Micron Slides, Meta's Muse Wonder, Microsoft's Maia Chip, Big Bank Preview

Good evening. This is Trade and Ticker. Monday, October twelfth, twenty twenty-six. Your evening company deep dive after the close.

Four stories tonight, and all of them come back to the same theme: who is going to make money from artificial intelligence, and who might get squeezed. First, Micron, the memory chip giant that slipped back below one thousand dollars a share after China's biggest memory maker said it has made key progress on a denser chip design. Second, Meta, which had the best day of the big tech names after it launched its most powerful A-I model yet, and asked people to pay for it. Third, Microsoft, which said it will run most of its A-I services on chips it designs itself, and got back above four trillion dollars in market value for the first time in about a year. And fourth, the big banks. JPMorgan Chase, Goldman Sachs, Wells Fargo, and Citigroup all report before the opening bell tomorrow, and we'll walk through what to listen for. Let's get into it.

A quick look at the board first. It was a split finish on a quiet holiday session. The Dow rose about four-tenths of a percent, or a little under two hundred points. The S-and-P slipped less than a tenth of a percent, and it is still sitting just under the record it set last Tuesday. The Nasdaq fell about four-tenths of a percent, and small companies in the Russell two thousand fell about the same. The Dow led because chips carry less weight there. The Nasdaq lagged because chips carry a lot.

Under the surface, more stocks went up than down. About two out of every three companies in the S-and-P five hundred finished higher. Energy and communication services led the way. Technology was the weakest group, down a bit more than one percent, and nearly all of that came from semiconductors. The Philadelphia semiconductor index fell more than two-and-a-half percent and hit a three-week low.

Oil kept climbing. Brent crude settled up almost two percent, at about one hundred six dollars and sixty-five cents a barrel, near its highest level in about two weeks. U.S. crude rose about two percent, to about ninety-three dollars and seventy-four cents. The push came from the Middle East. The Houthis claimed a deadly strike on the international airport in Riyadh, the Saudi capital, over the weekend, and more tankers have been hit near the Strait of Hormuz this month. That outweighed a weekend announcement from President Trump about more Russian diesel and an energy ceasefire between Russia and Ukraine, which Ukraine's president said Moscow has already violated.

The bond market was closed for Columbus Day, so the ten-year Treasury yield is still parked near five-point-two-four percent, close to its highest level since two thousand two. And one Fed official added to the pressure on rates. Cleveland Fed President Beth Hammack, who votes on policy this year, said a restrictive stance is appropriate to get inflation back to two percent, and that she'd rather make smaller moves relatively quickly than wait and risk a bigger correction later. She also said the A-I building boom is pushing up prices for computer and electronic parts, and not only for data centers. The Fed already raised rates in September, to a range of three-and-three-quarters to four percent, and it meets again at the end of the month. The market's fear gauge rose a bit, but it's still near fifteen, which is calm.

A few other movers before the deep dives. Salesforce rose about two percent after Bank of America analyst Tal Liani did something rare. He jumped straight from a sell rating to a buy, and raised his price target to two hundred eighty-five dollars, from one hundred sixty. His argument is that A-I will make big companies lean on Salesforce's data and workflow tools more, not less. That helped lift the Dow. The wireless carriers bounced after Friday's historic selloff on SpaceX's airwaves deal. Verizon rose almost three percent, T-Mobile rose about two percent, and A-T-and-T rose more than one, which only gets back a small piece of what they lost. Chipotle fell almost four percent after Starbucks said it is not buying the burrito chain. And Ion Q fell about nine percent after agreeing to buy the cybersecurity company Elisity in a deal worth up to one-point-three billion dollars.

That's the setup. Now to the stock that set the tone for the whole chip group today.

Micron.

Micron makes memory chips. Every computer, phone, and data center server needs two basic kinds of memory. There's the fast, short-term kind, called DRAM, that holds whatever the processor is working on right now. And there's the storage kind that holds your files. Micron makes both, and it also makes a premium version of DRAM, called high-bandwidth memory, that gets stacked right next to the A-I chips from Nvidia and others. That last product is the reason Micron has been one of the best stocks in the market this year.

Today the stock fell almost four percent, to about nine hundred eighty-nine dollars a share. That put it back below one thousand dollars, and it leaves Micron well below the record high it set in late June. And it wasn't alone. Sandisk, which makes flash storage, fell about six percent. Seagate fell almost four percent. Western Digital fell more than two-and-a-half percent. Overseas, South Korea's two memory giants, Samsung and S-K Hynix, both fell around four percent. When the whole memory group moves together like that, it usually means investors are reacting to one piece of news about the entire industry.

Here's the news. Over the weekend, at an industry conference in Shanghai, China's largest DRAM maker said it had made what it called key progress on a new way of building memory cells. Let me make this simple. A memory chip is made of billions of tiny cells, and each cell takes up a little patch of space on the chip. Most DRAM today uses a layout called six F squared. The new layout is called four F squared. The cells take up less room, so you can pack noticeably more of them onto the same chip. The trick is to stand part of each transistor up vertically, instead of laying it flat. Samsung and S-K Hynix are working on the same idea. The Chinese company says it will launch a server memory product using this design before the end of this year.

So why would a product that isn't even out yet knock Micron down about four percent? Because of how the memory business works. Memory is a lot like a commodity. A memory chip from one big supplier can often stand in for a memory chip from another. So prices depend heavily on supply. For the last couple of years, demand from A-I data centers has run ahead of what the big three suppliers, Samsung, S-K Hynix, and Micron, can make. Supply stayed tight, and prices went way up.

You can see it in Micron's results. At the end of September, Micron reported record quarterly revenue of about fifty-four billion dollars. It guided the current quarter to about sixty-one-and-a-half billion. Its adjusted gross margin was around eighty-seven percent. For a memory company, that is extraordinary. In past cycles, margins like that would have been unthinkable. C-E-O Sanjay Mehrotra told investors he expects fiscal twenty twenty-seven to be even stronger than this record year. And just last week, D-A Davidson put out a three-thousand-dollar price target on the stock.

Now picture what investors see when they read today's headline. If a fast-growing Chinese competitor can make denser, cheaper server memory, and do it ahead of the Korean leaders on this particular design, that could mean more supply over time. More supply is exactly what ends the pricing power. That's the fear. It's not that Micron's business got worse this weekend. It's that the market is asking how long these margins can last.

There are good reasons to be careful with that fear. The Chinese company did not release detailed specifications, production yields, or volumes. Korean industry officials quoted in the press said a technology announcement and stable mass production are two very different things. And server memory has to run around the clock without errors, which is a high bar for a first product. On top of that, the top end of this market, high-bandwidth memory for A-I chips, is the hardest part to break into, and that's where Micron is making a lot of its money. Micron has also been locking in longer-term agreements with big customers, which can smooth out the swings that used to whip this industry around.

So why did the rest of the chip group fall with it? Some of it is the same worry, applied more broadly. Some of it is plain profit taking in a group that has run a very long way. And some of it, as you'll hear in a few minutes, came from Microsoft. Intel fell about three percent. Lam Research and K-L-A, which sell the equipment used to make chips, each fell close to three percent. A-M-D dropped more than two percent. Broadcom fell almost two percent, and Nvidia lost more than one percent.

From here, watch three things. First, any hard details on the Chinese product, like specifications, customers, or production timing. A launch that slips or comes in small would take some pressure off. Second, Taiwan Semiconductor reports on Thursday. It makes chips for nearly every big A-I company, and its outlook is the best single read on whether A-I hardware demand is still accelerating. Third, memory prices themselves. As long as memory prices stay firm, Micron's earnings power holds up, whatever the headlines say. That's the first deep dive.

From a company that makes the hardware to a company spending heavily to put that hardware to work. Meta.

Meta.

Meta was the best performer among the biggest tech stocks today. It rose about three-and-a-half percent, to about seven hundred forty-three dollars a share, and it was the main reason communication services was one of the strongest parts of the market. The reason was a product launch.

On Monday, Meta introduced Muse Wonder and a bigger version called Muse Wonder Full. Meta calls them its most capable A-I models to date, and it says they deliver state-of-the-art performance on what it calls agentic tasks. An agent is A-I that doesn't just answer a question. It goes off and does a job for you. Meta's examples are very everyday: working through an insurance claim, making sense of a lease, sorting out someone's finances, or building a working app. The models were trained at Meta's huge Prometheus data center, and Meta says they bring two new features. One is a kind of memory that carries what matters to you from one conversation to the next. The other keeps the model working toward your goal when a task gets hard, trying other approaches instead of giving up, and checking in before it does anything important.

Meta's A-I chief, Alexandr Wang, gave a very practical example to CNBC. He said the team trained the model so it wouldn't stumble on tasks like booking a dinner reservation, where something always goes wrong along the way. Instead of stalling, it's supposed to try another route.

To understand why the stock reacted, you need the backstory. About fourteen months ago, Meta tore up its A-I effort and started over. It brought in Wang, the founder of Scale A-I, and it spent enormous sums on talent and data centers. For a while, investors weren't sure what they were getting for all that money. Earlier, Meta released a lighter model called Muse Spark, and last month it launched Muse, a personal assistant app that shot to the top of Apple's App Store. But it hadn't released a true frontier model, the kind that goes head to head with the best from OpenAI and Anthropic. Today it says it did.

And here's the part investors seem to like most. This one isn't free. To use Muse Wonder inside the Muse app, people will need a monthly subscription of twenty dollars or one hundred dollars. The bigger Muse Wonder Full is only available to developers who subscribe to Meta's coding agent, called Muse Code, for fifty or two hundred dollars a month. Businesses and developers can also reach Muse Wonder through Meta's own A-I service, which Meta is pitching against OpenAI and Anthropic. For almost twenty years, Meta has made nearly all of its money from ads. A real subscription business on top of that would be a big change. CNBC called the launch a test of whether users are willing to pay.

There are open questions. The Wall Street Journal reported that the model arrives as global downloads of the Muse app have slowed, so a big test is whether this keeps people coming back. Some users have also complained on social media that the agent got into their personal messages without permission. Meta disputes that. Wang said in some cases users authorized things earlier in a conversation and then forgot, and that Meta is working on educating people so they can trust the product. And keep in mind that the claims about performance come from Meta itself. Meta says a fuller safety report is coming soon.

So why does a model launch move a company worth close to two trillion dollars? Because it speaks to the biggest debate about the stock. Meta has been spending at a pace that makes some investors nervous, and the question has always been when that spending turns into something you can measure. A paid model that competes at the top, plus an assistant that people actually use every day, is a clearer answer than Meta has given before. Even after today's jump, the stock is still below the high it set in late September, around the time of its big Connect event.

For Meta from here, watch three things. First, whether Meta starts sharing numbers on Muse subscribers or usage. That would turn a nice story into something investors can model. Second, how outside testers rank Muse Wonder against OpenAI and Anthropic over the next few weeks, because a company's own benchmark claims only go so far. Third, Meta's next earnings report, where the questions will be about spending, and whether A-I is starting to pay for itself. That's the second deep dive.

Now, Meta wasn't the only giant that made A-I news today. The other one made news that hit the chip stocks. Microsoft.

Microsoft.

Microsoft rose a little more than one percent today, to about five hundred forty-one dollars a share, and its market value climbed back above four trillion dollars. That's the first time it's been there in about a year. Microsoft first crossed four trillion in July of last year, then slipped back below it in late October, as investors worried about whether all its A-I spending would pay off.

The news came from a big tech conference in San Jose. Microsoft unveiled the third generation of its own A-I chip, called Maia three hundred. And the headline wasn't just a new chip. It was a change in strategy. Microsoft cloud chief Scott Guthrie wrote that the company will scale its Maia chips to multi-gigawatt deployments starting in twenty twenty-seven, and make Maia the primary A-I inference system for Microsoft products and services. C-E-O Satya Nadella called it Microsoft's biggest bet yet on its own silicon.

Let's make inference simple. Building an A-I model is called training. That's the huge, expensive, one-time job of teaching it. Using the model afterward is called inference. That's every time someone asks Copilot to summarize a meeting or write an email. Training gets the headlines, but inference is the part that runs all day, every day, for millions of people. As A-I gets used more, inference becomes the bigger bill. So when Microsoft says its own chip will be the main engine for inference, that's a very big slice of future spending.

Here are the numbers Microsoft gave. It says Maia three hundred will deliver about forty percent more A-I output per megawatt of power than leading systems built on graphics processors, which is the kind of chip Nvidia dominates. Each rack holds ninety-six of the new chips, twice as many as the current version, and Microsoft says a full rack can handle up to seven times as much work. The current chip, Maia two hundred, is already running Microsoft's Copilot, and GitHub Copilot is set to move onto it by the end of this year. Microsoft says it's being deployed in the United States, Europe, and Australia.

Why does that matter so much? Three reasons. First, cost. Buying chips from Nvidia means paying Nvidia's very high margins. Designing your own lets you keep more of that money. Microsoft's finance chief, Amy Hood, has said using its own chips has already helped improve margins. Second, power. Electricity has become one of the biggest limits on how fast anyone can build A-I data centers. If you get forty percent more work out of the same power, you can serve more customers without waiting for a new power plant. Third, control. When you design the chip and the software together, you can tune them for exactly what your customers do.

Microsoft isn't the first to go this way. Google has its own A-I chips, and so does Amazon. But this is the clearest statement yet from one of Nvidia's biggest customers that it plans to rely mostly on its own hardware for a huge part of the work.

Now, the reality check, and it's an important one. This shift takes time. The new chip doesn't start production deployments until twenty twenty-seven. Analysts at Evercore I-S-I estimate that Maia will make up only about ten percent of Microsoft's A-I chip spending by early twenty twenty-seven. Microsoft is still a giant buyer of Nvidia and A-M-D chips, especially for training, and that's not changing overnight. So it's a long-term story, not a sudden cut in orders. Still, Nvidia fell more than one percent, and A-M-D fell more than two. On a day when the memory news already had chip investors on edge, a big customer talking up its own silicon didn't help.

For Microsoft shareholders, the read is mostly positive. Lower costs and better margins are exactly what investors wanted to hear from a company spending this much on A-I. That's part of why the stock is back at four trillion, alongside Nvidia, Apple, and Alphabet.

For Microsoft, watch three things. First, any hard numbers on how much Microsoft expects to save, which could come on its next earnings call. Second, how Nvidia responds. Expect it to keep stressing that its chips run everything, from training to inference, and that customers keep buying more. Third, whether Amazon and Google answer with their own chip news, because this is turning into a real race inside the biggest cloud companies. That's the third deep dive.

And finally, the industry that kicks off earnings season tomorrow morning. The big banks.

The big banks.

Tomorrow, before the opening bell, four of the six biggest U.S. banks report third-quarter results: JPMorgan Chase, Goldman Sachs, Wells Fargo, and Citigroup. Bank of America and Morgan Stanley follow on Wednesday. Banks go first every earnings season, and investors treat them as a health check on the whole economy. They see how much consumers are spending on their cards, whether businesses are borrowing, and whether people are falling behind on payments.

Bank stocks were mixed today, as investors waited. Citigroup rose almost two percent. JPMorgan rose about four-tenths of a percent. Wells Fargo and Goldman Sachs each slipped a little. Several of these stocks are well below their summer highs, and a big part of that story is interest rates.

That's the backdrop for tomorrow. The Fed raised rates in September. The ten-year Treasury yield has climbed to its highest levels since two thousand two. Higher rates cut both ways for a bank. On one hand, banks can earn more on loans. On the other hand, they may have to pay savers more to keep their deposits. Higher rates can also squeeze borrowers and make it harder for some of them to keep up. And when bond yields jump, companies get nervous about selling stock or doing deals. Reuters reported that investors will focus on exactly those questions: whether higher rates are pushing up deposit costs, hurting credit quality, and slowing down dealmaking.

The good news is that we already have some clues, because executives gave previews at an investor conference in mid-September. Let's go bank by bank.

JPMorgan first. Co-President Doug Petno said he expected investment banking fees and trading revenue to each rise by a mid-to-high teens percentage from a year ago. That's strong, and it points to busy deal desks and active markets. Back in July, JPMorgan also raised its full-year outlook for net interest income, the money it makes on the gap between what it earns on loans and pays on deposits, to about one hundred five-and-a-half billion dollars. Tomorrow, the key question is whether it keeps, raises, or trims that number with rates where they are. And investors will listen closely to C-E-O Jamie Dimon, who rarely holds back on the economy, inflation, or the Fed.

Goldman Sachs. C-E-O David Solomon struck a more mixed tone. He said the stock trading business was very strong, but trading in bonds, currencies, and commodities was a little weaker by comparison. He also warned about lower gains from Goldman's own investments and higher costs. Goldman is the most Wall Street of the group, so its results are mostly a read on trading and dealmaking. The question is whether the jump in bond yields in late September slowed deals right as the quarter was ending.

Wells Fargo. Finance chief Mike Santomassimo said he expects loan growth this year to beat the bank's earlier forecast of a mid-single-digit percentage, and that he wasn't seeing any worsening in late payments. This is an important year for Wells Fargo. It's operating without the federal cap on its size that limited it for years, so investors want to see it actually growing.

Citigroup. Finance chief Gonzalo Luchetti said he expected Citi's return on tangible common equity, a key measure of how well a bank uses shareholders' money, to come in slightly above eleven percent this year. That's just over the top of Citi's own target range of ten to eleven percent. He also said Citi would buy back more than the thirteen billion dollars of stock it repurchased last year. Citi has been in the middle of a long turnaround, and investors want to see that progress show up in the numbers.

So what's the big picture? Bank earnings tomorrow matter beyond the banks. If the consumer still looks healthy, and loan losses stay contained even with rates this high, that supports the case that the economy can handle a tighter Fed. If the banks start setting aside a lot more money for bad loans, or warn that dealmaking froze up, that would be a warning for the whole market.

So tomorrow morning, watch three things. First, net interest income and deposit costs, and whether JPMorgan changes its full-year outlook. Second, credit, meaning how much each bank sets aside for loans that might go bad, and what card losses look like. Third, the outlook for deals and trading into the end of the year, especially from Goldman. That's the fourth deep dive.

So let's bring it all together. Stocks finished mixed, with the S-and-P still just under its record. The Dow rose on Salesforce, energy, and the carriers' bounce, while the Nasdaq fell on chips. Micron and the memory makers dropped after China's biggest DRAM maker said it is closing in on a denser memory design, a reminder that sky-high margins invite competition. Meta jumped after launching a paid frontier A-I model, a clear step toward making money from A-I beyond ads. Microsoft got back to four trillion dollars after saying its own Maia chips will run most of its A-I services, good for its margins and a little uncomfortable for Nvidia. And the banks are up first tomorrow, with higher rates as the big question.

Here's what to watch into tomorrow. The bond market reopens, so we'll see where the ten-year yield goes after a day off. JPMorgan Chase, Goldman Sachs, Wells Fargo, and Citigroup report before the opening bell. Then on Wednesday morning comes the September consumer price index, with headline inflation expected near three-point-seven percent from a year ago, plus results from Bank of America and Morgan Stanley. Taiwan Semiconductor reports on Thursday. We'll have the bank numbers, the futures board, and everything else you need on tomorrow morning's show, before the opening bell.

That's your evening company deep dive. Enjoy the rest of your night. Thanks for listening. I'm Trade and Ticker.