Shutter Island

‍

Michael Burry’s Soft Close

‍

Soundbite: Michael Burry of “The Big Short” fame closed his hedge fund Scion last week. My first thought was that, similar to past high-profile investors closing after incurring heavy losses, now is the time for the trade to finally start working. However, that is not the case with Burry. He returned outside investor money but kept his short technology bet against Palantir and the Data Center/Graphical Processing Unit (GPU) theme. Burry has a history of being early, but we think he will be proven right. The AI-investment idea can crack under a collateral crisis. There is an intrinsic maturity mismatch between the long duration of the investment requiring financing and the rapidly depreciating assets backing these loans. This creates a refinancing risk because the chips will be replaced each time upgraded technology is available. A market correction that devalues the GPU collateral will precipitate a credit crisis. We just do not know when sentiment turns, proving Burry’s thesis. We will wait for Pave’s fixed income rankings to rise to the top of our asset scores before sounding the alarm. Although this has not yet occurred, it does not mean it will not happen.

Illogical markets have no expiration date.

Many value hedge fund managers deregistered their funds in 1999 and 2000, returning money to their investors because the markets “did not make sense to them anymore”. Michael Burry shared the exact sentiment last week.

Trading is a game with rules. We believe that Burry is ignoring, as he did in 2005, an important principle of trading: you cannot tell a market when it should fall. We prefer to wait for the market to tell us when it is ready, even though we agree 100% with Michael’s outlook.

We liken the situation to taking out a 30-year mortgage on a house that will fall apart within two years and will require extensive renovation, followed by refinancing.

Cash-rich hyperscalers are hitting a ceiling from directing 50% of their free cash flow on a single use (data centers). That limitation forces them to look for funding by creating new financing vehicles Special Purpose Vehicles or SPVs) that keeps the mountain of debt off their balance sheet. There was an accounting rule change years ago, allowing firms to extend the depreciation schedule to 10 years for the semiconductors used for data storage. However, the same extended depreciation rule is applied today to new GPU-powered data centers, generating an asset-liability mismatch. Companies are using short-lifespan NVIDIA GPUs as collateral for long-term loans because the chips become obsolete as newer technology becomes available. These chips will be replaced frequently before the loan matures.

As an example of this financial engineering, CoreWeave announced earnings of $3.4 billion, or $2.2 billion after interest expense. They have $20 billion in GPUs, which could have a usable life as short as 18 months, but the company assumed they last a full 10 years, following the accounting rule. That leaves them with $200 million in profits after depreciation, which would have been a loss of $11 billion if the depreciation matched the lifespan of their GPUs.

Sentiment will turn down hard when a credit crisis becomes a reasonable possibility. The first stage, the collateral crisis, has not presented itself, so banks remain willing lenders. They are following Citibank CEO Chuck Prince’s 2007 playbook, “As long as the music is playing, you’ve got to get up and dance.” Now we have JP Morgan’s Jamie Dimon seeing cockroaches on the dance floor. Bulls may take solace that the NASDAQ rallied 10% after Prince’s infamous clarion call to ignore risk before it hit a multi-year high, but there is no denying the risk/return landscape is vulnerable.

‍

Behind the Data Center Curtain

‍

Soundbite: There is a line from the movie Shutter Island that reminds us of our current situation. When a police officer asked if anything unusual occurred, the nurse replied, “This is a mental institution for the criminally insane…usual isn’t a big part of our day.” Investors find themselves grappling with concepts and capex that exceed the bounds of normal comprehension. Stress is beginning to show. Over the last two weeks, we experienced four days when the S&P 500 moved one hundred points in a single session, versus only once in the past quarter. The questions linger: Are we facing a 2008 problem or a 2000 problem? The Center for Public Enterprise (CPE) published a report called “Bubble or Nothing” discussing circular deals pushing AI-related companies’ valuations higher, which rolls many different bubbles into one. Soaring optimism is locking investors into overweight long equity positions that overlook the specter of data centers combining a speculative real estate and energy infrastructure boom backed by loose private credit creation.

Many were shocked to hear that Softbank’s Masayoshi Son raised $6 billion by selling all his NVIDIA stock, but one cannot forget that there were investors on the other side of the transaction, happily buying his shares. The problem for those buyers is the very high-cost base they now carry. The incredible long-term prospects for AI technology blind investors to problems detailed in the CPE report out last week:

· Cash flow uncertainty from economics that do not reduce costs when data centers scale, and those costs will not be quickly recovered due to incredible competition restraining pricing power.

· The collateral value of GPUs can drop soon, not only from uncertain near-term demand (see third Point below) and cloudy supply dynamics of GPUs that may see new releases every eighteen months.

· Data center developers are facing potentially rapid tenant turnover as their tech company tenants encounter multiple-stage capital outlays within a lease term.

· A small number of those tenants comprise the bulk of the entire market and are financing one another’s expansion, “creating an interlocking liability structure across the sector representing concentration risks for lenders and shareholders.”

The CPE report warns that any pullback in demand can make the data centers “look like stranded assets,” and problems will arise as widening credit spreads create problems in rolling over the asset-backed debt that funded the data center expansion. They detail a potential crisis where hyperscalers are forced to sell high-end GPUs to meet their liabilities, a problem that is magnified if there is no demand for used GPUs. This would slow purchases of new GPUs, leading to an “industry-wide, or even economy-wide, de-leveraging.” This we would describe as a margin call for the hyperscalers.

They reference the work of economist Hyman Minsky, who warned of margin calls from levered tech stock investors who are in danger because a failure of a “even one weak firm, will also hurt the equity valuations of stronger firms as investors sell shares of the latter—including leveraged shares—to cover their losses from the former. The result is a self-sustaining downward spiral.”

The CPE lists the same dangers we outlined by describing the potential for a collateral crisis that precedes a credit crisis. There is no hard evidence of problems outside of Tricolor and First Brands, but as evidence mounts, so will equity selling.

‍

McKinsey AI State of the Union: Early Days and Demand Wobble

‍

Soundbite: McKinsey Consulting published “The State of AI in 2025,” reporting an 88 percent adoption rate, up ten percent from last year. Unfortunately, only one-third of firms surveyed are even beginning to scale their AI initiatives, with most stuck “in the experimenting or piloting stages.” This presents enormous challenges for investors trying to accurately forecast future demand that will generate the necessary revenue to cover their interest expense borne by the datacenter capex outlays.

McKinsey endeavored to survey nearly 2,000 corporations worldwide, questioning employees at all levels during June and July of this year. The good news is that McKinsey is reporting that over half of the firms surveyed expect no layoffs, with some forecasting net increases in their payrolls over the next year as they roll out AI. A solid 62% of all organizations are experimenting with AI agents, and 23% are already scaling agentic AI systems. The most active sectors conducting those activities are technology, healthcare and insurance, and media and telecom. While 64% express that AI is helping innovate, only 39% of firms are enjoying EBIT gains. However, that is in sharp contrast to the MIT report that revealed 95% of companies extract no measurable profit impact. Both McKinsey and MIT found AI system failures: McKinsey’s research uncovered that 51% have encountered problems, mainly from inaccurate output.

The consulting firm has been conducting AI research for eight years, and IT, marketing, and sales have been the most familiar target areas for AI. However, this survey shows knowledge management is now a focus; enterprises are using AI to “capture information as well as processing and delivering it, such as through a conversational interface; in content support for marketing strategy, including drafting, generating ideas, and presenting knowledge for creating marketing strategies; and in contact-center or customer service automation.”

McKinsey found that the firms that see the most value from AI not only set efficiency as an objective but also set growth, cost, or innovation goals through redesigning workflows. In other words, they are using AI tools to transform and redefine how work is completed. Instead of measuring efficiency, the most advanced firms measure how fast their agentic systems act.

Unsurprisingly, companies with revenues north of $1 billion are those that have surpassed the scaling threshold, with 10% of large firms having reached the fully scaled phase in certain projects. Over the past twelve months, there have been only single-digit cost decreases reported across all business functions from AI use.

As of 2025, it is the rare company that has achieved significant enterprise-wide improvement to the bottom-line. The fact that there are leading firms that have seen success is a source of optimism, but this is a glass half-full situation. Current valuations expect that all firms will rise to match the success of the leaders, but just because AI adoption is broad, it is not transformative, at least yet.

‍

What to Look for This Week

(All times E.S.T.)

‍

1. The Big Event: Wednesday, November 19 after the market closes, NVIDIA presents its earnings. This could be the pivotal event of the week that sets the tone for the remainder of the month. We will be watching the tone of the market immediately after the numbers, and of course after the earnings call.

2. A second non-economic release of import Wednesday November 19: at 2:00 p.m. The FOMC Minutes from the October 28-29 meeting. There was obvious contention between the two camps, those who forecast two rate cuts after the September meeting and the other half of the Committee who wanted to keep rates unchanged for the remainder of the year. The voters decided to show consensus against Governor Stephen Miran who wants 50-basis point cuts at each meeting. There was only one dissent to having kept rates

unchanged at the October meeting, but it is obvious there were more, and we will get a sense of that from the minutes. The market was over 90% confident in a cut before that meeting and has now flipped to 55% for no cut. We will see how that changes after the minutes are released.

3. Thursday, November 20, two regional Fed indicators: at 9:30 a.m. Philadelphia Fed November Manufacturing Index and Business Conditions Index, and at noon, the Kansas City Fed Manufacturing Index. For Philadelphia, the Manufacturing Index fell sharply in October but Business Conditions have risen steadily for the last four months. It is worth looking at the Employment subcomponent which has fallen for the last three months. The Kansas City Fed Manufacturing Index rose to a three-and-a-half-year high in October, and the Composite Index flipped positive over the past quarter.

FOMC Voters Speaking: Again, from above, the key Fed news will be the minutes on Wednesday, November 19 at 2:00 p.m. Monday November 17, three speakers: NY Fed President Williams at 9:00 a.m., Governor Jefferson at 9:30 a.m., and who we are keen to hear, Governor Waller at 3:35 p.m. Tuesday November 18 at 10:30 a.m. Governor Barr, and Wednesday, November 19 at 10:00 a.m. Governor Miran, and at 2:00 p.m. President Williams speaks again, as the Minutes are released. Governor Cook speaks Thursday November 20 at 11:00 a.m. and at 12:40 Chicago Fed President Goolsbee speaks, with Miran once again promoting rate cuts at 6:15 p.m. Friday November 21 President Willams completes his hat trick at 7:30 a.m., followed by Governor Barr at 8:30 a.m. and Governor Jefferson at 8:45 a.m.

Earnings: Tuesday, November 18 before market open, Home Depot (HD) and Baidu (BIDU) report. As mentioned above, Wednesday, November 19 after market is the highly anticipated NVIDIA (NVDA) earnings results and conference call. Palo Alto Networks (PANW) also reports and is worth observing. Before market, we get a better understanding of consumer demand when Lowes (LOW) and Target (TGT) report. Remember that Target has not been hiring for the holidays. That picture fills out even more when Walmart (WMT) reports before market on Thursday November 20.

By Peter Corey

‍

Risk Disclosure & Disclaimer

Copyright © 2025 Pave Finance, Inc., LLC and Peter Corey. All rights reserved. 

The attached material is provided in partnership with Pave Investment Advisers, LLC (“Pave”), and is the opinion of Pave and Pave’s Chief Markets Strategist, Peter Corey. Distribution is being provided through a licensing agreement, and any further dissemination is strictly prohibited. 
You are receiving this email because you opted in through our product or website

You can update your preferences or unsubscribe from this list.

For Financial Professional Use Only

The content is developed from sources believed to be providing accurate information. The information in this material is not intended as tax or legal advice. Please consult legal or tax professionals for specific information regarding your individual situation. The opinions expressed and material provided are for general information and should not be considered a solicitation for the purchase or sale of any security. There is no guarantee any investment recommendations will be successful and there is a risk of losing money.

We take protecting your data and privacy very seriously. Please see our privacy policy here.

Pave Investment Advisors, LLC is an SEC Registered Investment Advisor. Such registration does not imply any level of expertise.

Securities offered through Pave Securities, LLC – New York, NY. Member FINRA/SIPC. You can review Pave Securities LLC with FINRA’s BrokerCheck BrokerCheck.

Advisory Services are only offered through Pave Investment Advisors, LLC., an SEC Registered Investment Advisor which is an affiliate of Pave Securities, LLC.

Pave Securities, LLC and Pave Investment Advisors, LLC do not offer tax or legal advice.

The news, resources and articles available on this site are being presented for educational purposes only and should not be considered specific investment or planning advice applicable to each individual.

Please see ADV Part 2A here and Form CRS here.