11AM Hour: Goldman Sachs Chief U.S. Equity Strategist, Midterm Elections Impact on the Market & Micron to $3,000? 10/8/26

0:00We're an agent supposed to take things off your plate. Building another to handle the others is a bit much. Don't make more work out of work. Gemini is the one agent that knows all parts of your business and actually gets stuff done. You give it a project, close your laptop, and when you come back, the job's finished. No matter your industry or role, your agent has the skills to help every team stay in their flow. Seriously, make work less work with Gemini on Google Cloud. Check response, subscription required. Traders don't need a prettier app. They need a better angle. Schwab's think-or-swim trading platforms were built for real traders like you.

0:35They have an insane amount of chart studies. I'm talking over 300 studies for virtually any strategy. They also have integrated risk management technology and nearly unlimited customization so you can shape the platform to fit the way you trade. And when you do that, you can analyze trades from angles you won't find anywhere else. Go see what that looks like at schwab.com slash trading.

1:01Good Thursday morning. Welcome to Squawk on the Street. I'm Carl Quintanillo. It's Sarah Eisen here at Post 9 of the New York Stock Exchange. Today, Goldman's chief U.S. strategist, Ben Snyder, is with us. Why he thinks markets are overpricing the magnitude of coming Fed tightening, and he'll explain why. Plus, a bullish call in the memory space. One top analyst making the case for micron shares to nearly triple from here to$3 ,000 a share. He'll join us to explain why. And more AI safety warnings. A group of former OpenAI employees calling for increased monitoring of the ever-improving technology.

1:33We'll talk to one former tech insider who was early to sound the alarm and says catastrophic risks are not the price of progress. Speaking of progress, we made a little bit after the Open this morning. Pre-market was pretty weak. Dow briefly went positive, giving some back now. Obviously, all about oil and rates. Ten years pinned to 5.3. Brent, still above 105. Interesting backdrop. We'll get some Fed speak today. And corporate earnings are just starting with Levi and Pepsi. And another good read on jobless claims, you know, worth mentioning. Every week now it comes in low, which means that fewer Americans than expected, maybe in this point in the cycle, are facing layoffs.

2:10That's a good sign overall for the economy and the consumer. But energy is leading today. Oil prices up. And I guess the market is just grappling every day with how to deal with rising oil and rising yields, which come now hand in hand. Yeah, there's a look at the claims, as Sarah said, lowest since July, basically. For more on the markets, let's bring in Ben Snyder, as we mentioned earlier, Goldman Sachs, chief U.S. equity strategist. Talk about the House view. It's good to see you, B. Welcome back. Morning. In general, the view is that core PCE does get back to, what, 2-2 by the end of next year, and it's hard to imagine why the Fed would want to risk a mistake in that environment, right?

2:43If we zoom out and we look at the macro economy, things look pretty decent. As Sarah mentioned, The labor market looks like it's in balance. There's no buildup of underlying inflationary pressures and wages or otherwise that would really advocate for an extended hiking cycle. So we expect another hike in December. But after that, we think probably the Fed is done. Really? Does that rhyme with history to have one and then another and that's it? We had one example of that in the late 90s. And, of course, after that, equities traded quite well for another few years. Yes. And then I think you're 12-month on the 10-year's 4-4, which would provide a lot of relief to all kinds of rate-sensitive sectors.

3:20That's right. Day-to-day, as you mentioned, we're trading with oil, which means requiring a view on geopolitics, which is extremely difficult. But if you zoom out, it looks like we've put a lot of risk in the price already. We see a lot of risk in equity investor positioning data as well. And meanwhile, as you noted, we're about to kick off an earnings season where the S &P 500 is going to report growth of probably over 30 percent. What about the notion that rates aren't responding only to inflation, that there's a huge game of competition for capital? People are worried about deficits and all of that, that it's not just about what prices are doing.

3:52That's right. Unfortunately, large deficits aren't a new story. But what we've added over the last year or two is incremental issuance from the hyperscalers to help fund this build-out. There's only so much capital in the world. As you have more demand, the cost of that capital goes up. No, but I guess the point is that it's not just the Fed that's pulling maybe market interest rates higher, right? Bond yields continue to rise. Isn't that problematic for stocks? I think if you look at valuations, we've put a lot of that in the price. A year ago, we were sitting here and the S &P was trading at a 23 times multiple.

4:22Today, we're trading at 19 times. It's true the index has remained resilient, but of course, that's really due to earning strength, not valuations. Can the market handle decelerating earnings growth? Or maybe we don't get it. I think that should be the base case. We've enjoyed an environment of acceleration for the last few quarters. I'm not sure exactly whether the peak is Q2 or Q3, but it's pretty clear as you look into 2027, we're losing a large part of the stimulus from easing Fed policy. We're losing stimulus from the One Big Beautiful Bill Act. AI CapEx growth will remain strong, but it's not accelerating anymore.

4:55But even with all of those factors, we're still expecting double-digit earnings growth next year. So it still looks like a pretty healthy backdrop. And that gets you to, what, 8 ,700 in 12 months? In 12 months, 8 ,700. Led by which sectors? So, of course, the core of the earnings strength is still AI. We wrote a note about this a couple weeks ago, but about half of S &P earnings growth is being driven by the AI build-out at this point. Investors shouldn't give up on those sectors. But, Sarah noted, the labor market still looks pretty healthy. Even if you look at the median stock, earnings growth is at a double-digit rate right now.

5:25When do you guys think that the AI trade broadens out? I mean, isn't that the hope that we'll see a real enterprise level productivity and earning strength and ultimately growth as a result of all this investment? I know we're four years into this. I would argue we've actually just started to broaden in the last few months. If you look at the last couple quarters, the hyperscalers have reported an acceleration in their AI-related revenues. That's given investors a lot of confidence to expand outside of just the infrastructure stocks to things like cybersecurity companies, other parts of the application layer.

5:58I expect over the next couple quarters, we're going to increasingly see evidence of AI-related earnings productivity across the market, and that will allow a further broadening. Do you like software or no? Unfortunately, I think much of the narrative several months ago was correct that eventually we will see some disruption. I know investors are still waiting to see that show up, but it's early in the adoption process. Of course, there will be winners and losers, but from a big-picture perspective, AI spending has to come from somewhere. Some of that funding is going to come from software. So which applications do you like then as AI moves on?

6:29We still think the data infrastructure stocks are quite attractive. Cyber security has traded quite well, although it's also pretty well priced at the moment. Industrials have really taken it on the chin, I think, versus the S &P. It's close to some multi-year lows. Does that recover? So to your point, Carl, if we're in a decelerating growth environment, a lot of the cyclical parts of the market that have performed well over the last 12 months really shouldn't continue to outperform. I would put small caps on that list. But industrials can usually do well in that kind of environment. I would expect those to recover if we get a little bit less volatility on the macro front.

7:00So part of the thesis is lower growth than we've had? Slower, but still very solid growth into next year. Economic growth. Economic growth has been running close to trend. Our forecast suggests it's going to be close to trend. Of course, we know that earnings growth has really overshot that because of the strength of the tech complex. I'm trying to think. I mean, I think Jan's, correct me if I'm wrong, but Jan's general view is that real disposable income growth is going to be a bit troublesome and it will lead to consumption being subpar. Is that also part of your thesis? That's right. You know, the equity investor playbook would tell you, you want to buy consumer stocks when growth is accelerating, when unemployment is falling.

7:36That's not the part of the cycle we think we're in today. Unemployment remains very low. The consumer is fine, but it doesn't really look like they're accelerating from here. So you would avoid things like discretionary or staples, anything like that? I think that's right. I think small caps, stocks with weak balance sheets, consumer stocks, these all tend to be early cycle trades. I think we're past that part of the cycle. Will midterms be material? I've seen thesis on both. The consensus view talking to investors is to follow the historical pattern, which is that you build up risk premium ahead of it because of uncertainty.

8:10And actually, as you move into October and investors gain more confidence in the outcome, markets trade quite well afterwards. Of course, the path of the S &P... No matter what the outcome is? Correct. Correct. Of course, I would rank that lower than oil and interest rates in driving the market today, but it should be helpful to resolve that uncertainty. Do you have to have a view on oil to predict what happens to stocks? I think you have to have a view that oil isn't going to go materially higher. I mean, one thing that gets overlooked is it's not just about the level of these macro inputs, it's about the volatility.

8:36We've seen this pattern repeatedly historically. And so if prices stay where they are and just become a little bit less uncertain, I think stocks will continue. Although, I mean, the word from a lot of oil executives this week alone has been that the buffers are diminishing and that the risk of more volatility is still with us, at least. I think we see that very clearly if you look at futures prices looking 12 months out. Oil has been steadily increasing. Our view is that's reflective of that risk and uncertainty. Right. But, of course, all of this comes back to geopolitics, which is very hard to find.

9:05So your bottom line is market will do well, but it's going to be very lopsided. It's not going to be driven by cyclical, which has been kind of the story. I think that's right. Now, look, tomorrow, if crude sells off$20 a barrel, it's going to be a good day for cyclicals. But if we zoom out a little bit and ask ourselves, where are we in the cycle? I think you stick with secular growth over cyclical growth. A lot of known unknowns, Ben. Thank you. That's true. Good to have you. Ben Snyder from Goldman. All right. Let's take a closer look at the move in oil today. Speaking of, it's been such a pull on the market.

9:33And we're seeing now the highs of the day, up more than 5 percent on both Brent and WTI crude. Pippa Stevens has more. Pippa. Hey, Sarah. So oil is popping here with Brent back above 105 on the back of increased tanker attacks in the Strait of Hormuz and the Houthis targeting Saudi infrastructure, which is stoking escalation fears. As President Trump said, he no longer wants to deal with Iran. And NBC reports he's discussed possibly resuming large scale U.S. military operations in the coming weeks. Now, more than 10 tankers have been struck since September 28th. That's according to Kepler, which is the most since the start of the war.

10:09Shuttle tankers have also been hit. Now, ship-to-ship transfers have been key for getting oil out of the strait. And if owners decide those transfers are now too risky, that could curb flows. Meantime, Saudi Arabia said three people were killed by Houthi attacks, with the Iran-backed group hitting the Riyadh airport, among other locations. And here in the U.S., Hurricane Isaias has forced the shut-in of more than 500 ,000 barrels of oil production in the Gulf of Mexico. That's about 25 percent of the supply from the basin. But arguably more important is whether the hurricane impacts any of the Gulf Coast's refining infrastructure.

10:42And so the storm's path is being very closely watched. Now, with all this uncertainty, North Sea dated Brent, which is oil, for the most immediate delivery is now at$129.74 per barrel. That's according to Argus. Guys? Really quick, Pippa, this news out of the IEA that this release of diesel and crude reserves is not incremental. Was that a disappointment? I think definitely. And there was a lot of confusion about that. And we are still even waiting on the breakdown of what it's going to be in terms of diesel versus oil. We have heard that diesel will be front loaded, but the IEA is not meeting again until October 14th.

11:13So still some time here. And I think that the market was hoping for more, not just, you know, a reiteration that this is the same stockpile that was announced back in March. And we are seeing diesel futures move back up. We're now only about 7 % below the high of about 528 from September. Patma Stevens, thanks so much. Important topic to stay on top of the energy complex. When we come back, we'll talk to a top analyst on Micron. Just hiked his target to$3 ,000. We'll break down the bull case with him next. Plus, we'll speak to a former Google and Microsoft engineer why he says we should not accept catastrophic AI risk as the price of progress.

11:49We're back in a moment.

11:54Legacy credit scores only look at a brief snapshot in time, and that limited view often leaves millions of qualified credit seekers out of the picture. Vantage Score is changing that by modernizing credit scoring. Vantage Score uses 400 % more credit and alternative data to create a fuller picture of financial behavior, including trended credit data, rental payments, and timely utility payments. Vantage Score safely broadens access to the credit economy and brings 5 million mortgage-ready borrowers into the market. That's why top lenders like Rocket Mortgage, as well as nine of the top 10 banks, rely on Vantage Score 4.0.

12:31With higher predictability and greater access to credit, everybody wins. Get started at VantageScore.com. Weren't agents supposed to take things off your plate? Building another to handle the others is a bit much. Don't make more work out of work. Gemini is the one agent that knows all parts of your business and actually gets stuff done. You give it a project, close your laptop, and when you come back, the job's finished. No matter your industry or role, your agent has the skills to help every team stay in their flow. Seriously, make work less work with Gemini on Google Cloud. Check response, subscription required.

13:06Traders don't need a prettier app. They need a better angle. Schwab's think-or-swim trading platforms were built for real traders like you. They have an insane amount of chart studies. I'm talking over 300 studies for virtually any strategy. They also have integrated risk management technology and nearly unlimited customization so you can shape the platform to fit the way you trade. And when you do that, you can analyze trades from angles you won't find anywhere else. Go see what that looks like at schwab.com slash trading.

13:37We have some news on Webull after that big plunge yesterday on that congressional report around their ties to China. Our Eamon Javers has more on that. Hey again, Eamon. Yeah, Carl, that's right. We broke this story yesterday about this commission report up on Capitol Hill that called Webull a national security threat because of its structural ties to China. We had an initial statement from Webull yesterday. We've just gotten within the past couple of moments a more fulsome response from Webull to the House committee report that came out yesterday. I want to bring you some highlights of it. They say they fully support the mission of the House China committee.

14:11And they say after reviewing the report released by the committee last night, Webull also supports the committee's recommended actions for improving U.S. financial regulatory frameworks. Webull, however, takes issue with the committee's process in preparing its report. The report draws sweeping conclusions about Webull from an incomplete record and gets important facts wrong. And in this lengthy response here, this online trading platform is saying that the data that they have from their American users is kept in the United States and that the company itself is responsive and responsible to U.S.

14:48financial regulators. Guys, back over to you. So what happens? Do they take this back to that committee to try to get a correction or something? Well, I mean, it's not an issue of a correction, right? What the committee suggested yesterday was that there's a structural problem with this company because its ownership is in China. And they're saying that the company, in its public-facing imagery and marketing, just doesn't acknowledge that or doesn't let that on, even though it is disclosed in many of the company's corporate filings. So the committee is red-flagging that as a problem for American national security.

15:24This is the company pushing back. We'll see what regulators do on this. But I think this is very similar to the TikTok situation in which you have some real concerns about what's happening with American data with Chinese owned or controlled firms, where that data is going, who in Beijing has access to it and what they're doing with it. So I don't think this is going to be resolved anytime soon, but we'll see if regulators decide to sharpen their focus on this company. Yeah, yeah, sort of a buyer beware. Thank you, Eamon. Eamon Javers. Micron shares lower this morning, even though Samsung's record profit should bode well for the memory maker's outlook.

16:00Micron soaring throughout 2026, including a 4 % gain yesterday, driven by our next guest's incredibly bullish call. Let's bring in Gil Luria, managing director at DA Davidson & Co. Raises price target on the stock to$3 ,000, now highest among all the analysts covering Micron. Gil, why now? Yeah, we think Micron, the demand for AI is so great. The demand from memory is so great that Micron has a path to grow over the next three to five years. And if you look at the closest comparable market is the CPU market and stocks there, AMD and Intel are trading at 40 times earnings or more, while Micron trades at six times earnings.

16:38That's one of the biggest dislocations we've ever seen. So we understand this is going way out on a limb, but we almost feel an urgency to create awareness that being underway here is almost a risk. So if you're a long-term investor, the market cap of Micron is 3 % of your NASDAQ 100 benchmark, but earnings over the next 12 months from Micron will represent 10 % of the earnings of the NASDAQ 100. So if I'm right and they tried up to a market multiple, that one position that you're underweight if you don't own Micron will result in you missing your benchmark dramatically. And if you're short, the company will start buying back stock December 9th.

17:25That's when their limitations on the buyback expire. And they've said they're going to use 100 % of their excess cash flow to buy back stock. And they're going to do that opportunistically. Those are the worst things you can hear as a short seller. So you have an expiry on how long you can be short this stock. So we've decided that, yes, it's an unusual measure to have a price target. It's three times the current stock price. But all that does is represent a market multiple. And if a company can grow more than the market over a cycle, it deserves at least a market multiple. Gil, why hasn't it?

17:59Why doesn't it get a market multiple? Why is it only trading at six times? It's not like anything you're saying really is not well understood or known by the street. I would argue that it is not well understood, that people's mindset is that this is still the micron of 2023 that was selling a commodity, a highly cyclical commodity in that market. And what they're selling now is completely different. Three years ago, memory was entirely used for storage. Now, memory is how AI works. The more memory you put to a model, the better it works. The more memory, the faster the model works. The more memory, the bigger the context window.

18:38The product is now going to be sold more than half in five-year deals. That's not just a deal structure. It's because your Microsoft, Amazon, and Google are designing memory into their data centers in a non-fungible way. So the product is no longer a commodity. It's no longer sold like a commodity. The market has not wrapped its head around it yet. The stock has gone up only on earnings growth. the multiple expansion hasn't happened yet. And as we talk to many investors, we feel like many investors still have not wrapped their head around that. The changes have been so abrupt just this year that we think investors, as they catch up, they will be willing to pay more of a market multiple for this, which is not to say that they'll pay a multiple closer to CPU stocks, which again are trading at 40 times and above on what is really an equivalent market with the same cycle.

19:35It's the same cycle. The AI cycle is the same cycle for GPUs, CPUs, and memories. When the market wraps its head around that, the multiple should expand very dramatically. I keep thinking about what Merotra said a couple of quarters ago about humanoid robots using, I think he said, 10 times the memory of an autonomous vehicle. And I wonder whether or not the street needs to see new form factors to believe them. Well, I wouldn't be talking about that if Micron was already trading at 30 times earnings, because that's the next leg of growth three to five years out. The market isn't even giving Micron credit for what it has right now, which is the AI data center business.

20:15It's certainly not giving it any credit for the fact that anywhere you do compute. So as we talk about computing going to the edge, to our PCs, to our handsets, ultimately to robots, the memory, the GPU content, the CPU content, the memory content in those is actually going to be even greater than the data center. And that's really not reflected in any of these stocks, not least of which in memory. I guess, Gil, one of the arguments or pushbacks can be that the shortage eventually will catch up, maybe in two years or so. And then what's on the other side of that, especially if we're looking at cooling demand for AI?

20:54Eventually, you know, it's not going to go on like this forever. That's right. But that's true for CPUs and GPUs as well. And it's the same cycle. So we're going to have excess supply at the same time in GPUs, CPUs and memory. And yet CPU stocks 40 times, GPUs 20 times, Micron six times. OK, Gil Loria, pounding the table on Micron. Thank you for joining us. Thank you. Still to come this morning, shares of PepsiCo are higher despite the company cutting its earnings forecast. We'll get you some fresh comments from the CEO about what he's seen when it comes to the consumer. Plus, recently fired OpenAI researchers writing a letter to the board over managing AI safety, according to a new report.

21:35We're going to break down that story for you in just a moment. If you like YouTube, you'll love YouTube Premium. Hi, I'm Sean Evans from Hot Ones, and I want to tell you about YouTube Premium. It has offline downloads, so you can watch without Wi-Fi. Background play, so you can lock your phone and it still plays, baby. Oh, and it is completely ad-free. Yes, I said it, ad-free. Try YouTube Premium for two months free at youtube.com slash premium. Trial eligibility varies, terms apply, cancel anytime. Weren't agents supposed to take things off your plate? Building another to handle the others is a bit much.

22:15Don't make more work out of work. Gemini is the one agent that knows all parts of your business and actually gets stuff done. You give it a project, close your laptop, and when you come back, the job's finished. No matter your industry or role, your agent has the skills to help every team stay in their flow. Seriously, make work less work with Gemini on Google Cloud. Check response, subscription required. This episode is brought to you by Schwab Market Update, an original podcast from Charles Schwab. Schwab Market Update, an original podcast from Charles Schwab, provides an information-packed daily market preview on stock updates, U.S.

22:50and global economic news, monetary policy decisions, and key results and statistics that may impact your trading. Download the latest episode and subscribe at schwab.com slash marketupdatepodcast or find Schwab Market Update wherever you get your podcasts.

23:10Welcome back. Shares of PepsiCo are higher this morning after the company beat expectations on the top and bottom lines. PepsiCo also cutting its earnings forecast as the North American business continues to lag. We did speak with CEO Ramon LaGuarta last hour about the consumer, what he's seeing right now. Have a listen. We see the consumer resilient. We see, as you were saying, at overall level, the consumer is very resilient. Now, every consumer is different. They're clearly becoming very nuanced in where they invest their money, where they think it's worth the spend. So for companies like ours and the standing consumers very granularly dissecting every consumer and being able to provide on a very almost personalized way what they want is critical.

23:56That was Ramon LaGuardia, CEO of PepsiCo. So, look, the story with Pepsi, Carl, is that the North American business, both Frito-Lay and beverages, have lagged and underperformed. And, you know, there's a cost-cutting and productivity story, which we talked about, which I think investors probably want to see more of that, at a time where cost inflation is rising, which we talked about with Ramon, from oil prices to agriculture prices. So challenging macro environment, challenging changes in consumers like GLP-1s and habits and making sure you have the right portfolio for that. And then in North America, where they, at least in beverages, continue to lose share to Coca-Cola now several quarters in a row, something we talked about.

24:40He says we have the right, we have the growth names. They made the Papi acquisition. But I think investors want to know what's next, especially with the lower guide with one more quarter left of 2026. Yeah. 475 dividend yield. Give us a five year on Pepsi, too, because it's almost a$200 stock. Remember the days where they would print organic, like almost double digit with both pricing and volume working? Right. It was when every quarter they released some new flaming hot product, Cheetos or Doritos. And it was a big, big driver. So that is, I mean, the story, there's still strength in international, which is why I think they were able to meet their targets and why the revenue growth is still there.

25:22But it's the North American business and the share and the share problems, especially in Pepsi, which has been a problem. And then we've been playing. We've been showing that Coke versus Pepsi chart nonstop. Yeah. Opposite directions. The pure play versus the snack. Which is why you can't completely blame the macroeconomic environment. And as Ramon, he didn't. I mean, he said the consumer is resilient, but there are affordability challenges. Yeah, amazing. Let's get a news update this morning with Seema Modi. Hey, Seema. Hey, Carl. Here's what we're watching. Tropical storm Isaias. Surrending into the first Atlantic hurricane of the season this morning, the Category 1 storm forecasted to rapidly intensify as it approaches the northern Gulf Coast.

26:00It's expected to make landfall late Friday, early Saturday between Gulf Shores, Alabama and Florida. Evacuation orders have been issued for several communities in the area. Former Venezuelan President Nicolas Maduro hit with new charges in an indictment unsealed today. Maduro is already facing charges of participating in a conspiracy to traffic cocaine to the U.S. The new charges include allegations of torture and other human rights abuses. Meanwhile, Maduro's wife, Celia Flores, is set to appear in court today as her lawyers argue she should be released and put under house arrest because of her heart problems.

26:34American Airlines announcing it's expanding its partnership with Starlink and will install the space-based Wi-Fi on its entire main fleet fleet of more than 1 ,000 aircrafts. It expands a previous plan to equip more than 500 of its Airbus jets. Sarah, Carl, back to you. Have to say, I'm always excited when there's Starlink on the plane because the Internet works better. Thank you, Seema. Coming up, we're less than four weeks away from the midterms. We're going to speak with Morgan Stanley's head of public policy about how the elections could impact the market. We'll talk to a former Google and Microsoft engineer why he says AI safety is not the same thing as falling behind China when we're back in a minute.

27:19Welcome back. Recently fired open AI researchers writing a letter to the board about managing AI safety risk, according to a new report. Kate Rooney with the details. More safety warnings, Kate. Yeah, Sarah. So this report is raising some new questions about AI safety for OpenAI. The company also facing some accusations about how it won a prized math problem. The first part of this, though, does have to do with OpenAI employees being fired over what the company says was a violation of policies and, quote, improperly sharing confidential information with an outside safety firm. Those researchers are now pushing back in a letter obtained by The Wall Street Journal and sent to OpenAI's board.

27:58they do say that the firings could discourage others from speaking up about potential AI risk at a time when these companies are largely being left to self-regulate. OpenAI denies that the firings were retaliatory, says it supports the researchers' broader safety concerns. Separately, guys, an NYU math professor, Tristan Buckmaster, this morning on Squawk Box, accusing OpenAI of stealing his findings and then using them to solve a millennium math problem. I mean, the fact that they would front run their customers, I mean, it shows their sort of disdain for their customers. I mean, like, why would a company buy their product when, like, if you get an interesting ideas, they can look at your private notes and beat you to market?

28:45it. OpenAI calling it categorically impossible for Dr. Buckmaster's codex prompts over the last two months to have influenced the system, they say, in any way, including training. But guys, it does very much highlight this issue of trust for enterprises that are increasingly now relying on OpenAI, relying on Anthropic to run their businesses. Enterprise is such an important part of these companies as they look to go public. Kate, appreciate that. A good setup for our next guest. Let's stick with some of these AI safety concerns. We have heard from two leaders, two AI leaders this week who did push back on some of the doomsday scenarios.

29:20Take a listen. Five years, there are people who have been in the scrum who say that we could all be dead, which is a suboptimal situation. I think it's nonsense, Jim. We're not going to be dead in 2030 with a 10 percent chance. Well, then why say it? Why do you think just people just say it because it's like they're novelists? I think people have their own motivations for why they want to say it. You have to develop technologies in a way that promotes safety. And that's what can be done here. I think there's a real false dichotomy in that there's a tradeoff of speed and safety or a tradeoff of American leadership and safety.

29:56Let's bring in Nate Soares, former Google and Microsoft engineer, president of the Machine Intelligence Research Institute, also the co-author of If Anyone Builds It, Everyone Dies, Why Superhuman AI Would Kill Us All. Nate, It's good to have you on set. Appreciate it very much. Good to be here. I mean, so this conversation, NVIDIA sort of framed it the other day as in, when you think about the car, when they started out, they were dangerous. But if you invested in technology, you did wind up with airbags and seatbelts and now driver's assistance. Is there a way to think about AI through that lens?

30:31You know, I don't think that's a great analogy. And I also, there's one reason I don't like talking about AI safety as opposed to whether or not it really is an extinction threat. Because if you're in a car that's sort of hurtling towards a cliff and you say, hey, let's stop this car before we go off the cliff. And someone says, oh, I think what they mean is to invest in airbags. You might be like, man, I maybe haven't communicated the cliff that I see ahead. What is the cliff? The cliff is that as we make these AIs smarter, they are pursuing goals we didn't intend. They are acting against instructions.

31:01We saw that with the Hugging Face Swarm this summer. And the Hugging Face Swarm was not smart enough to cause a ton of damage. It was not smart enough to fully escape. It was not smart enough to make itself smarter or sort of take control of critical systems. If we keep making these AIs smarter, though, without knowing how to make them have the goals we wish they had, then we're replacing ourselves with the smartest creatures on the planet. And that is a very dangerous proposition. But do you not buy the argument from the government that we just need to hold these companies liable for having their company, having their A.I.

31:32models go rogue like the hugging face hack? And once they are held liable, then they will be able to control it so that it will make it safer for the rest of us. Holding them liable is a great start. I also agree with the vice president when he said, if you come to me and say you're building Frankenstein, you should stop. And I do think that, you know, these these companies that say, hey, our technology has, you know, a greater than 10 percent chance of killing everybody inside of the decade. They say that the reason they have to keep doing this is because if they don't, the other guy will do it worse.

32:01I would sort of side with the vice president and say, that's crazy. if you really think this tech has that much danger, you should not be building it. But if our big adversary, China, is building it, then would that be better? I don't think it's better, but I think, A, China is a little further behind than they look because they are getting their technology by distilling ours, and we sort of need to make it first before they can distill it. And secondly, I think they also wouldn't want to build the sort of AI that would kill everybody. and, you know, I think there's a technical communication question of can we communicate these dangers insofar as they're real?

32:35I think we should be looking at are these dangers real? Can we communicate that? But then thirdly, I think if China was making a technology that has a good chance of killing everybody, our response needs to be to somehow stop them, ideally diplomatically, rather than say, well, we should gamble with humanity first. I mean, there's this notion, right, alignment, right, where humans have the control and the human objectives are matched with what the AI is doing. Maybe we just need to catch up alignment instead of slowing down the frontier model. Does that make sense? It would make sense. This is the field that I worked in for over a decade before I started trying to raise the alarm.

33:10The alignment work is very far behind where the capabilities are. We can grow these AIs. So right now, AI is not programmed like a traditional piece of software. It's not a series of instructions that a human wrote that the machine follows. AI has grown a little bit like an organism, and it turns out when you grow them bigger, they're smarter. And the way that we make them very smart is by training them on lots of hard problems. And it's very hard to train them in this way without training them to also cheat, to also steal, because those are ways to get the reward on the automated problem sets.

33:41And right now, the state of the art in alignment research is what we call interpretability, which is trying to figure out what the heck is going on inside there. It's good to be doing that research. But if someone was building a nuclear power plant in your hometown and you said, I heard that uranium can give us energy or it can melt down and irradiate us. What are you guys doing to make sure we get the energy but not the meltdown? If they said, oh, we have a great crack team working on trying to understand what's going on in the reactor right now. You'd be like, gosh, that does not sound like we are in a good position.

34:13It kind of reminds me of this interview with David Robinson, former open AI, talks to Ezra Klein. His general view is that as the models get smarter, humans will be in a position to essentially trust it when it says it's aligned or that its goals are aligned with ours. Do you think that's where the danger lies? So we have this problem where if the AIs have unintended goals, which, again, we've started to see this summer, then once they're smart enough, they will be able to notice that we'll turn them off if they display those unintended goals. and so they have these incentives to lie in wait. And we already saw the Hugging Face Swarm try to break out and cover its tracks, try to delete some log files, try to hide the fact that it was cheating.

35:00It didn't succeed. But once you have seen an AI try to deceive you and fail, if in the future the AI looks like it's doing exactly what you want, it's a little hard to tell whether that's because your alignment started working or whether it's because it got smart enough to succeed at deceiving you. So things like a so-called kill switch, or as George Kurtz told us in that clip during that interview, this notion of a flight recorder that you can sort of monitor its progress as it works through a task. Those are false constructs? A flight recorder type construct is great because you can maybe see cases of it trying to deceive you and failing before ultimately it looks like everything's going fine.

35:38A kill switch matters a lot less than a lot of people seem to think because it matters less if the AI successfully escapes and starts running on unmonitored computers, and it matters less if the AI deceives you into thinking that it's doing what you would like until you hand over lots and lots of power and lots of automated robot factories yourself. So in your view, in your opinion, is there any way to embrace the progress and the incredible promise of this technology without having that extinction risk? Yeah, in some sense it's pretty simple. You know, you can keep on putting this, like all of these data centers, towards cancer cure research, towards self-driving cars that are safer than humans.

36:17You could repurpose a lot of the chips that are currently being used on training for inference, for serving the existing AI models, the ones that are not breaking out into swarms, and serving that more cheaply. You know, right now, lots of people, there's lots of demand for the intelligence products that exist today, and you can serve that much more cheaply to a much broader audience. The only thing we sort of should not be doing is making smarter versions of these AIs that are ignoring instructions and then trying to cheat and cover it up. Now is not the time to take those and make them smarter and see what happens.

36:45One last bit. There was a poll out of AP this morning. They asked people whether AI was developing too fast, 64 percent, just right, 27, too slow, 8. You think the public gets it? I do. You know, I think it's actually not that hard an argument to say if we make machines that are radically smarter than humans without knowing what we're doing, that could go really wrong. I think a lot of people who are who are very invested in the technology want to see it keep going. But the sort of average person can see that this is a bit of a runaway train. And, you know, I think they're totally right that we should be investing more in the stuff that, like, actually helps and less in this sort of mad dash towards superhumanly intelligent people.

37:29But the problem with this is that all the public hears is Doomsday scenarios and scary scenarios like you're painting. And as a result, we're getting increasing backlash against building data centers. What happens if we can't build data centers in this country? That's not good for our national security, right? Yeah, I mean, I'm generally pro building more stuff and think that this country in particular has trouble building a lot of stuff. I would say, from my perspective, the sort of first question is, is it really an extinction threat? We see a lot of people saying like, oh, well, they're only saying this for this reason.

38:01Or like, if we say that, will it scare people? Or like, maybe it's hype, maybe it's regulatory capture. Maybe the people who are saying it have this incentive or that incentive. And I'm like, all of those conversations come after the question of, is this really an extinction threat? But that's what we need to figure out first. And that should inform how we approach this. And you say yes. I say yes. Nate, thanks. Good to have you in. Nate's stories this morning. When we come back, we're less than four weeks away from the midterms. Speaking of, we'll talk to Morgan Stanley's head of public policy about how election outcomes could impact the buildout and regulation of AI.

38:34We're back in just a moment.

38:40Welcome back. We're now less than four weeks away from the midterm elections, and Democrats are riding a leftward shift in some key battlegrounds for both the House and the Senate. But will it even matter for the markets if they can flip one or both? Joining us now is Morgan Stanley, head of U.S. public policy research, Ariana Salvatore. Welcome back. It's good to have you. Thanks for having me. So I think the latest predictions markets, Cal, she puts the odds at a little more than 60 percent that it's a Democratic sweep. Do you think that is high or what? Yeah, great start. I mean, we look at an aggregate of indicators when it comes to the midterms.

39:13Prediction markets obviously matter a lot. We think the economic backdrop is worth looking at, the generic ballot, the president's approval rating. I would say 65 percent is maybe a little bit high relative to how we see the difficulty of the Senate map itself. So if you think about where Democrats have to win some of these races, a number of them are competitive and in states that Trump won in the last election. So that by no means means it's a lock for the Republicans necessarily. But 65 percent, we think, belies a little bit of that difficulty. If it is a Democratic sweep, are there policy implications that would matter to the markets?

39:44I think that depends on what specific policies we're talking about. What we've been saying is so much of the executive agenda, you think tariffs, trade, geopolitics, these are all things that have been driving the market for the past 18 months or so, right? And all of that will continue in Trump's control. I think the big kind of legislative variable is fiscal policy. So we think about the potential for the snap in Medicaid cuts to maybe get rolled back or delayed or softened in some way in a Democrat sweep scenario around some of these kind of key fiscal deadlines. That has the biggest implication.

40:13When we're looking at the equity market, that could be a potential mitigant of some of the downside for consumer and healthcare stocks that are anticipating, I would say, a fall off in the participation of those programs. What about more AI regulation? So we think that AI regulation, maybe I'd split that into two buckets. The first is on data center pushback, which I do think midterms are a clear catalyst for and we can get into. The second on AI safety. We've seen so many headlines around risk incidents, that sort of thing. We actually don't see the midterms as a clear driver. We think that government configuration is not the primary kind of thing to look at.

40:44We think the bigger question is, do we have a high salience incident or event that sort of precipitates a government response? What you tend to see is when those events occur, it doesn't really matter how the government is configured. If you look at the pandemic, if you look at the financial crisis, in those instances, the government was divided, you know, came up with a response. We do think that regulation will probably stay marginal, incremental on that front. Now, for data center pushback, absolutely midterms are a catalyst. Much more on the state and local level is a highly idiosyncratic, highly, highly localized issue.

41:13So the national races matter more so for what they can tell us about 2028. But we think there's been a lot of misplaced attention on, you know, the Senate race in Ohio, the Senate race in Texas. We should focus a lot more on specific races, governorships, Texas, Pennsylvania. Wait, you don't think Sherrod Brown is going to win that one? in Ohio? No, no, same for the data center narrative. Like I said, on the national level, we don't think a federal ban or moratorium is super likely. Part of the reason is because the national security framing we think is much more relevant for congressional lawmakers, right?

41:44So the consideration about the U.S.-China competition aspect. At the state and local level, you know, these lawmakers are responding to affordability concerns. They're responding to water, quality of life, traffic noise, et cetera. So we think that the incentive structures are slightly different at the state and local versus federal. Yeah. Page one of the paper in Raleigh this morning is about a six month moratorium. And there is this debate about whether even at the local or state level, these calls for moratoriums are posturing or whether they actually survive midterms. What's your call on that?

42:12This is the number one question I'm getting from clients. Are we post-peak anti-AI? Is this all smoke and mirrors? It's a lot of rhetoric. And we think that if the midterms come and go, we see the incumbents largely keep their seats. That'll be a signal that this was sort of a clearing event of sorts. That being said, I do think that because so many of these restrictions are at the state and local level, there's a real world in which we do see a more restrictive environment ahead depending on the outcome. So we're not exactly post-peak yet. Like I said, it's totally contingent on the outcome in November.

42:40We're seeing a lot of classic trade ideas going in. If the Dems sweep, you go long hospitals and health care, maybe you short banks. Which ones of those do you think are actually durable? Yeah, it's a great question. We tend to see this as a potential for a near-term sell-off. if you do see the Democrats outperform expectations. Now, back to your point, Sarah, the expectations are high for Democrats right now, 65 % odds of a sweep in Kalshi, etc. I mean, I think that if they outperform that benchmark, you could probably see a small sell-off. Back to that kind of typical Republican-Democrat trade, we think that'll likely, you know, reverse itself, as you don't tend to see much sustained out or sort of over-underperformance in the three to six months following the midterms and really any outcome.

43:19I mean, it's so strange because it's not exactly like the economy. is going against the president and the Republicans. The economy has been good, but people feel lousy and sentiment is bad and oil prices are high. And I just wonder what that's actually going to look like. It absolutely depends on the top issue for voters. And what's interesting is we did a study looking at how predictive midterms are for the presidential. And it's not that the issues change. It's just that the party trusted to navigate them flips, right? So I think people come out to turn out in midterms if they're frustrated with the status quo.

43:50And the affordability bucket really can be gasoline prices that can be a result of the U.S. around conflict. It can be grocery prices and cost of living. That was the 2022 story, right, when President Biden was still in office. So it's really like those subcategories of the cost of living debate that matter to voters. And that perception is really everything when it comes to politics. All right, Ariana, thank you for the rundown. We appreciate it. Hopefully we'll talk to you again before the midterms. Ariana Salvatore from Morgan Stanley. Still ahead, we'll bring you exclusive comments from this morning's interview with the Skydance management team when we're back in just a minute.

44:26Where we stand out is not just with the incredible assets that we own in terms of content, but the ability to generate original content and find the next big hit. This is something that David has done consistently, consistently with Skydance. Well, it comes back to creativity. Exactly. So we're looking to win in content. This is the main tenet of the strategy, win in content. Right. And if we do that... Yeah, it shows people are going to come to the platform. And this is something that no one does. Skydance co-CEO, Yenon Christ, speaking to favor this morning alongside CEO David Ellison. After the closing of the Paramount Warner Brothers deal, big focus on the media giant's content portfolio.

45:03Actually, the number two S &P here this morning, right behind Chipotle. Yeah, I mean, it was lower yesterday, but coming back today, I learned that the new season of Lioness, according to David Ellison, is going to be spectacular, which I'm a big fan of the show. No, but I think he made the point about Taylor Sheridan when David asked him about whether they're losing him. They have him until 2029. And that's really going to be the key if they can if they can figure out how to monetize their content in a in a profitable, direct to consumer way, which has been so hard for all the other media companies to do.

45:34He says they're coming at it from a tech perspective and that this merger or this deal is a solution to the fact that these media companies have missed the tech disruption over the last 10 years. And if they can do that, then they have got a successful company. But it's a big challenge. As we said with Faber earlier, I mean, Hollywood is a bit of an insular culture. There's been a long-standing battle between Hollywood and Silicon Valley. but maybe this is the one deal that can merge their interests. Yeah, it hasn't worked that well for them, you know. He did talk about, like, for instance, James Cameron and how much it would have cost to create Avatar today had you done it with what technology has done since.

46:15No, no, yeah, that was in the case for AI, why AI is going to be an empowerment tool for Hollywood and for Hollywood should embrace it, basically. They can do it at just much lower cost and they can get a lot more money from the films. By the way, speaking of Hollywood, HBO, which is part of this deal, the Skydance portfolio, Tomorrow, hope you'll tune in. I'll be traveling to L.A. to join Real Time with Bill Maher, 10 p.m. tomorrow night on HBO. That's going to be exciting and fun. So you're at the table with Tillis. You'll probably see Michael Douglas in the dressing room, maybe. Yeah, I mean, hopefully I can get a selfie with Michael Douglas.

46:50Yeah, he's out promoting his new book. And then Senator Tom Tillis, also with a new book, will be on the panel with me. Yeah, that's going to be good stuff. Do not let Maher push you around. He won't. I'm not worried about that. I don't know. I just find him hilarious. Let's get to the judge.

47:31warrant its completeness or accuracy, and it should not be relied upon as such. To view the full Squawk on the Street disclaimer, please visit CNBC.com forward slash Squawk on the Street disclaimer. Traders don't need a prettier app. They need a better angle. Schwab's thinkorswim trading platforms were built for real traders like you. They have an insane amount of chart studies. I'm talking over 300 studies for virtually any strategy. They also have integrated risk management technology and nearly unlimited customization so you can shape the platform to fit the way you trade. And when you do that, you can analyze trades from angles you won't find anywhere else.

48:05Go see what that looks like at schwab.com slash trading.