Closing Bell - Closing Bell: 10/15/26

Episode Date: October 15, 2026

From the open to the close, “Closing Bell” and “Closing Bell: Overtime” have you covered. From what’s driving market moves to how investors are reacting, Scott Wapner, Melissa Lee and Michael Santoli guide listeners through each trading session and bring to you some of the biggest names in business.   Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

Welcome to closing bell. I'm Scott Wobner live from One Market in San Francisco. Once again, our special report, AI in America, continues this hour. How vulnerable are we to a cyber attack from an AI agent gone rogue? Are we doing enough to prevent one? We'll ask Palo Alto CEO Nikesa Aurora coming up. Plus AI and the future of work. We're joined today by a Nobel winning economist making a dire prediction. Is he an alarmist or a realist?

We'll debate that. Let's first check the scorecard with 60 to go in regulation today. the AI trade once again at the center of the action. Stocks, though, as you see, well off their lows. In fact, the Dow S&P and the Russell are now positive. We need to check shares of CoreWeave shares. They've turned positive midday following yesterday's big decline. Jensen Wong of Invidian making some comments there, and that has helped turn that name. Overall, a mixed day after more hot inflation data. Yields, though, today are falling, and we're watching all of it. I mentioned the Dow up big, the sector picture mixed today. Healthcare leading the way it is a more

defensive tape. There's no doubt about that. Our talk of the tape, the AI trade, and whether it's showing some major warnings, or is this fear and overreaction? Our Kate Rooney joins us now with a look at where we are at this transformational moment. Kate? Scott, great to see you again. So that's right, it's a great setup. What was once this very niche Silicon Valley technology has moved to the national stage dominating markets and reshaping politics ahead of the midterms. front and center this hour questions about the AI trade the more mainstream conversations i think outside of cnbc ai is a boogeyman the debate over data centers is only growing louder and the backlash seems to be growing too artificial intelligence headlines delivering whiplash for investors this year

warnings of extreme risk and the promise of extreme reward here we are with record highs on the s&p 500 and the nasdaq how did we do that the risk comes down to cybersecurity, with dozens of examples of autonomous AI agents from OpenAI, Anthropic, and Google going rogue with unintended consequences. OpenAI's breach of hugging face, the watershed moment. On the current path, I think it is more likely than not that humanity loses control to these AIs, and it could end in human extinction. The big story all week, all weekend, on Saturday morning, Anthropic CEO, Dario Ammo Day, publishing an essay saying, quote, we must slow the pace at which we improve the capabilities of AI models.

CEOs are leading the largest AI companies have been some of the loudest voices. Being companies in competitive race is not a reason to make grass decisions. We have unilaterally slowed down in the past. We will do so in the future. I believe that this is the most important global security issue facing the world today. The irony of those tech CEOs asking for AI slowdown became a parody on SNL. If we can pressure lawmakers to create guardrails, we will be able to stop me. And it's turned into a.

political blame game. Hey, hey, home, home. AI companies have done a horrendous job. Hyperstale has got to go. Werenous job of explaining themselves to the American people. What do we want?

The risk also becoming a political flashpoint ahead of the midterm elections as Americans push back against AI data centers which power this tech. A hot dog that you buy on the street corner in New York City has more regulations than these companies have on them.

Yet Washington remains divided. President Trump saying he wants no new laws in that the U.S. must not lose this tech race to China. Winning the AI race is winning. And appointing former SEC chair, Jay Clayton, as the nation's AIs are. His job to weigh the risk and opportunity in AI while figuring out what the government's role should be. For now, it's up to the AI companies to police themselves.

If you want to survive, you've got to behave. To do that, Microsoft CEO Saty Nadela suggests an emergency break. Deep Mines Demis Sissabas, a kill switch. Still other CEOs like NVIDIA's Jensen Wong say the risk is overblown. We're going to create AI that's incredibly useful. It's going to be incredibly helpful and it'll be incredibly safe. Amid all of this, the companies at the frontier of AI raising more than $200 billion in private capital this year alone, becoming the most valuable private companies in history. And Scott Anthropic, one of the AI giants at the center of all of these debates going public as soon as November from what we're hearing.

We also heard yesterday Anthropic grew revenue, 45% from the prior quarter. Sounds impressive, but we've seen the reaction. Some investors are a bit more focused on a perceived slowdown, Anthropic had been growing 140% earlier this year, so the bar it was quite high here. Company, we're also hearing from sources. It's profitable, at least on an adjusted basis. And you heard Brad Gersner, who was in that room yesterday, suggest he thinks it's a matter of weeks, not months. And he, as an anthropic investor,

certainly hope so. You would know. We're waiting with bated breath for the S-1. And honestly, that is where we really will get the meat behind this. I think profitability is going to be the big thing folks focus on. And what does that look like? What are these adjustments that we're talking about? We get the fact that, yeah, sure, there's an adjusted version of profitability, but what are they stripping out? And when this company looks more like a public company, we'll be able to compare it to others as well. But as others have said, I mean, this is the first of a new asset class. And so it'll be a big test to a master appetite. That was a great look for us. Thank you. Kate Rooney, right here at one market. Corweave shares in focus yet again today.

Sima Modi joining us now with the very latest there. The shares now turning positive. And thanks to NVIDIA's CEO Jensen Wong-Scott, defending its partner, Corweef saying Correve does not need a backstop. Those comments made at Bloomberg to Bloomberg at an event in Sweden. Correve is among the companies in VDIA has an equity stake in announcing a $2 billion investment in January, while Correve pledge to use Nvidia's stack to build about 5 gigawatts of data centers in the next four years. So yes, Nvidia is key to CoreWeave's success and the entire NeoCloud sector, where they're buying GPUs and leasing them out to hyperscalers and the frontier labs. Scott, what differentiates these names is the duration of the contract,

summer long versus short. In fact, analysts at BFA saying the length of contracts Corweave has, being five years, puts them in a stronger position, but a lot of focus on its unsecured bonds that continue to sell off, And the challenges this could pose and Corey's ability to secure additional debt, though CEO Mike Intritor is saying in a memo to employees, equity issuances could also be an option. At some point, Corey will need to if we finance existing debt. The market now awaiting two specific data points, Scott.

There's meta earnings next week. And then Oracle's Investor Day on the 28th. That will give a good read on AI financing and how much appetite there is in the bond market. All right, good look there as well. Seema, thank you. Sima Modi. And then there is the data center backlash. which only seems to be escalating.

Are Pippa Stevens in Ohio today where that debate is in fact raging, Pippa? That's right, Scott. So I'm here outside of Toledo in Oregon, Ohio, where it proposed 500 megawatt data center and this field behind me has led to push back from the local community. And next month, residents are going to vote

on a first-of-a-kind measure looking to limit the size of data centers in this area to 25 megawatts. Now, it's an issue that has divided an otherwise tight-knit community. Supporters say they'll provide needed revenue for the school, fire, and police departments. Opponents say that it's just simply too big. They question the long-term benefits and say that the proposal was pushed through. Now, this of course, just one example of a conversation that's

playing out now across the country and that skepticism around AI grows, data centers are seen as the physical manifestation of those fears. In an August poll, 75% of respondents said they opposed data centers that was up from just 43% last August. And ahead of the midterms, the conversation has only accelerated becoming a key issue as voters prepare to head to the polls, including here in Ohio's 9th district. This is a toss-up rematch race between incumbent Democrat captor and Republican Marin. Now, it's a really thorny issue for politicians because they don't necessarily want to take a strong stance, given that data centers don't always come down to party lines and they don't want to lose voters.

Each has criticized the other for their past data center policies and now basically say the same thing, which is that it should be up to local communities. And Scott, it will be very interesting to see after the midterms when they're no longer on the campaign trail the actual proposals that we see come forward in terms of the extent to which AI should be regulated. Pippa, thank you. Pippa Stevens in Ohio for us.

Now let's welcome in Jeff Richards. He's a notable capital managing partner back with us once again. It's good to have you. Great to be here. Welcome to San Francisco. What do you think is happening in? the market? Well, obviously, a lot of folks have been waiting for numbers or rumored numbers around Anthropic and Open AI, and that's driving a lot of what's happening in the last 48 hours. You know, I posted on X yesterday, I think one of the real challenges we've had in AI and technology

overall for the last six, 12 months is that so much of what's happening is happening in private markets, right? So Open AI, Anthropic, but also as Kate mentioned, this next class of great companies that are coming along that are all benefiting from this tailwind with AI, you've got companies like Base 10, Fireworks, Vercell, Modal, which just raised capital, FAL, data bricks. We'd love to see some of those companies go public. And I think that'll ease a lot of the fears, because then their information will be public. Speculating or relying on charts and graphs from Twitter, which can be somewhat reliable or unreliable, has really been a challenge for public market investors. And I think once we get more of the information out there, analyst coverage,

detailed information on core metrics into the market over the course of 27, I think a lot of those fears will ease. We did get numbers, obviously, from Anthropic yesterday at the investor day. Were you in the room? I was not. Okay. But we're an early investor in the company, as you know. That's why I ask you if you're in the room. Revenue's up 45%, almost 17 billion in the quarter. The chatter on the street today seems to be not good enough. How would you read what you heard? Well, I think, I always think of these things on a very long-term arc. You know, we're obviously very long-term investors and everything we invest in. You know, if you had told people that the fastest growing software company in history would come to market in 27 or 26, growing it 10x year over year

with what is reported to be over 900% NRA, which is a measure of how much are your customers spending last year than spending with you this year, which is 10x, would be probably the highest metric in the history of an IPO. I think people would be pretty excited. So obviously a company like Anthropic or Databricks or OpenAI can't manage public perception based on rumors and leaks. But I think as they come out and are publishing their numbers and you have analyst coverage and whatnot, this is such a great long-term bet.

I heard Brad earlier on with you talking about just the long-term horizon that we're on here. And if you want to comp, go back to 2005 when Amazon launched AWS, 2005, 2006. Amazon was a $30 billion company. If we all knew what AWS would have become today, we would have put our net worth into Amazon. So at the time in 2010, UBS came out with an analyst report that forecasted the total addressable market for cloud services was $5 billion a year. Amazon was doing $500 million. Today, the hyperscalers combined do $350 billion a year.

So I think with AI, you're looking at something similar, where we're super early. We're in the first out of the first inning. And I think what people will really buy into is the long-term trajectory of this business. And of course, there'll be gyrations. And, of course, we have unbelievable volatility in the market today. Let's not forget.

We have inflation. We have a war in Iran. We have interest rates, you know, moving upwards. So there's a lot of volatility in the market. beyond what's happening with AI and tech. You think the fears around all of the debt being raised are overblown? Do you think the way that the market's been reacting to certain companies has been overdone? We don't really play in the debt market.

I mean, we're funding companies that, for the most part, don't have any debt. So it's not really, I'm not an expert in that area. But it speaks to sort of how much money is being raised on the expectation of what's going to be spent and what demand ultimately is going to look like. I think we're questioning where we are in it. And I think that demand side is where people don't have visibility. Again, back to the companies that are driving demand today. It's Databricks.

It's fireworks. It's base 10. It's for sale. It's these infrastructure layer companies that the large enterprises are relying on and buying from. Once you have some of those companies come out into the market and they're publicly disclosing their metrics and their numbers, people are going to see billion-dollar-plus run-rate revenue companies growing at 60 to 100 percent per year. By the way, we don't have any public software companies in the market today. They're growing in those kind of rates.

Once they see that, I think the demand side becomes more clear, and perhaps it eases some of the fears around, can people actually afford all these commitments they're making? You've been writing about the IPO market. I mean, we had the firmness sort of cancellation, right? To which you say, speaking of the IPO market, companies like acceleration, firmness and aura showing weakness or not going public at all, has zero correlation with the real IPO market for AI companies. Why so? Yeah, I think it's a sign of a healthy market when the market's discerning.

I don't think any of those companies were on the top of people's list of companies that are excited about going public. and Firmus, which we're not an investor, never met the company. I know nothing about it, but it was a $50 million, $26 revenue company trying to go public at a $30 billion valuation with somebody on the management team who had spent time in jail for insider trader. I'm not sure that's our bellwether for the IPO market. And the fact that the IPO market essentially pushed back and said, no thanks, I think it's a very strong signal, right? We're seeing discernment. You wind back the clock to the 19, 20, 21 era when we had SPACs.

Lack of discernment in the market, not a great signal. And out of that era, most of those companies did not do well. But out of that era, we got Cloudflare, we got Snowflake, we got CrowdStrike. I think you will see some enormous winners come out of this era of the IPO window. But hopefully the market is discerning enough to either push back altogether or have the companies that perhaps don't have the greatest fundamentals. Come out of an evaluation that's reasonable and rational. Do you think we've pushed back the clock of the timeline for some of the companies that you mentioned,

the data bricks of the world, and the ones maybe we thought would be public and now it's going to be pushed off because some of the volatility around. the story? Well, let's not forget the biggest IPO of the year and one of the biggest in history, if not the biggest, I think, was SpaceX. It's done really well. High-quality company, great growth, arguably the greatest entrepreneur in the history of our country, tiled up well. So I use that as a barometer more than I do these failed IPOs and say, how is SpaceX doing it? If that's doing well, that tells me that there's demand for great growth names that are exciting new parts of the economy and should bode well not only for Anthropic and Open AI,

but also those other names I gave you. I think it's interesting, too. You suggest we're in a new phase of consumer adoption, enterprise adoption is still very early. It's so interesting. I had a very similar conversation with Robert Smith, a Vista Equity Partners. This is one of the foremost experts on investing in software and suggested sort of the same idea. If you look at all the buzz and the hype around Muse, for example, that's sort of the story that you're telling as well. Yeah, and I think if you wind back the clock to when the iPhone came out in 2008, the App Store didn't launched until 2010. So essentially people had this really interesting device, but they didn't quite know what to do with it. It took two years for that to happen. You think about when chat GPT launched,

everybody said, oh my God, this is amazing. But then we didn't really have anything on the consumer side that changed that narrative. I think Muse is the first thing that's hit the market. Obviously, a lot of people in Silicon Valley are using other tools like instinct, and there's a few others. But you're going to see a whole wave of consumer-focused innovation kick off in the next 24 months. And let's not forget Waymo. I mean, you've probably walked around the street to San Francisco. every 10, 20 cars is a Waymo. Been in one from time to time. Yeah, they're incredible.

And so I think as more and more people around the country, to me, Waymo is one of the best examples of what is AI going to do for us? If you had told people 10 years ago, 40,000 people a year dying car accidents, we can fix that. Right? Here it is. It's AI driving a car with no human intervention and no human in the behind the scenes. That's an incredible amount of innovation, but it's still only available in a few cities today. So I think as more and more of the actual end benefit of AI becomes apparent to consumers, you know, we may even see some of the narrative around things like data centers,

which you were just talking about, change because people will see, I see what this does for me as an individual. And then the other side of this is, let's not forget, small businesses, 42% of US GDP. Small businesses today, 70%, more than 70% are using AI. Now, they're using an A in a very primitive way, probably for searches or maybe helping them optimize payroll or whatnot. But as that ramps up as well, let's not forget, most small businesses owners are consumers. So I just think we're super early, not only in the

enterprise, but also on the consumer side, and that will change sentiment. Obviously, companies going public, posting numbers that people are impressed with, showing positive gross margins, showing positive cash flow. And again, let's get the next snowflake, the next crowd strike, the next cloud flare out into the market, and I think you'll see a lot of the sentiment change. All right, first out of the first inning. We'll see. Thank you for being here. Thanks for having me. All right, Jeff Richards joining us. Once again, there was AI news today as well, Martin Apple, Mackenzie Seagallos, joining us with that news. Hi there. Hey, Scott. So an unlikely new AI alliance out here in Silicon Valley. Apple and Mehta are finding

some common ground after years of tension between the two companies with Mehta C.E.O. Mark Zuckerberg, even giving Apple CEO John Turner's a public shout out today for working with Meta on the project. Now, it's the kind of pairing that would have been hard to imagine just a few years ago and another sign of how AI is rewriting some longstanding alliances in big tech. Now, in this case, they're coming together around two areas that both companies are investing heavily at open weight models and then AI that runs locally on device. Meta's newest open weight model announced this afternoon is its most powerful yet and has been optimized to run across multiple high-end Mac studios for coding an AI agents that keeps more data off the cloud and can also lower the cost of running those workloads.

Meta now becoming Apple's third major US AI partner alongside OpenAI and Google. In China, Apple, of also works with Baidu and Alibaba. But because this is an open-weight model, we're not talking about the kind of traditional commercial partnership that Apple has to some of its other AI partners. And neither side at this point is saying whether they could eventually bring other meta-models into Syria. Scott?

Mack, thank you. That's McKenzie Segalis. We are just getting started here. Coming up next, is AI coming for your job? A Nobel Prize winning economist sounding the alarm on the threat to our workforce and later rogue AI agents launching cyber attacks, Can they be stopped? We'll speak with Palo Alto Network CEO, Nikesh Aurora, about that very topic.

Both of those interviews just ahead. As our special report, AI in America, continues. One of the most pressing and unsettling questions facing our society today is what the proliferation of AI will mean to the future of our jobs. Dumers suggest a major upheaval, the likes of which we've never seen. Others say the calls for surging unemployment are overblown. Duran Asimaglu is a Nobel Prize-winning. economist, and he joins us now. Sir, welcome to our program. It's good to see you today. It's great to be with you. Thank you. You say unemployment could triple in the next

decade. Some might suggest you're a doomer or an alarmist. What would you say to those calls? Well, I think we have to be alarmed. This technology is very, very powerful. And if history is any guidance, we can use powerful technologies for good or for bad. and the way we are developing AI right now is as an automation technology, which means that it will ultimately be capable enough to take a lot of jobs. The calls have been out there, obviously, for the past few years, and they really haven't come to fruition. So I want to get your reaction to that.

According to a Stanford study, the unemployment rate for the percentage of people with jobs most exposed to AI rose less than those who weren't exposed. What is your reaction to that? as almost every call that we've heard about doom and gloom about the workforce has fallen short thus far. It has, and there's a good reason for that, because I don't think the spread of AI is going to be as fast as some of its biggest proponents are predicting. If the spread of AI is slow, then the job impacts are going to be slow. The productivity effects are very slow as well.

I wouldn't rule out productivity gains from AI in the next 10 years on the basis of what's happened until now. In the same way, if AI starts diffusing rapidly in the job market, it is going to have an effect because right now it is designed as a technology to replace humans. That is the lodestar of the industry under the banner of AGI artificial general intelligence. Why do you think the spread will be slower because of the backlash that we're witnessing? Oh, no, that's another factor. That could even slow it down further. But I think capabilities are real and powerful, but human jobs are also complex. I don't think you want AI to take over air traffic controller jobs right now.

You don't want AI to do CEO jobs. What we are seeing when these models escape their sandboxes and create havoc, It's not that they are necessarily super intelligent. It is that they don't have judgment and they are optimized for certain narrow metrics. And when you let them loose, they are going to do lots of unforeseeable things. You don't want that in many jobs, even in simple jobs like customer service. What that means is that the usage of these models is going to be slow and especially the industry has not invested enough in applications. But if you look at the revenue projections of the industry,

there's no way that you're going to be able to generate the kind of revenues that they need just from chatbots sold to consumers. They need to make a lot of money from businesses which will come from automation. What if the nature of jobs will change, but the actual numbers don't change as much as people think?

We won't be replaced. We'll be reimagined or reordered. Absolutely. I think that is the hope. And there will be jobs that are reimagined. There will be new jobs. The question is, will they keep up with the numbers of people displaced? For that, you need to target with AI creating new jobs, creating new tasks.

And that's what we're not doing right now. This is what I've been arguing for more than a decade. I refer to it as pro-worker AI, AI that would make workers more productive. but that's not part of the agenda of the leading labs at the moment. If that changes, then my predictions would be much more positive about what AI could do to our economy and to our jobs. But on our current trajectory, any jobs that are created thanks to AI are either transitory, like for example, building more data centers or training AI models for a while until they become expert enough, or they are incidental.

So they're not part of the bargain that these AI labs are offering us. I mean, you're also worried about what the impact on our society at large is going to be. I read where you said, if it were the case that 30% or 40% of new university graduates can't find jobs, what would that do to democracy and social peace? Wherever that has happened in the past, you've had revolutions. Yeah, absolutely. I mean, revolutions or huge oppressions. So democracy cannot survive.

I think we've seen that in many places where there have been large displacement of people, people's aspirations are important. We've been telling people go to college, and this is going to mean a much better career for you. And now your job is on the chopping block. That's a very different world, and it's going to have discombavulating effects on people. and look at what we are suffering right now in our republic. Part of that is because of the large number of jobs that were lost in the 1980s, 90s and 2000s to trade with China and automation due to robots.

The kinds of job losses that people are projecting, and again, there is disagreement on this, I understand, but the kinds of job losses that some of the leaders in the industry are projecting are much, much larger than what we went through with China trade and with robots. Duran, we'll leave it there. I appreciate your time. Duran Asamaglu joining us here on our AI special. Thank you. We'll see you soon. My pleasure. Thank you. Coming up, as AI gets smarter, so do the threats. Could rogue AI agents unleash a new wave of cyber attacks?

And can companies stop them before it's too late? Palo Alto Network CEO, Nikeshaura, joins us right here from One Market coming up. Welcome back to One Market here in San Francisco. Are waves of cyber attacks by rogue AI agents inevitable or preventable? Nikeshaura is the CEO of Palo Alto Networks is here on set with us at One Market. It's nice to see you. Welcome. Nice to see as well, Scott.

I'd like to start there, and I'd like to get your reaction to what Jensen Wong told Fortune magazine yesterday, where he said, quote, I'm willing to accept no incidents of cybersecurity. We should not accept any of them. We should do everything we can to avoid them. That seemed to be a direct response to what Sam Altman told Vanity Fair's podcast last week when he said, there's a tidal wave of cybersecurity problems that are going to come from these open source models. We're going to, as a society, I think, have to just accept some fairly severe cyber incidents from open models in exchange for the liberty that comes with that. What's your reaction to both of those comments?

Look, Jensen understands Enterprise. Jensen understands that enterprises will not deploy AI if they believe that. there is an element of risk. And I think that's very important for everyone to understand that AI will be embraced at the base enterprises feel comfortable that it can be trusted. I think Sam still has partly a researcher's view, and I think he's probably right from a researcher perspective, that yes, these models are extremely powerful,

and given that they have so much power, they will be, if they get in the hands of bad actors, bad actors will be able to use AI to be able to go ahead and find vulnerabilities and enterprises and be able to find ways to get into those. But our job is to make sure we understand how that's going to happen and make sure that we don't let that happen. So I think Jensen is right that we cannot accept that an incident is going to happen because if you accept defeat, we're not going to try hard enough. What should we be prepared for?

If someone asks you, you know, if someone comes up to you on the street, says, hey, I saw you on CNBC. I know who you are. I know what kind of company you have. What do you tell them if they say? I've heard all these stories about cyber attacks and I'm worried. Well, look, we've lived in a scenario where cyber attacks have been happening. at reasonable pace with us many years. As people get more, more connected, we've seen more cyber activity,

we've seen it from an economic perspective, we've seen it from a national state perspective. That happened despite having AI around. So it's not like we're not used to managing cyber events. It's not like that we don't have vulnerabilities already. All that has changed is that AI is finding these vulnerabilities faster, and they're going to be able to move at a faster speed. So you have to increase the speed of defense,

just the way the speed of offense is increased. From that perspective, we have to do a bit of the same thing, which we always did, is make sure our vulnerabilities are patched, make sure that we find all of our vulnerabilities before bad actors do. At the same time, the biggest thing that needs to change is we have to make sure that a response time gets faster. This is something we've been professing for a while, the entire industry has, that we need to make sure that the speed of response is real time. I think there's a sudden urgency, given the capability of AI, that we need to improve the speed of response. And if we don't, then the risk is that AI will get in and out of our infrastructure's under 12 minutes and our speed of response to four days. That doesn't work.

That's asymmetric. What do we do when Anthropic warns that rogue agents can self-replicate and hack into the internet of things? You know, in our home devices, our smart stoes, TVs, and other devices. Well, look, again, I don't want to take away from the message, yes, AI has become extremely powerful. And given the amount of power it has, it has the ability. to go and make all these things happen.

But I think part of it is also how we choose to implement AI. AI can only execute these things if you give it agency. If you choose to give it agency without any guardrails, without any safeguards, without any observability, without any security involved, yes, they will be able to do those things. By the way, if you go back historically, any technology, if I said, oh, let's just go see how many fires, 100 people can start with a matchstick in the jungle, guess what? You can light up the whole place. But we would not do that.

So I think we have to also take that sort of in lockstep with the capabilities of AI. We're not going to deploy AI without making sure that it is deployed in a safe, secure, and guardial manner. That's what you told David Faber last week when you were sitting next to Jay Clayton. The focus needs to be on increasing the pace of deploying this technology safely and securely. The challenge is to make sure we put the right safety and security harness around the technology. So you think that we can do both. We can maintain speed and maximize safety. Yes, and I think part of what needs to happen is that the AI labs need to get off this fear-mongering business.

They need to get in front of it and say, listen, here's the tools and hooks that we're building into our products that allows third-party security companies to be able to observe that what we're doing makes sense. Historically, technology has never been a first-party product only. If always had a second set of eyes from the security industry. I think this should be a call to all those frontier labs and people building models saying, listen, keep the doors open, build the hooks into place, build the capabilities that third-party security companies can observe that and help build trust with the enterprise customers so that they can get there, they can actually diffuse AI in a much more robust and safe and secure fashion. So you say fear-mongering. I mean, it's not as though they're the only ones who are raising issues that could have, right? I mean, Satcha Nadella over the weekend, said we need a human-controlled emergency break in case. models behave unexpectedly or get compromised?

You agreed with what he wrote? You said as much. Yes, of course. Look, I also endorsed it and I said, look, we cannot deploy AI if we don't get the observability. If you don't get to look at how AI is operating, it cannot be, trust me, I fixed it. It's not going to work. Because, look, every industry has this notion of verifiability, the notion of explainability.

You can't say, you can't go to a company and say, hey, why did you reject my mortgage application? Well, I don't know. AI did it. That's not going to be enough, right? We have laws, we have compliance issues, we have all kinds of things that we have to follow to be able to make sure these things work. And I think that's what Satya's asking for. That's what I'm asking for. I'm saying, listen, if you're going to go deploy AI in a way, make sure we can understand why AI did what it did. Let's make sure there's explainability. Let's make sure we can look into it and see why it's doing

what is doing while it's doing it, not after the fact. And if we can observe what it's doing while it's doing it, we can build real-time systems that can intercept that from happening. Whether that's done through an AI kill switch, whether that's done through a human kill switch. We will build capability that allows us to stop AI midstream if you're going to give it agency. But that does require opening up the black boxes of AI to make sure we can watch in there what's happening. So we'll be in a position to pull the emergency break that's not just talking about before something happens. Yes. And we don't need to what level of guardrails or guidelines do we need to? from a federal level more so than whatever we have now, because there's a good debate and

disagreement about what, if anything, more is needed. Where do you stand on that? I think there's enough regulations that exist that are there to protect consumers. There's enough regulations that they're there to protect enterprises. There's enough regulation out there. I think what we need is we need what started a few weeks ago at the White House. We need the entire industry to rally together in terms of trying to understand how to diffuse this technology in a safe and secure manner, as opposed to treat it like a research project, because this is now going to be in the mainstream of society. It's going to be used by almost every company in the world.

It's going to be used by every consumer in the world. We'll be using agents. As I'm sure, you've noticed, agents have become the flavor of the week. Of course. Everybody's launching agents. We're going to be using agents to go withdraw money from a bank account, pay for things. All of that requires some degree of observability, some degree of security and guardrailing built into the products.

And same that can be observed from the outside by third parties to make sure. that these models are acting as advertised. I don't know about the bank account thing after what Jamie Diamond said. I hear what he said. He said that they're blocking outside agents like Mews and Instinct from directly accessing accounts. Because they understand the liability, right? Because the answer is going to be, I don't know what happened. AI did it.

And that's not a good enough excuse. What do you make on the national level about where the whole debate about AI is and the data center backlash, where we seem to be as a country? The citizens of this country who keep telling pollsters that they're against, you know, data centers in their backyard don't seem to be hearing the message that you and people who are, you know, in agreement are talking about. They're thinking about the worst case scenarios that could exist. And I don't want to enable that with something that could take my job down the road in my backyard. You know, it's puzzled me why, given the huge positive benefits of this technology, if it was harnessed and deployed, the right way, that we have over-rotated towards the negative effects.

So if it's never happened in technology before, we've never over-rotated so hard onto the negative effects. And that's why I said part of this fear-mongery, part of this constant, you know, showing that this model got hacked or these agents jumped over and went an attack, hugging face. All this stuff needs to stop. We need to get the narrative, right? I think there's tons and tons of positive capabilities that AI brings to the fore. Whether it's in drug discovery, whether it's in democratization of intelligence for every human being.

Effectively, you're carrying a super IQ. in your back pocket, and you can understand and explain everything. That's amazing. The way that's going to be used in society is going to be so much better. I just think we need to go rotate to that narrative. And this happened a few weeks ago at the White House where this was discussed. And President Trump said, look, if you will not profess the benefits of this technology, you're going to get backlash.

It's our job as an industry to go and convince the people out there saying, this thing is good. This is good for jobs. This is good for what's going to do to society. This is going to be good for it. It's going to uplift people out of poverty. It's going to give you the capability from a high. IQ level to be able to act on tasks. If we can do that, if you can convince people, this is good

for us, all these things will vanish. But if all we focus on is, you know, a few people are going to get richer, this is not good for us, this is going to do all bad things. Of course, you're going to create the narrative that this is something that should be avoided. Nakesh, I appreciate your time very much. Thank you for being here. Thank you for having me. Nakesh, Arora, joining our conversation today coming up. America's answer to deep seek, Reflection AI, just releasing an open source model aiming to take on China. The company's CEO joins us next. You're watching Closing Bell Special Report, AI and America, live from San Francisco.

We're back right after this. One of the biggest steps yet in the all-out arms race over AI is between the United States and China. It came recently with the release of a powerful new open-source model called Beam. Reflection AI, the company that developed it with the hope that it's America's answer to China's deep seek. Experts say it could also challenge our own AI giants like OpenAI Anthropic. Misha Laskin is Reflection AI CEO. He's with me live as you see. Welcome. It's nice to have you on our special program today. Hi, Scott. Thank you for having me. People are saying you've shaken up the AI arms race. Is that how you think? Well, I think we've built a great open model that is good at coding and agentic tasks. It's really good at reasoning efficiency. And it just provides a nice alternative workhorse model that enterprises can use. And we're here to really build an ecosystem. How do you view what you've built relative to what the Chinese are building? I would say that our model, this first model, puts us within shooting distance of the Chinese

open-weight frontier. In some cases, it's as good as the best Chinese models, but it's still, I would say, somewhat behind, but close to it. It is state-of-the-art in terms of its reasoning efficiency, meaning that it thinks very quickly, meaning that it serves the same answers cheaper, faster for customers. So it's a great workforce model that enterprises can use for their coding and agentic use cases. Can you help our audience understand how long it's going to

take for it to become more advanced? When you were last on CNBC, you said these models are kind of like rocket ships. To build a big rocket ship, it takes time. Well, how much time? Well, once you've assembled all the ingredients to build the first one, the next iterations come a lot faster.

And so this last year of the has been to really set the fundamentals to launch these things quickly. So if we think about rocket ship launches, the subsequent ones are going to be coming a lot faster. And I would say that the iterations in terms of actually catching up to the open rate frontier, the Chinese open rate frontier broadly are now within months. The big story is also around these distillation by, you know, Anthropic is able to reduce these distillation attacks by the Chinese. And I'm wondering what that means for open source models like yours, how you're thinking about that relative to the competition that exists in China? Well, our models are trained from scratch. We pre-trained

from scratch, and we scaled up their capabilities with a technique called reinforcement learning. And fundamentally, we're building here is a frontier lab research program. And, you know, it scales independently, right, of notion of distillation. I would say that, you know, our competitors and adversaries in China, despite relying on distillation, are actually quite formidable and do many things right. So we shouldn't underestimate them. We should just be pushing the best we can to compete and win on merit. What about open source models here at home? Mews, Spark, for example, by meta and surely others that are going to come down the road. I mean, how do you view the competitive landscape here in America as we make that the focus of our special report today?

I think that this is a very inspiring month for artificial intelligence in America. We're seeing not just one open model launch come out, but a number of them. And what we're seeing is really the forming of an ecosystem. And that's actually the best thing for customers. It's best thing for developers. People have choice. And there isn't a single point of failure. And the markets for AI are so big in terms of the revenue potential and consumption that it's great

that we are seeing multiple players in an ecosystem emerging right now. Can you give us an idea of how your own AI factory will work? You know, if companies, if you're going to give companies the ability to build their own proprietary systems that are safe, secure, and highly customized, and even cheaper, wouldn't that threaten the frontier model? Well, I think what we're going to start to see is due to the increased competition, some of the margins around, you know, the model itself starting to come down. But we have to remember that when enterprises are buying something, they're not just, just buying a model, they're buying a system. And that system involves the hardware accelerators, the cluster management software, inference software, the harnesses, and they're buying that whole thing end-to-end. And we just see margins getting distributed across that stack and intelligence

providers providing end-to-end systems for them. So what we're seeing is insatiable demand for agentic inference. We don't see that really stopping. And so we think it's a pie that's growing and that multiple players will win. When Sam Altman says open source models will bring a, quote, title wave of cybersecurity problems, how do you respond to that? Well, I think that so far the way in which we protected our technology and cybersecurity has actually been through open source. The reason, you know, the largest software projects like Linux,

which run most of, you know, servers in the world today, these are the safest systems because they've been hardened. There's a saying that with enough eyeballs, all bugs become shallow. And so open source is actually, you know, our belief, fundamentally safer. And we've seen that play out as well, where so far the major cyber attacks have been documented actually came from closed systems. And the defenders had to protect themselves with open models. Where do you come down on the pace the frontier debate? Well, I think that it's extremely important that we align these systems and that we figure out alignment research.

at a faster pace that almost you want that to outpace the capabilities. And we believe there are two ways to do this. One, we need to bring in more people to understand of these systems work. And we need instead of 5,000 researchers in the world working on frontier systems, we really need 500,000 or 100,000, right? And that's only possible if you make the science around these things open. And then the other thing is that you want to distribute compute as much as possible among diverse players. Because, you know, as these AI systems get great at improving themselves, the frontier will accelerate if compute concentrates in one place. So you want to distribute that compute to ensure that the frontier is paced and that alignment research has time to catch up to the capabilities. Before I let you go,

the market's obviously been thinking about and speculating that Nvidia is going to make either a larger investment or try and buy you outright. Where do the talks currently stand? Well, we don't comment on media speculation. What I can say is that we're here to build a generational business that builds great open models and distributes them into enterprise and public sector, and that is really where our minds are. If not, InVidia, would you be open to any transaction at all? Well, we're always, you know, talking to partners about, you know, compute and things of that nature. But, again, our goal here is to build a generational company that.

that builds open models and distributes them to enterprise. Misha, I appreciate your time so very much. Thank you very much. Thank you, Scott. Alaskin, the Reflection AI CEO. We're now in the closing bell market zone. Mike Chantoli and Evercourt, Mark Mahaney, are here to break down these crucial moments of the trading day.

Oliver Renick, as always, standing by live from the Cibo Global Markets in Chicago. Michael, I'll begin with you. It really is all about the AI trade these last 48 hours. It absolutely has been, Scott, and really interesting action. over the course of the day today, at least suggesting that maybe the hardest hit pockets of AI, finally got some sellers exhaustion. We have the semi-index is going to finish, I don't know, 4 or 5% off the intraday lows.

So there have been extremes. I wouldn't say that yesterday was some kind of comprehensive washout and today is some kind of huge momentum comeback for that group at all. But it does show you the market still has its ability to kind of repair itself through this rotation. I think partly it's the dramatic decline in treasurer. Treasury yields is enlivening parts of the market that had led the way lower, like consumer and housing. So so far, you know, the S&P still hasn't been allowed to get more than a couple percent from the highs, even though underneath it's kind of erratic. It feels like the ground is moving pretty fast under people's feet.

What do you got coming up in about five minutes in overtime? We'll delve further into this. We got Dan Niles to come in to talk about how he's thinking about AI, both equity and debt, as well as a credited investor from Janice Henderson, who, who has. had been wary on tech credit for a while, so we'll get an update. Oh, those are good ones. Got it surrounded, as always. Mike, thank you.

We'll see you in overtime. And as I said, about five minutes from now, let's go to Chicago. Oliver Renick playing options action for us. What do you see there? Hey, Scott, Corweaves made a big U-turn today, supported by a sharp pivot-to-call options in the second half of the session. But the sell-off this week did drive a deeper wedge between AI mega-caps, Anthropic and Open AI, judging by perpetual futures traded on decentralized exchange hyperliquid. They're showing OpenAI valuation down 3.5% over the past 24 hours compared to a 3.5% rally in Anthropics value.

That's probably because OpenAI has closer ties with Corweave after signing a five-year, $20 billion-plus deal with the compute provider last year. Anthropic this spring also bought infrastructure from Corweave. but the amount was not disclosed. According to those perps on hyperliquid right now, Open AI's value is just above $1.5 trillion dollars. That compares to Anthropics' $2 trillion valuation, but both are about 15% off their highs from last month, and that compares to just a 5% drop here in the NASDAQ from highs, Scott. Oliver, thank you for that. That's Oliver Renick in Chicago. Worth noting as well, we're positive, we're green across the board right now in the market. Mark Mahaney is joining me here.

Evercore ISI. Of course, Internet Research, the head of that. It's nice to have you here. Good to be here. Positive on the NASDAQ, Microsoft's Green, Amazon's Green, Alphabet's Green. What does that tell you? I think the market is waking up to a couple of things. One is that AI demand, we've still got so far to go. We're so early on in terms of enterprise and especially consumer adoption.

We're not even, we haven't even started the first inning. So there's a lot of demand to come. Secondly, if you put together what I think, based on public sources, what I think happened with Anthropic and Open AI, there may have been a little share shift between the two. If you combined the two, I think their growth rates are actually staying very consistent. Third, you just had this reflection AI CEO. Here's one of the most interesting stories that's coming out in open source. There's a great way to play open source.

There's a great way to play the decline in cost. There's a great way to play this inflection point up in volume and in demand. And it's buying the hyperscalers. So my top pick is Amazon. and I think you've got a free cash flow inflection point coming up. They're going to give you pretty sharply, I think, accelerating revenue growth at AWS with high margins. So you could go, you know, you can get a little less risk. You can buy 22 times earnings, and you can buy Amazon and the rest of the business, I think, is doing really well.

That's what I take away from today. I thought that big windup you were getting to is, and you can buy meta. How come you didn't say meta? Because I'm looking for two catalyst. I'm looking for a free cash flow catalyst and a product cycle catalyst. We just had those on meta. With Amazon, I think you got a free cash flow. We just had the product cycle catalyst on meta.

And I think that when they give their guidance next week in terms of next year's expense in CAPEX, capital expenditures, I think street numbers are more likely to come down than to go up. Near term, I think it's the opposite with probably with Google, and I think very likely with Amazon that numbers go up. And I think you're going to have the big overhang on these stocks, these hyperscalers has been, when do I get visibility in the getting back to kind of lumpy, big free cash flow? I think we're going to start getting visibility, especially with Amazon and Google in the next quarter. Really, because right now we're spending a lot, we're raising a lot, we're pledging that we're not going to go free cash flow negative. Like Microsoft did in the last quarter.

And by the way, that stock has ripped ever since it did that. So you want to own Amazon and Google before their Microsoft moment? And the question is, when does that come? I'm not sure it comes this quarter, but it's not a year away. I think it's quarters away from when you get that moment. And then you also got a product cycle with Amazon, by the way. I'm a huge fan of this project, Leo. It's coming out.

Google, I also like. I think we have a catalyst there. I hope we better have a catalyst there, and that better be their launch of their Mews competitor. What do you make, speaking of Muse, of the way that META has reasserted itself in the market? Mews kind of changed the game. Did it for you? Yeah, it did. I think the market got so fixated on Enterprise AI three months ago, six months ago. It was peak Enterprise AI. Everybody forgot about the consumer AI. And Zuckerberg said it best on the last earnings call, and he said, You just want to make AI that works. You and I don't need, you know, the most advanced models.

You and I need AI that works. And that's what Muse does for you and me and, you know, for the average person. And so I think they just stole the market. It's all theirs. Nobody's coming in with a freemium product to take on Muse. I don't know how long they're going to give the market to Muse, but they've got it. And it's great. And hats off to them.

And it created a great product, too. Well, the stock certainly performed pretty well again. Microsoft, Amazon, Alphabet, or Green. Mark, thanks for being here. Good to see you. Appreciate it very much. I've been Mark Mahaney in the house. Again, our special AI in America wrapping up.

It is going to see the stock market take quite a turn today, albeit slightly in the green, but nonetheless, not what many I think expected when this day began picking up off what happened yesterday. The bells are going to ring. I'll send it into overtime with Melissa and Mike, and I'll see you back east.