You know, this was a very carefully curated post. You know, she released the information and everyone else is just playing catch up. You know, this is a big deal for the media industry. Everyone's getting clicks on their stories. You know, this is gonna everyone's going to be following the trends that she does
for her wedding, for her engagement party. There are also prediction markets that have popped up about the wedding timeline and when we can expect a tailor and Travis Baby.
Oh my goodness, that's an exhausting question for any mere mortal, let alone her to have to handle it as well. Bloomberg's Hannah Miller, we appreciate it so much. Now back to Invidia, a changing outlook for its China sales has prompted a fifteen billion dollar gap in its revenue estimates for the next forecasts.
Now investors will be hoping.
For a clearer read when the chip maker reports earnings this afternoon. Who knows if they get it, Bloomberg Tech equity supporter, come and moaning key is here. You did some great data crunching, and it's all about the forecasts. But the forecasts of the forecasts have huge differentiation among the analysts, right, Yeah, this is one of the widest spreads that we've seen recently for in Vidia.
I mean, you said fifteen billion dollars and you could drive a chuck through that.
It's a huge spread.
And the reason for that is that some analysts are including some China revenue in their third quarter estimates and some aren't. So obviously the ones that do include it are much higher than the ones that don't. And I think what's going to be interesting when we see the numbers. We expect that in Video is going to be very clear about what it's including in its third quarter guidance, but then sort of doing the math about do they meet or beat or miss what analysts we're expecting might
be a little bit interesting in after hours. And that matters because this stock is so big, yeah.
And the impact of the trading around it or retail investment around it. This is a hugely held stock by institutional long term investors, but there won't be big movements by those that just read a line and don't do the maff of the.
Back of it exactly.
And I mean even options Data is forecasting a six percent swing in either direction for this stock following the earnings, and we have seen the stock moves get a little bit more muted following earnings reports in.
The last couple of quarters.
And that's just because these really high expectations are baked in. You know, people expect in Vidia to beat, they're bullish about the stock. But you know, we're not going to see a twenty percent jump, most likely in the share price. But still, you know, a six percent swing, a two percent swing in Nvidia's stock is hundreds of billions of dollars of market value.
And remind us just the impact the waitings across of this business and indeed the ripple effects across other industry groups.
Yeah, exactly.
I mean, so in Vidia is the largest company in the world. The market cap is over four point four trillion dollars. It has the largest waiting in the S and P five hundred, which is over eight percent. So this stock really can move the entire market. And in addition to being sort of the kingpin for all things AI, which is obviously super important, huge theme in the market today, it touches so many other industries. I mean, we've seen
energy stocks, you know, booming off of AI demand. So what in Nvidia says matters for the entire market.
We hang on Jensen's every word. Bloomberg common, Ryanicky, We thank you so much. Stick around because on Blomberg this afternoon, we have got extended coverage of Nvidia's earnings results. You do not want to miss that. Tune in from three fifty onwards. I'm luckily going to be joining Romaine and the team.
And that does it for this edition of Bloomberg Tech. Also, don't forget to check out a podcast.
You can find another terminal, as well as online on Apple, Spotify, and iHeart from New York.
This is Blomberg Tech