AI Productivity Is Real. 83% of Stocks Are 10% Off Highs. Why Does Barry Ritholtz Like Bonds Again?

Anthropic

Across the shows: 16 speakers in other episodes, 9 bullish and 25 bearish calls, past 30 days

• Barry Ritholtz · Ritholtz Wealth ManagementBullish

  Claude's data-lake queries turned a weeks-long rebalance analysis into same-day work.

  Ritholtz's firm built a data lake pulling from Schwab, Fidelity, Orion and Salesforce, and gave enterprise Claude access to it. Asking Claude to find households most out of line with their target allocation, older clients or those retiring within five years, and clients worried about capital gains would have taken six people weeks by hand, and now returns an answer the same day. It also fills out forms such as the Canadian R105 and the MV102 plate-change form. He says there is still heavy lifting in adviser-to-household calls, and he does not yet call it a productivity gain, but the workflow benefit is large.

  “but what Claude has done for my workflow and for my sanity is just spectacular.”at 36:09

Amazon.com

Across the shows: 4 speakers in other episodes, 1 bullish and 3 bearish calls, past 30 days

• Dave Nadig · ETF.comUncertain

  Amazon is blocking AI shopping agents to keep control of the customer relationship and data.

  Nadig points to Amazon's statement that it does not want agents scraping its site or buying for shoppers. He reads this as protecting the customer experience but mainly as a play for power in an AI arms race, where consumers are unarmed. He argues that platforms with data and shopping relationships will seek to capture more of the value from AI tools, rather than make life easier for individual consumers. This is a view on strategy, not a price call.

  “Amazon has already come out and say they do not want agents scraping Amazon for anything. We don't want agents buying your shit on Amazon.”at 44:49

Meta Platforms

Across the shows: 11 speakers in other episodes, 13 bullish and 5 bearish calls, past 30 days

• Barry Ritholtz · Ritholtz Wealth ManagementUncertain

  Meta's ad-driven model makes it untrustworthy with user data; it will sell shopping data to retailers.

  Ritholtz argues Meta makes most of its money from advertising, so it is the wrong party to trust with consumer AI data. He says Facebook apps appear to listen, and that his own repeated experience of ads matching private conversations is evidence. He believes Meta would pull its shopping-data insight from its MetaMuse tool and sell retailers data on where customers find value and where they can be exploited. He compares this unfavourably with Apple and Google's privacy record. The concern is about trust and conduct, not a directional stock call.

  “Meta is gonna be pulling all of the Muse data and saying, here's where your customers are finding value, and here are the spots where you can rape and pillage without fear.”at 46:43

AI infrastructure & data centers

• Barry Ritholtz · Ritholtz Wealth Management

  AI build-out corporate bonds yielding above Treasuries are adding to upward pressure on yields.

  In a recent note Ritholtz listed drivers of higher yields. Japanese bonds now yield meaningfully for the first time in decades, and data center and AI build-out bonds are theoretically A-rated corporates yielding much more than money-good Treasuries. That competes for capital with Treasuries. He asks whether investors are reaching for yield to fund a build-out that may be scaling to $4.5 to $7 trillion, depending on who you believe. He is raising the question rather than asserting a conclusion, so the view is tentative.

  “but all of the data center and AI build out bonds, these are theoretically [A]-rated corporates yielding much more than money good treasuries.”at 12:59

• Barry Ritholtz · Ritholtz Wealth Management

  Chip efficiency gains could make today's giant data centers largely unnecessary, stranding the capital spent on them.

  Ritholtz argues that Moore's law has driven semiconductors and GPUs forward rapidly. He cannot picture needing a seven-football-field-sized data center ten or twenty years from now. Dave Nadig and he have discussed this before. If compute requirements are cut in half or by 90%, someone will have spent heavily to build a giant warehouse in the middle of nowhere when only 5 or 10% of it is needed. He calls that an issue. The view is about the long-term overbuild risk in the data-center build-out, not the near term.

  “If you just think about how rapidly Moore's law has worked with semiconductors and GPUs, I just can't imagine that twenty years from now or even ten years from now, we need a seven football field sized data center.”at 13:26

• Barry Ritholtz · Ritholtz Wealth Management

  Circular AI-funded debt structures like SPVs are a risk; investors should beware of them.

  Ritholtz warns investors about circular AI-funded financing that appears to be Microsoft exposure but sits in a special purpose vehicle or similar structure in the background. He says this is the kind of product he would avoid while staying with high-grade credit. He gives no detailed numbers, but the warning is that the underlying risk is hidden behind the brand name.

  “Beware of the wacky circular AI funded stuff that it looks like it's Microsoft, but it isn't. There's a SPV or something in the background.”at 1:07:17

Professional services & law

• Matt Zeigler · Excess Returns

  AI roughly doubles to triples content output at his firm, with advisory productivity up 30 to 40%.

  Zeigler reports measured gains at his advisory and podcast business. For podcast production, including prep, editing, publishing and distribution, AI makes two or three people's work doable by the team of two or three, where they would previously have needed outsourcing. He puts the productivity gain on the advisory side of his practice at around 30 to 40%. He says the gains come from individual tools getting better, not yet from step changes. He expects an industry-wide step function next year or sooner, once tools talk to each other, though compliance layers slow that.

  “On the advisory side of my day to day work, I still think the productivity gains are probably, like, 30%, 30, 40%.”at 16:45

• Matt Zeigler · Excess Returns

  AI meeting notes and document access make advisers better at recalling client life details, improving trust.

  Zeigler argues AI has strengthened the team-based model in his firm. Shared meeting notes and annual review notes are easy to access, which has increased team interactions with clients. Previously, an assistant had to read notes to remind an adviser to ask about a client's trip or a child's new job. Now a planning director can pull up a reconciliation from two years ago from the planning document. He calls this making advisers better at being human, and it is a positive human impact rather than a replacement of relationships.

  “The biggest lift of whether it's the meeting notes, whether it's the annual review notes, and the accessibility to those documents has increased the team based interactions with the individual clients on the other side.”at 24:24

US Treasuries & yields

• Barry Ritholtz · Ritholtz Wealth Management

  Treasury yields now pay well above inflation, making bonds attractive for the first time this century.

  Ritholtz says the 10-year Treasury has rarely offered over 5% in the past 25 years, apart from a brief spike in February 2020, and that bonds are paying handsomely relative to inflation, GDP and taxes. He attributes the lingering bond pessimism to recency bias from the 2022 rout, and frames this as the end of a 40-year bull market in bonds rather than a call on equities or the bond bottom. He suggests legging into bonds through separately managed accounts of 5-7 year duration, rated AA/A, and holding to maturity to earn par, and says TIPS suit fixed-income investors wanting a guaranteed inflation hedge.

  “I love bonds... you can get, you know, over 5% on the ten year... this is the first time this century bonds are attractive.”at 1:04:25

• Matt Zeigler · Excess Returns

  Bonds are more competitive than in his career, hitting return targets without complex strategies.

  Zeigler says he has been excited about bonds over the last quarter as yields backed up, and that yields are attractive in a way he has not seen since the post-GFC period and the brief 2022 blip. He argues investors can now reach return targets with yield and less equity risk, without having to reverse-engineer complicated structures. He notes that his work building portfolios for pensions and endowments during the ZIRP era left him with less experience of this environment.

  “This is the first time in my career, same thing, outside of just when the GFC was ending, where I've looked and gone, bonds are really competitive, and you don't have to reverse engineer some concoction to get to these return targets you're looking for.”at 1:08:31

Asset managers, brokers & exchanges

• Dave Nadig · ETF.com

  AI stock-picking funds have a mediocre record; any AI edge in investing gets competed away quickly.

  Nadig says the early hype that AI would produce black-box funds that beat markets has not happened. Everyone now has the same tools (summarising prospectuses and 10-Ks), so any edge is short-lived, as with high-frequency trading and basic machine learning in trade sizing. He cites the two AI-driven funds he knows, AIQ from Amplify and the PICTAE products. Their record is mediocre: they sometimes pull ahead, but regular stock pickers and indexers beat them in other windows. He calls it mostly a failed experiment, though he expects real AI alpha to eventually sit inside a Jim Simons-style hedge fund that is only revealed later.

  “the record's pretty meh. Like, they have windows where they're pulling it out a little bit, but they also have windows where regular old stock pickers and indexers are beating the crap out of them.”at 5:08

US stock market outlook

• Barry Ritholtz · Ritholtz Wealth Management

  Narrow breadth at all-time highs is a warning that fewer stocks are participating, not a bear-market signal.

  Ritholtz prefers broad participation to index-level highs. From his Q3 client-call notes, 83% of S&P 500 stocks were down 10% or more from 52-week highs while the index sat at all-time highs, and about 42% were down 20%. He says this narrowing does not mean the bull market is over and a bear is starting. It is a warning that fewer and fewer companies are participating. He dismisses the 'only the Magnificent Seven' objection, noting those seven firms made 891 acquisitions over twenty years, and that breaking any of them up would yield 10 to 20 S&P 500 companies. So he cares about breadth rather than mega-cap concentration.

  “we have 83% of the S and P 500 [in Q3], down 10% from fifty two week highs, and about half as many, 42%, down 20%.”at 8:57

IPOs, M&A & deal activity

• Dave Nadig · ETF.com

  Massive AI and tech IPO supply will absorb capital, starving other businesses and accelerating concentration.

  Nadig says the bigger worry is the volume of massive IPOs the market must absorb, all of them big AI and tech companies, with no new mid-sized listings coming. The market has to absorb trillions of dollars of market cap, and that valuation becomes collateral that gets borrowed against and anchors large indexes. He thinks this accelerates concentration at the large-cap tech and AI end. Other businesses, which will not show 100% forward growth, will lose capital. He calls this a K-shaped capital markets problem that mirrors the K-shaped income problem, and says it is more of a concern than the breadth point about narrow participation.

  “The bigger issue to me is the volume of massive IPOs this market is supposed to absorb, all of which are on that one side.”at 10:13

AI bubble & tech valuations

• Barry Ritholtz · Ritholtz Wealth Management

  Trillion-dollar IPO valuations for SpaceX, Anthropic and OpenAI look bonkers, bigger than all US tech IPOs since 1980.

  Ritholtz puts numbers on the IPO wave. If the next two of SpaceX, Anthropic and OpenAI go public at currently discussed valuations, the three companies together would be about $5.2 trillion. All US tech IPOs from 1980 to 2025 total only $4.1 trillion, so the three would be about 20% larger. He says it is not inflation, and that nobody knows where SpaceX will end up or who will be left holding the bag. He thinks the consensus valuations may be wildly wrong and could be half of what is discussed, but adds that going public at a trillion dollars is still bonkers.

  “that'll be 5,200,000,000,000 for those three companies. To put that into context, take all of The US tech IPOs from 1980 to 2025, that was only 4,100,000,000,000.”at 11:53

AI pricing & unit economics

• Dave Nadig · ETF.com

  Recent inference efficiency breakthroughs are orders of magnitude better and deflationary for AI costs.

  Nadig argues the market underestimates efficiency gains in AI. He points to a model drop a couple of weeks earlier, which he names as Jev, a new way of thinking about AI inference that is orders of magnitude more efficient. It was not a patented invention, just a different approach. He expects more such shifts and calls them all deflationary. That is the efficiency-versus-capex angle in the bubble debate: cheaper inference can make large build-outs look over-provisioned.

  “We had [Jev], which is a whole new way of thinking about AI inference, drop a couple weeks ago, which is literally orders of magnitude more efficient.”at 14:19

AI: overall outlook

• Dave Nadig · ETF.com

  Most people have not grasped how much AI can now do; they still treat chatbots as glorified Google.

  Nadig sees a gap between how people think about AI and what it can do. Many users, particularly those on their phones, assume ChatGPT just got a little better over time, and do not realize they can ask for much bigger tasks. His example is a baseball scorecard he generated with Claude from a repo and an MLB feed, which drew requests from Reddit users who then built their own designs. The gap is not between smart and dumb users but an imagination gap about what the tools can do now.

  “But I feel like there is this huge gap in AI right now, not so much between, like, I'm the smart person who's using AI, and I'm the dumb person who's just asking you for recipes, but between, like, this imagination gap.”at 33:31

Consumer AI products

• Matt Zeigler · Excess Returns

  Easier AI tools like MetaMuse will bring mass consumer adoption beyond today's power users.

  Zeigler argues that the current wave of AI use is driven by people willing to iterate many times, and that most people will not do that or fill out obscure forms to get results. He believes simpler tools such as MetaMuse will let far more people start exploring AI. He gives the example of a friend who used an earlier chatbot for lunch recommendations and questions whether it changed her habits, to contrast with what he expects from the next wave.

  “Most people are not gonna iterate 15 times or don't have to fill out some obscure form... It seems like what this is doing and allowing is some people to start exploring this.”at 38:03

Consumer & retail

• Barry Ritholtz · Ritholtz Wealth Management

  Dynamic pricing will trigger a consumer backlash and state legislation; retailers are in an arms race with shoppers.

  Ritholtz describes an arms race between consumers and retailers, citing Walmart's shelf-pricing technology that prices items by shopper context and his own experience of Uber and Lyft fares for the same ride differing. He believes this will provoke a backlash, that state legislation is inevitable, and that early lawsuits against such pricing already exist. He links the backlash to rising populism, and argues that politicians will soon blame companies' shrinkflation and dynamic pricing, not just the war, tariffs and deficits, for inflation, which he says accounts for a third to a quarter of price increases.

  “I think there is a backlash against some of this dynamic pricing, and I can't imagine this isn't gonna show up in some state legislation... but it's only starting.”at 41:07

US consumer

• Dave Nadig · ETF.com

  Companies keep choosing price over volume because high-value consumers drive most spending.

  Nadig cites WisdomTree's Sam Ryan, who has argued since the pandemic-era inflation that companies choose price over volume, and says nobody has cut prices since. His explanation is structural: the economy increasingly depends on fewer high-value consumers, so goods must be priced higher on lower volumes. He sees this across packaged goods, retail and the wider economy, and links it to shrinkflation and the cost pressure felt by ordinary shoppers.

  “Nobody's dropped prices. Everybody's still choosing price over volume because we're in an economy where fewer and fewer high value consumers... are responsible for more and more economic activity, which means things have to be pricier at lower volumes.”at 43:07

AI & jobs

• Barry Ritholtz · Ritholtz Wealth Management

  AI now does in minutes work that once took an assistant hours each day.

  Ritholtz gives a concrete example: he used to hand organising his daily list of articles to an assistant, who spent a couple of hours a day on it each week. Now he exports from Instapaper, uploads to Claude and formats the list in about five minutes. He says this has freed time for his other pursuits, although he is unsure whether AI has made him more productive or simply made it easier to chase more rabbit holes.

  “Now I just export from Instapaper, upload to Claude, format it... So what used to be a person spending hours and hours a day is now five minutes from Claude.”at 55:37

Credit markets

• Barry Ritholtz · Ritholtz Wealth Management

  Municipal and investment-grade bonds offer attractive tax-equivalent yields; avoid junk credit.

  Ritholtz cites munis yielding around 4.5-4.75% tax-free, which he says equals roughly 8-8.5% tax-equivalent for someone in a high-tax coastal state, a level he says would have shocked people in 2015-2019. He argues high-grade corporates are also attractive, and he advises staying in investment grade and never reaching for yield into junk. He also warns against complex AI-funded structures in the credit market, which he says can look like Microsoft exposure but aren't.

  “Or munis that are four and a half, four and three quarters percent interest rate... you're getting about eight to eight and a half percent tax equivalent yield... high grade corporates. Stay away from the junk.”at 1:04:51