Barclays Warns Debt Limits Threaten Trillion-Dollar AI Capex Boom
Bloomberg
Widening credit spreads at peripheral infrastructure providers show the artificial intelligence buildout faces rigid capital constraints that enterprise revenues cannot bridge.
Institutional equity investors remain overwhelmingly long the AI trade, but debt markets are displaying severe cracks as net corporate bond issuance competes against a projected $1.1 trillion pipeline alongside heavy Treasury supply. (5:00)
Peripheral 'neocloud' infrastructure providers face direct debt financing bottlenecks and contagion risks, contrasting with core hyperscalers whose legacy cash-volcano businesses self-fund heavy capex. (7:10)
Projected forward hyperscaler capital expenditures are tracking at roughly 12% of GDP, with approximately 75% of next year's S&P 500 earnings growth concentrated in semiconductors and memory chips—effectively leading to the 'KOSPI-fication' of the index. (43:24)
Open-weight models have captured up to 80% market share while token costs have deflated to 1/100th of prior levels, pushing everyday inference toward local on-device silicon rather than centralized hyperscale data centers. (22:30)
Detailed notes
## Background
Published by Bloomberg on October 29, 2026, as an episode of the Odd Lots podcast, this program examines how Wall Street is processing recent wobbles and shifting sentiment around the AI trade following catalysts like Anthropic's S-1 filing and major hyperscaler earnings. The show notes highlight discussions with Alexander 'Alty' Altmann of Barclays on whether the laws of economics apply to AI, the 'KOSPI-fication' of the S&P 500, historical precedents for modern capex, and the sector's sensitivity to interest rates.
## TL;DR
Hosts Joe Weisenthal and Tracy Alloway sit down with Alexander Altmann, Global Head of Equities Tactical Strategies at Barclays, to evaluate growing cracks in the artificial intelligence trade. Altmann details how debt and credit markets are already pushing back against aggressive leverage in peripheral 'neoclouds' like CoreWeave, contrasting their vulnerabilities with self-funded hyperscalers possessing core cash-generating franchises. The conversation explores the tension between multi-trillion-dollar total addressable market projections and finite corporate profit pools, noting the deflationary collapse of token costs and the rise of local on-device inference. Altmann and the hosts also break down multi-manager hedge fund dynamics, the sudden breaking earnings from Microsoft and Alphabet, and the 'KOSPI-fication' of the S&P 500, where semiconductors drive an overwhelming share of earnings growth.
## Atmosphere & style
The episode carries an energetic, fast-paced Wall Street trading desk energy, punctuated by dry humor and authentic market bewilderment as the hosts confess to feeling like they have 'lost the plot.' The dynamic between Joe, Tracy, and Altmann is brisk, deeply technical, and conversational, blending sharp macro-credit analysis with casual Silicon Valley jargon and real-time reactions to breaking earnings prints. It feels like an insider look into institutional anxiety, balancing the dizzying scale of tech capex against the grounded, nervous reality of corporate bond markets and multi-manager pod shop positioning.
## Bulletin
- `05:00` Institutional equity investors remain overwhelmingly long the AI trade, but debt markets are displaying severe cracks as net corporate bond issuance competes against a projected $1.1 trillion pipeline alongside heavy Treasury supply.
- `07:10` Peripheral 'neocloud' infrastructure providers face direct debt financing bottlenecks and contagion risks, contrasting with core hyperscalers whose legacy cash-volcano businesses self-fund heavy capex.
- `43:24` Projected forward hyperscaler capital expenditures are tracking at roughly 12% of GDP, with approximately 75% of next year's S&P 500 earnings growth concentrated in semiconductors and memory chips—effectively leading to the 'KOSPI-fication' of the index.
- `22:30` Open-weight models have captured up to 80% market share while token costs have deflated to 1/100th of prior levels, pushing everyday inference toward local on-device silicon rather than centralized hyperscale data centers.
## Key takeaways
- `05:00` Institutional investor sentiment remains overwhelmingly bullish on AI, mirroring the denial phase of classic bubble cycles even as select Asian memory chipmakers trade down to roughly 3x P/E due to peak-cycle fears.
- `07:10` A structural fault line exists between core hyperscalers with massive operating cash flows and peripheral neoclouds that rely purely on debt markets to fund speculative capacity.
- `07:10` Cracks in peripheral debt financing, such as Cipher's dilutive convertible offering, have triggered contagion across neocloud credit, hitting CoreWeave bond pricing.
- `10:00` Over 60% of surveyed Barclays institutional clients anticipate systemic bottlenecks in AI financing as aggregate corporate debt issuance competes for finite global balance-sheet capacity.
- `12:35` The long-standing negative correlation between high benchmark interest rates and debt issuance has broken down, with corporate debt pipelines expanding aggressively despite elevated yields.
- `20:00` Aggressive TAM projections, such as SpaceX's pitched $30 trillion market, clash with corporate fundamentals given that the entire S&P 500 generates only $5 trillion to $5.5 trillion in net profit pool.
- `22:30` Token prices have experienced dramatic deflation while data center GPU rental rates remain elevated, creating a margin squeeze for intermediate infrastructure players.
- `25:00` The advancement of open-weight models optimized for consumer laptops and mobile hardware enables edge inference, reducing the necessity of centralized data centers for routine agentic tasks.
- `30:00` Only specialized sectors like biotechnology, chip design, and quantitative trading exhibit unlimited demand for frontier intelligence, leaving consumer adoption as the necessary driver to justify broader capex.
- `35:00` The primary catalyst capable of derailing current equity momentum would be an explicit downward revision in capex guidance by a major hyperscaler.
- `40:00` The explosion of multi-manager pod shops to over $1 trillion in AUM has compressed liquid equity trading into violent, factor-crowded churn beneath historically low headline index volatility.
- `40:00` The AI factor defies static quantitative hedging models like Barra or Axioma, because underlying component correlations and market narratives change continuously.
- `43:24` Hyperscaler forward capex commitments rival monumental historical infrastructure builds, transforming the S&P 500 into a hardware-concentrated benchmark akin to South Korea's KOSPI.
- `47:11` The widespread tech narrative that hyper-growth renders AI infrastructure immune to benchmark interest rates is fundamentally flawed for debt-dependent providers.
## Detailed walkthrough
### `00:00` Odd Lots Intro: Market Reality Check and the AI Dilemma
Tracy Allaway and Joe Weisenthal open the episode noting that they feel as if they have "lost the plot" with current financial markets, struggling to identify what the actual consensus trades are and what market participants are genuinely worried about—particularly around AI data center buildouts, leverage, and cloud providers like CoreWeave. Recording on October 28, the hosts note that the episode falls at a critical macro juncture: right after an FOMC decision, following the release of an S-1 filing from Anthropic, and amidst fresh financial guidance from Oracle unveiled during its Investor Day.
Joe points out that Oracle's stock surged 12% following its Investor Day announcements, arguing that corporate capital expenditure plans and the immense debt hyperscalers must take on have eclipsed the Federal Reserve's moves in the minds of investors. Tracy jokes that Federal Reserve press conferences have become an exercise in deflecting tough questions, making corporate AI updates far more consequential for market sentiment. To unpack what Wall Street genuinely thinks about the sustainability of the AI theme, they bring back Alexander "Alty" Altmann, Global Head of Equities Tactical Strategies at Barclays.
- **Recording Date:** October 28.
- **Key Corporate and Macro Catalysts:** The Anthropic S-1 registration filing, an FOMC rate decision, and Oracle Investor Day financial guidance driving its shares up 12%.
---
`05:00` The Consensus AI Outlook, Bubble Cycles, and Memory Stock Valuations
When asked if a clear consensus outlook on AI still exists, Alexander Altmann responds that investors remain overwhelmingly bullish: “The consensus AI outlook is that AI is still going to—this train is in motion, we're not stopping this train.” If forced to choose between being bullish or bearish, institutional cohorts across the US, Europe, and Asia almost unanimously choose to stay long. However, Altmann cautions that on classic market bubble cycle charts, the first downward roll following an initial peak represents the "denial" phase. While he does not claim that the AI bubble has formally popped, the prevailing environment mirrors that psychology, as virtually everyone stays staunchly optimistic regardless of emerging strains.
```mermaid
flowchart TD
accTitle: Classic Bubble Lifecycle and AI Sentiment
accDescr: A flow illustrating the market cycle progression highlighting current investor denial despite low semiconductor multiples.
subgraph Cycle ["Market Bubble Progression"]
A["Rapid Expansion & Optimism"] --> B["First Market Dip / Wobble"]
B --> C["Denial Phase: Overwhelming Bullish Consensus"]
C --> D["Valuation De-rating & Cyclical Peak"]
end
subgraph Market ["Market Reality (Semis)"]
E["Memory Chip Stocks (e.g., SK Hynix)"]
F["P/E Compresses to ~3x Earnings"]
E --- F
end
C -.->|"Juxtaposed against"| Market
```
As evidence of underlying skepticism, Altmann points out the disconnect in the semiconductor space, noting that several Asian memory chipmakers (such as SK Hynix) trade at roughly three times price-to-earnings (P/E). Joe expresses surprise that multiples have sunk that low. Altmann and Joe explain that this stems from a cardinal investing rule: never pay top dollar for cyclical companies at the absolute peak of their earnings cycle. The rock-bottom multiples show that public equity markets do not price in perpetual exponential growth for commodity memory hardware, contrasting sharply with the broader hype.
- **Survey Sentiment:** When polled, institutional investors across global regions overwhelmingly prefer remaining long the AI theme.
- **Valuation Anomaly:** Select memory suppliers trade as low as 3x P/E due to cyclicality fears and assumptions that peak earnings have already arrived.
---
`07:10` Neoclouds as the Periphery: Financing Cracks and Contagion Risks
Joe shifts the discussion to the emergence of "neoclouds" (such as CoreWeave) and why credit anxiety—manifested in spiking CDS spreads and volatile bond pricing—has begun to dominate conversations despite the bullish equity consensus. Altmann frames the dynamic by drawing a structural boundary across the AI ecosystem: core hyperscalers versus peripheral neoclouds. Hyperscalers like Alphabet and Amazon possess self-funding, high-margin, cash-generating core franchises—which Altmann describes as massive "cash volcano businesses." In contrast, neoclouds are mandated to build out speculative and pre-committed compute capacity with little to no existing operating cash flow.
Altmann emphasizes that neoclouds represent the tip of the spear and sit at the high-risk edge of AI debt financing. He compares this dynamic to the prelude of past financial crises: “The problem didn't start with G-SIB banks, right? It started with subprime vehicles 12, 18, 24 months beforehand.” He highlights that a clear financing bottleneck is emerging in this peripheral layer. For instance, when an alternative neocloud named Cipher attempted to tap liquidity via a convertible bond offering, the deal heavily penalized its own share price because direct corporate debt markets were effectively closed to it. That distress quickly spilled across the sector, causing CoreWeave's bonds to drop in sympathy even though CoreWeave itself had issued no negative news.
```mermaid
flowchart LR
accTitle: Core Hyperscalers vs Peripheral Neoclouds
accDescr: Contrast between self-funded hyperscalers and debt-dependent neoclouds vulnerable to contagion.
subgraph Core ["Core Hyperscalers"]
H1["Alphabet / Amazon"]
H2["Massive Operating Cash Flows ('Cash Volcanos')"]
H1 --- H2
end
subgraph Periphery ["Peripheral Neoclouds"]
N1["CoreWeave / Cipher"]
N2["Minimal Cash Flow / Pure Capex"]
N3["Debt & Convertible Market Vulnerability"]
N1 --- N2
N2 --- N3
end
N3 -.->|"Spillover contagion (CDS / Bonds)"| Core
```
- **Core vs. Periphery:** Hyperscalers rely on vast internal operational cash flow, whereas neoclouds rely almost purely on external debt and capital markets to acquire compute capacity.
- **Contagion Example:** Cipher's dilutive convertible debt offering led to a direct selloff in CoreWeave bonds, highlighting how fragile financing conditions at peripheral firms can ricochet across the entire neocloud subsector.
`10:00` The Myth of Unlimited AI Financing and Cracks in the Debt Market
A sudden shift in investor perception has begun exposing the vulnerability of the AI trade's funding pipeline. While equity markets have treated the artificial intelligence buildout as an unstoppable juggernaut, warning signs have been quietly flashing across debt markets for months. Bond spreads on specialized data center projects—including dedicated financing vehicles like those tied to Meta—had already been widening well before recent equity wobbles, alongside a noticeable spread creep across investment-grade hyperscalers and opportunistic players like Cipher Mining, which pivoted from Bitcoin mining into high-performance computing data centers.
```mermaid
flowchart TD
accTitle: The AI Debt Contagion Flow
accDescr: Visualizing how credit widening at the periphery spreads to primary AI borrowers and challenges assumptions of unlimited borrowing.
subgraph Periphery ["Periphery: Debt Market Cracks"]
A["Widening Spreads in Data Center Bonds"]
B["Spread Creep in IG Hyperscalers"]
C["Pivoted Crypto Miners (e.g., Cipher Mining)"]
end
subgraph Catalyst ["Market Catalyst & Hubris"]
D["CoreWeave Credit Sell-Off<br>(FT Interview on 'Unlimited Borrowing')"]
end
subgraph Reality ["Macro Reality Check"]
E["Competition for Capital<br>($1.1T Net Corporate Issuance)"]
F["Denial of Real Financing Limits"]
end
Periphery --> D
D --> E
E --> F
```
To gauge sentiment, Alexander Altmann’s tactical strategies team surveyed institutional clients on whether AI financing would become a systemic problem in the foreseeable future. Roughly 60% of respondents answered yes. Altmann contends that after recent market developments, that number should rationally be closer to 100%, and that the current market mood reflects an outright refusal to accept tightening credit realities.
> “So what's happening in AI to me is evidence of like almost a denial of a reality that's already here.” — Alexander Altmann
A prime catalyst for this reckoning occurred when CoreWeave bonds sold off sharply following a *Financial Times* interview in which co-founder Brannin McBee openly touted the prospect of seemingly infinite leverage. Investors had operated under the collective illusion that borrowers could tap the investment-grade debt and syndicated loan markets indefinitely. As Altmann observes from global meetings with institutional clients, markets had adopted the widespread dogma that AI capital expenditures were somehow impervious to the cost of money—a belief that has quickly unraveled.
- Specialized data center bond issues had been steadily widening for months prior to broader public equity sell-offs.
- In a Barclays investor survey, around 60% of respondents anticipated financing bottlenecks for the AI buildout, leaving roughly 40% still expecting frictionless capital availability.
- CoreWeave's credit sell-off was accelerated by executive commentary in the *Financial Times* suggesting access to virtually unlimited borrowing capacity.
---
`12:35` The Breakdown of Macro Norms and the Competition for Capital
The assumption that tech capex operates outside normal economic laws was tested further by disclosures from hyperscalers like Meta, whose filings and commentary highlighted clear financing limitations. During the summer, Altmann’s group analyzed holistic corporate and sovereign net debt issuance across investment-grade credit, high-yield bonds, U.S. Treasuries, and equity, mapping each segment's net issuance against its broader market capitalization dating back to 1995.
Historically, the data revealed an intuitive macro rule: lower interest rates spurred issuance, while high interest rates depressed debt supply. However, the data for 2025 and 2026 shattered these three-decade norms. As Tracy Alloway highlights, plotting net issuance against the two-year Treasury yield demonstrates that borrowers are pushing out record volumes of debt relative to prevailing benchmark rates.
> “And the reason why I bring it back to the sort of AI financing is because there is this competition for capital: it is real. AI is not issuing debt in a vacuum. There are other verticals that need to tap financing, and there is a finite amount, right?” — Alexander Altmann
Altmann underscores that investment-grade and high-yield net issuance is projected to reach approximately $1.1 trillion in a single year. While market participants casually toss around trillion-dollar estimates, capital pools are finite, and the sheer supply from the AI sector is aggressively crowding out and competing against other vital industries for balance-sheet capacity.
| Era / Metric | Benchmark Rates vs. Issuance Dynamics | Government Balance Sheet Context |
|:--- |:--- |:--- |
| **Historical Baseline (1995–2024)** | Inverse: Falling yields drive elevated debt supply; rising yields compress issuance. | Cyclical federal deficits with standard corporate funding patterns. |
| **Dot-Com Era (Late 1990s)** | High interest rates paired with massive equity issuance rather than pure debt supply. | U.S. federal government was running an active budget surplus. |
| **Current Market (2025–2026)** | Historical correlation shatters: Massive net debt issuance despite restrictive benchmark rates. | High Treasury issuance combined with an estimated $1.1T net corporate debt pipeline. |
---
`15:00` Intermission: Sponsor Messages and Bloomberg Tech Minute
During the mid-roll break, Carol Masser delivers the *Bloomberg Tech Minute* (sponsored by ChatGPT), spotlighting SpaceX’s rigorous hiring process as reported by *Bloomberg Businessweek*’s Lauren Grush. Landing a position at SpaceX requires proving raw, innate intelligence and deep hands-on accomplishments rather than relying on advanced academic credentials, with Elon Musk personally reviewing select candidates' resumes. Applicants must demonstrate unwavering loyalty to Musk's mission of space exploration and show willingness to work intense hours, including weekends, rewarded primarily through equity grants and vesting restricted stock units (RSUs).
Additional commercial messages highlight Cincinnati Insurance’s network of independent agents helping clients protect homes, businesses, and valuables, as well as the Delta SkyMiles Reserve Business American Express card offering enhanced bonus miles and Delta Sky Club lounge passes for business travelers.
---
`18:55` Reconciling End-User GPU Demand with Creditor Hesitation
As the conversation resumes, Joe Weisenthal challenges the growing pessimism in debt markets. He asks how to square the rising caution among credit investors with the rampant, seemingly unshakeable bullishness across equity markets and technology corridors.
If tech giants, startups, and sovereign buyers are desperate for compute capacity, why should creditors worry about niche infrastructure providers like CoreWeave, Iren, and Cipher Mining? Joe points out that the bull thesis insists someone will always step up to rent or buy these cutting-edge GPUs—provided operators have the requisite electrical power. The core debate turns on whether physical demand for processing power is enough to insulate infrastructure intermediaries from the real financial risks of asset depreciation, refinancing costs, and aggressive leverage.
`20:00` Trillion-Dollar TAM Fantasies vs. Corporate Revenue Realities
The discussion turns to the underlying logic of the AI capital expenditure boom: the idea that companies are buying up GPUs that will retain high intrinsic value, forming a "build it and they will come" foundation. Alexander Altmann acknowledges that while GPUs will not simply drop to zero value, market participants must remain alert to early warning signs. He points back to a Barclays note published in late August titled *"Placing Your AI Chips More Carefully,"* which tackled the absurd scale of market projections. Around that time, SpaceX had pitched a total addressable market (TAM) estimated at $30 trillion—an amount roughly equivalent to the entire annual GDP of the United States.
```mermaid
flowchart TD
accTitle: S&P 500 Revenue Breakdown vs AI Expenditure Fantasies
accDescr: Diagram contrasting S&P 500 corporate revenues and net earnings against multi-trillion-dollar AI TAM projections.
subgraph TAMDisconnect["Projected Tech TAMs"]
SpaceX["SpaceX TAM Pitch: ~$30 Trillion"]
Hist["Combined Microsoft + Apple Historical Revenue (Since 1990): <$10 Trillion"]
end
Prof -.->|"Must fund all AI rent & token spend"| TAMDisconnect
```
Altmann contextualizes these numbers using a statistic from his colleague Cameron Crise: looking back across modern tech history to 1990, the inflation-adjusted cumulative revenue generated by Microsoft and Apple combined barely reached $10 trillion. Pitching a future market of $30 trillion raises the fundamental accounting question of where that capital will actually come from. Across the entire S&P 500, top-line revenue totals approximately $17 trillion against $11 to $12 trillion in operating expenses, leaving an aggregate net corporate profit pool of only $5 to $5.5 trillion. That finite profit pool must fund all token fees, cloud rental contracts, and computing expenditures, yet alternative tracking sources like Ramp corporate card data and OpenRouter query activity already show corporate token expenditure plateauing in the near term.
- The S&P 500 aggregate revenue is roughly $17 trillion, with expenses near $11–$12 trillion and total net earnings around $5–$5.5 trillion.
- Cumulative inflation-adjusted revenues for Microsoft and Apple combined since 1990 sit below $10 trillion.
- Alexander Altmann warns investors to watch the leading indicators closely: “So when you start to see cracks or canaries in coal mines or however you want to characterize it, you should probably pay attention.”
---
`22:30` Frontier Labs, S-1 Disclosures, and the Open-Weight Squeeze
Joe Weisenthal and Tracy Alloway debate the divergence between public benchmark tracking and corporate financial filings. Acknowledging that Barclays has a banking role regarding Anthropic that prevents Altmann from commenting on the firm directly, Joe notes that Anthropic’s S-1 filing showed top-line revenue climbing sharply upward, defying narratives of an immediate slowdown. He argues that datasets from aggregator services like OpenRouter merely capture consumers seeking discounted Chinese models rather than the broader enterprise spend. Tracy counters that while top-line revenue looks impressive on an S-1, the core market tension remains the massive gulf between single-digit billions in revenue progression and valuations sitting in the hundreds of billions or trillions.
Turning to the competitive landscape of frontier models, the hosts address the rapid encroachment of open-weight alternatives. While proprietary frontier labs promise superior agents worth paying for, open-weight models have rapidly closed the benchmark performance gap at dramatically lower operating costs. Altmann highlights that open-weight options have expanded from a modest 20% slice to as high as an 80% market share across multiple telemetry datasets. Compounding this, the cost per token has experienced relentless deflation, dropping to roughly 1/100th of earlier levels, even while raw GPU data center rental rates have moved upward.
- Open-weight model adoption has surged from roughly 20% to nearly 80% market share according to multi-source industry usage data.
- Token pricing has plummeted, making it one of the steepest deflating tech assets in modern history.
- Contrasting the token cost collapse, data center GPU rental rates have continued to trend higher due to tight compute capacity.
---
`25:00` On-Device Local Compute and Shifting the Efficiency Curve
The debate over frontier dominance versus open models is evolving into a question of compute architecture: whether massive centralized data centers will remain necessary for routine productivity. Altmann describes this strategic fork as being either "long or short human creativity." Drawing an analogy to mobile hardware, he notes that in 2007, an engineer told that a smartphone would pack orders of magnitude more compute 20 years later without changing battery size would have assumed the device had to be the size of a boardroom conference table. Instead, extreme efficiency gains brought supercomputers into users' pockets.
```mermaid
flowchart LR
accTitle: Shift from Hyperscale Data Centers to Edge Inference
accDescr: Flowchart showing how user queries move from distant hyperscale clusters to local on-device inference chips.
subgraph CentralizedCompute["Hyperscale Cloud Route"]
Hoppers["Data Center Hoppers / Vera Rubin GPUs"]
Ohio["Remote Clusters (e.g., Ohio)"]
Hoppers --- Ohio
end
subgraph EdgeCompute["Local On-Device Execution"]
Phone["Smartphone Hardware Engine"]
LocalModel["Open-Weight Local Models (e.g., Meta / NVIDIA Nemotron)"]
Phone --- LocalModel
end
The assumption that everyday consumer and workplace agent tasks—such as drafting an angry hotel complaint or sorting inbox workflows—require querying clusters of NVIDIA Hoppers or Vera Rubin systems in an Ohio data center is being undermined by local edge execution. With tech companies such as Meta and NVIDIA releasing multi-billion-parameter models (such as Nemotron) configured to execute directly on consumer laptops and mobile processors, the unit economics of AI inference are decoupling from centralized hyperscaler cloud infrastructure. While frontier scale remains necessary for advanced breakthroughs like novel biotechnology drug discovery, daily utility tasks are migrating down the efficiency curve to local silicon.
- Alexander Altmann frames the evolution of hardware efficiency: “The idea that we necessarily need to ping a bunch of Hoppers or Vera Rubins in an Ohio data center just so that we can sort of get our, you know, agent manage my email for me, yeah, or to, you know, to send an angry response letter to the hotel to do that with the chip on your phone by next year.”
- Meta and NVIDIA have already deployed multi-billion parameter open-weight models optimized to run entirely on local personal laptops and mobile chips.
---
`27:30` Sponsored Interlude: SpaceX Culture and Enterprise Risk
The broadcast pauses for network programming, beginning with the *Bloomberg Tech Minute* reported by Carol Masser. The segment highlights a *Bloomberg Businessweek* piece by Lauren Grush examining the rigorous hiring ecosystem at SpaceX. Securing a role at the firm demands intense dedication to Elon Musk's workplace culture, where demonstrable hands-on problem-solving and innate intelligence take precedence over advanced academic credentials. Recruiters explicitly condition prospective applicants on demanding schedules that include extended evening and weekend shifts, motivated by the larger mission of deep space exploration alongside compensation packages heavy in equity and long-term vesting restricted stock units (RSUs).
The programming then transitions into commercial features for Cincinnati Insurance—emphasizing operational continuity, risk management through independent agent networks, and readiness for unexpected asset losses—followed by a brief spot for Delta SkyMiles business services.
`30:00` The Frontier Sink and the Missing Consumer Layer
Joe Weisenthal brings up a concept he encountered during a recent visit to San Francisco: the idea of a "frontier sink." The question at its core is which specific industries actually demand PhD-level and beyond frontier intelligence. Biotechnology is an obvious candidate, as medicine can essentially never have too much intelligence; semiconductor design and quantitative trading are other plausible sectors where cognitive ability cannot be oversupplied. However, for a massive portion of the broader economy, a standard level of capability is already more than sufficient, meaning true frontier intelligence may not be broadly required across ordinary industries.
```mermaid
flowchart TD
accTitle: Frontier Intelligence Demand vs Economic Reality
accDescr: Contrasts the few sectors needing frontier intelligence against the vast economy that only needs basic agents.
subgraph Sinks["Frontier Sinks (Endless Intelligence Demand)"]
Bio["Biotechnology & Medicine"]
Semi["Semiconductor Design"]
Quant["Quantitative Trading"]
end
Sinks -.->|"Too niche to fund"| Capex["Trillion-Dollar AI Capex Buildout"]
Consumer -->|"Must justify & plug"| Capex
```
Alexander "Alty" Altmann compares today's early AI wave to the debut of the iPhone in 2007–2008, when people were amazed by trivial applications like the digital beer pint app where tilting the phone simulated drinking. While Joe dismisses such apps as mere whimsy—declaring he hates fun—Altmann emphasizes that technology rarely starts with its ultimate killer use case. Altmann argues that society does not yet know what the future of AI will look like, and for everyday life, people do not need more than an assistant with an average human IQ sitting in their pocket to handle mundane friction, such as disputing an unfair $2,000 ding charge from a car rental company.
Enterprise spending alone, Altmann contends, simply cannot bridge the massive revenue gap demanded by the current capital expenditure boom on the timelines Wall Street expects. Instead, true long-term justification must come from the consumer, who drives the bulk of the economy. Furthermore, human technological leaps often stem not from pushing scientific frontiers, but from finding clever applications for existing tech. In Altmann's view:
> “I said like when we figure out that killer consumer application layer because at the end of the day the consumer is everything it's the economy it's where to the point about revenues like enterprise revenues i just don't think is going to plug that gap really i just don't see how like we yeah no you said you're right the numbers i don't see it can probably get there eventually but it can't get there in the time frame that the market wants to see the roi”
- Altmann predicts that deploying a 100-IQ agentic workflow that becomes ubiquitous over the next 12 to 18 months could power the next real upswing in AI adoption.
- Enterprise revenues are seen as structurally insufficient to deliver the immediate return on investment (ROI) that the market demands for hyperscaler capex.
---
`35:00` Solo Parenting with ChatGPT, West Coast Echo Chambers, and Capex Cracks
Joe shares a personal anecdote from a recent weekend of solo parenting while his wife was away. Despite having a busy weekend schedule packed with activities and play rehearsals, he relied on ChatGPT to parse cluttered email threads to confirm pickup times—such as discerning whether his daughter was required for the full 1:00 p.m. to 6:00 p.m. rehearsal block or could leave at 3:00 p.m. Tracy Alloway jokingly mentions having seen Joe tweet about this, prompting Joe to protest that she had seen it on his private alternative account ("alt"), which he never intended for public broadcast. Altmann interrupts to question the economic value of such an AI assistant, asking what a user would actually pay per month for an administrative tool of that sort, because answering that question is critical to establishing whether consumer AI can generate the revenue needed to justify current infrastructure outlays.
Turning back to the fragility of the AI bull thesis, Tracy asks what catalyst could shake investor conviction and degrade market faith. Altmann suggests that the first major crack would be an announcement from a hyperscaler reining in its capital expenditure. He notes a stark geographic divide: on the West Coast, proximity to the technology's epicenter fosters an echo chamber where investors and executives are deeply bought into the bullish narrative. Altmann expects that any tempering of enthusiasm will likely take until the new year, given that Google previously reignited market momentum with statements promising significantly higher capex, while Oracle managed to reassure investors at its investor day by limiting incremental fundraising to around $20 billion through May. When asked what could spook markets, Altmann observes:
> “Off the top of my head... I would say it would be some level of capex being reined in by one of the hyperscalers in fact on the walk over here i was asked by one of my clients who was saying whether questioning whether that would actually even come up potentially in the results this week and famous last words It's probably still premature.”
---
`38:00` A Knife Fight in a Phone Booth: The Volatility Paradox
The conversation transitions into broader market mechanics and index construction. Altmann describes the current market environment as one where the headline S&P 500 appears largely dormant, but beneath the surface lies intense, turbulent churn. Shifts in capital expenditure expectations cause sharp cross-sector rotations—where developments favorable to hyperscalers end up penalizing memory manufacturers—leading Altmann to characterize managing long/short portfolios in this environment as a "knife fight in a phone booth."
Tracy then asks why standard metrics of volatility appear so subdued—with both realized volatility and the VIX remaining low and benign—despite widespread anxiety and extreme positioning around the single most critical trade in global financial markets. Altmann acknowledges that this paradox is the single most common question he receives from institutional clients. He explains that the discrepancy breaks down into two distinct explanations: a structural shift in how equity capital is currently deployed and managed (rooted in the reality that equities are fundamentally engineered as long-only assets), alongside purely technical market dynamics.
> “And you just get this constant violent rotation. And I call it like a knife fight in a phone booth, you know, for all these funds that have to run long, short portfolios.”
`40:00` Pod Shops, Factor Crowding, and the Unmappable AI Beta
Alexander Altmann walks through the seismic shift in hedge fund mechanics over the past several years. In 2020, assets under management (AUM) deployed in multi-manager funds stood around $300 billion; today, that figure exceeds $1 trillion, even before factoring in the leverage extended through prime brokers. Altmann explains that higher interest rates have provided an ideal backdrop for market-neutral long/short equity models. Rather than locking up pure capital across gross long and gross short books, funds deploy units of risk and post margin, while sweeping the vast pool of unencumbered cash into high-yielding, interest-bearing instruments to clip an attractive coupon. The objective is simply to generate idiosyncratic alpha on top of that cash yield. However, because the total pool of AUM and leverage has more than tripled while the universe of liquid tradeable single stocks has stayed flat or shrunk, intense crowding is inevitable.
To mitigate that crowding, funds rely on ubiquitous quantitative risk models like Barra and Axioma to neutralize standard style factors—such as value, growth, momentum, size, and quality. Yet Altmann observes that because every pod shop runs the same optimization routines, they have simply created new layers of factor crowding beneath the surface. Volatility rarely shows up at the headline index level because the multi-manager capital is largely index-agnostic, but factor rotations underneath have turned exceptionally violent.
Compounding the problem is an inability to model or hedge AI itself. When institutional clients ask Altmann if he has managed to map the AI factor, his direct answer is no: “the AI factor is changing all of the time.” Earlier in the year, being short enterprise software was considered an integral part of playing the AI theme, but with software rebounding toward 52-week highs, that correlation broke down entirely. As Altmann concludes:
> “And bottom line is, is you're either long AI or you're short AI. And let's face it, the past three years, being short AI probably hasn't been a particularly lucrative trade.”
- Multi-manager pod shop AUM expanded from approximately $300 billion in 2020 to well over $1 trillion.
- Prime brokerage leverage amplifies this capital while the underlying universe of tradable equities has remained stagnant or contracted.
- Risk models (Barra, Axioma) neutralize traditional style factors, but the collective rush to hedge those factors produces aggressive cross-asset factor crowding.
- The AI factor resists static quantitative modeling because its market narrative and component betas shift constantly (e.g., short software transitioning to software reaching 12-month highs).
`43:24` Breaking Mega-Cap Earnings and the Unprecedented Scale of AI Capex
Mid-discussion, Tracy Alloway interrupts with breaking earnings headlines flashing across her Bloomberg terminal for Alphabet and Microsoft, presenting a vivid real-time demonstration of how violent single-stock dispersion nets out at the index level. Alphabet tumbled 7.9% after revenue ex-TAC missed estimates, while Microsoft surged 6.4% on cloud strength. Testing Altmann's market feel, Tracy asks him to guess Microsoft's Q1 Azure revenue growth without checking the terminal. Altmann guesses 47%, extraordinarily close to the reported 48% (beating the Bloomberg consensus estimate of 44.8%). Google Cloud revenue hit $30 billion, printing an astonishing 103% growth figure on the terminal. Altmann points out that when hyperscalers operate legacy cash-volcano businesses, they have the luxury of funding the largest capital expenditure experiment in American history.
```mermaid
flowchart TD
accTitle: Hyperscaler Capex Engine
accDescr: Diagram illustrating how legacy monopoly cash flows fund massive AI infrastructure capex.
subgraph InfrastructureBuild["Unprecedented AI Capex"]
B --> C["Hyperscaler Infrastructure<br/>Chips, Datacenters, Power"]
C --> D["Cloud Growth Reacceleration<br/>Azure: +48% | Google Cloud: +103%"]
end
subgraph MacroScale["Macroeconomic Footprint"]
C --> E["Projected Capex Run-Rate<br/>~12% of GDP (~2x Netherlands)"]
C --> F["S&P Concentration<br/>75% Earnings Growth from Semis/Memory"]
end
```
Contemplating the sheer scale of the forward capex commitments, Altmann notes that projected spending stands at roughly 12% of GDP—equivalent to "two Netherlands." In dollar terms over a compressed timeline, it eclipses historical American infrastructure projects like the transcontinental railroads or the Louisiana Purchase, leaving ancient marvels like the Pyramids or the Great Wall of China as the only conceptual points of comparison. Furthermore, this capex deluge has warped the benchmark: around 75% of next year's S&P 500 earnings growth is projected to originate solely from memory and semiconductors. Altmann and Tracy remark that the American index has undergone an extreme structural transformation, turning into the tech-hardware-heavy South Korean benchmark:
> “So you basically, S&P has become KOSPI.”
- Microsoft Q1 Azure revenue grew 48% year-over-year, beating Bloomberg survey estimates of 44.8%.
- Alphabet dropped 7.9% on its revenue ex-TAC miss, even as Google Cloud revenue surged past $30 billion.
- Hyperscaler forward capex is tracking near 12% of GDP, dwarfing historical benchmarks in both private and public infrastructure.
- Approximately 75% of projected S&P 500 earnings growth next year is driven by semiconductors and memory chips, sparking the coining of the S&P's "KOSPI-fication."
`47:11` Post-Interview Debrief: The Myth of Rate Immunity and Undifferentiated Alpha
Following Altmann's departure, Joe Weisenthal and Tracy Alloway reflect on the key takeaways from the conversation, focusing on the shifting market regime from indiscriminate capital abundance to harsh differentiation. Joe points out the recurring fallacy among tech bulls who claim that hyper-growth renders the sector entirely immune to monetary policy:
> “If you're growing 44 percent in a year, what do you care about? Well, the rates are five percent or five point two five. It's completely irrelevant. You're just going to keep building, etc. But that is a lie that investors tell themselves.”
Joe emphasizes that interest rates undeniably matter for capital-intensive, debt-reliant players such as neoclouds and leveraged infrastructure providers.
Tracy and Joe also dismantle the core premise of modern pod shops. While multi-manager platforms market themselves on delivering diversified, market-neutral, idiosyncratic alpha across independent pods (covering industrials, chips, software, and utilities), the macro reality is far simpler: anyone who generated meaningful positive returns at scale over the prior twelve months was, deliberately or accidentally, just expressing a long AI factor trade. The entire valuation scaffolding of the hyperscalers depends on whether external corporate customers will actually see the productivity and earnings gains necessary to justify buying these cloud services over the long run, rather than the ecosystem simply cycling cash within an insular semiconductor loop.
`50:00` The "AI-Pilled" Debate and Staggering Token Metrics
Reflecting on the wider market narrative, the hosts recount an exchange with an observer who claimed that Wall Street simply is not "AI-pilled" enough. That claim seemed immediately bizarre on its face, given the massive rally in equities where tech and semiconductor stocks have been rocketing straight to the moon and valuations across trading screens tell a story of extreme enthusiasm. However, the observer’s logic was that anyone even willing to entertain the possibility of an impending GPU overabundance fundamentally fails to grasp the sheer scale and inevitability of the technology. As the host relates, the prevailing attitude among true believers is: “That person's view was that if you're even having a discussion about the possibility of overabundance of GPUs, you just don't get it.” In their eyes, entertaining traditional supply-demand constraints or questions about excess capacity merely signals that one does not understand artificial intelligence sufficiently.
The conversation turns to the mind-boggling operational figures being reported during big tech earnings to justify the continued capital expenditures. Alphabet, for instance, reported that its models are processing 32 billion tokens per minute—a figure that surged 40% quarter-over-quarter. Confronted with the surreal scale of computing jargon now dominating earnings calls, the hosts laugh at the dizzying vocabulary entering equity analysis, remarking, “I don't even know what these words mean.” With market expectations soaring and metrics decoupling from traditional financial benchmarks, they conclude that it remains a uniquely wild, fascinating period to cover equities.
- Alphabet reported processing 32 billion tokens per minute, representing a 40% increase quarter-over-quarter.
- Skepticism around future GPU overcapacity is increasingly treated by AI evangelists as evidence of an analytical failure to understand the technology's trajectory.
---
`50:38` Odd Lots Wrap-Up and Community Channels
Tracy Alloway and Joe Weisenthal close out the episode and direct listeners to their respective social handles and community hubs. They encourage listeners to follow them online (@TracyAlloway and @TheStalwart), as well as the show’s production team—Carmen Rodriguez (@CarmenArmand), Dashiell Bennett (@Dashbot), Kale Brooks (@KaleBrooks), and Kevin Lozano (@KevinLloydLozano). Listeners looking for ongoing discussion are directed to the *Odd Lots* Discord community, which runs 24/7 discussions on market movements, macro themes, and the AI infrastructure buildout.
Tracy highlights the daily *Odd Lots* newsletter available at Bloomberg.com, which provides additional reporting and market analysis, and encourages listeners to rate, review, and subscribe across podcast platforms and video channels. For Bloomberg subscribers, all podcast episodes are accessible ad-free via the Bloomberg channel on Apple Podcasts.
## Notable quotes
> `12:35` AI is not issuing debt in a vacuum. There are other verticals that need to tap financing, and there is a finite amount, right? — Alexander Altmann
> `40:00` You're either long AI or you're short AI. And let's face it, the past three years, being short AI probably hasn't been a particularly lucrative trade. — Alexander Altmann
## Mentioned
Alexander 'Alty' Altmann, Alphabet, Amazon, Anthropic, Apple, Axioma, Barclays, Barra, Brannin McBee, Cameron Crise, Carol Masser, ChatGPT, Cipher Mining, Cincinnati Insurance, CoreWeave, Delta SkyMiles, Elon Musk, Federal Reserve, Google Cloud, Iren, Joe Weisenthal, Lauren Grush, Meta, Microsoft, Microsoft Azure, Nemotron, NVIDIA, OpenRouter, Oracle, Ramp, SK Hynix, SpaceX, Tracy Alloway
## Video description
<p>It has been a big week for the AI trade — we've seen wobbles in the market, many of them in response to headlines like the Anthropic S-1 and earnings from the hyperscalers (Oracle, Microsoft, etc.). So what does Wall Street make of all the news, and how are they processing an increasing sense that the laws of the economy barely apply to AI? Today, we speak with Alexander 'Alty' Altmann, global head of equities tactical strategies at Barclays, and he helps us make sense of how Wall Street really feels about the AI trade right now. We also talk about the KOSPI-fication of the S&P 500, if there are any historical precedents for the capex going into the AI buildout, and if AI is really immune to interest rates.<br><br>Read More: <a href="https://www.bloomberg.com/news/articles/2026-10-28/microsoft-reports-quarterly-cloud-sales-that-beat-estimates?srnd=homepage-americas?utm_medium=referral&utm_source=podcast&utm_campaign=odd_lots&utm_content=article">Microsoft Reports Rising Cloud Sales That Show Payoff From AI</a><br><br>Only Bloomberg.com subscribers can get the Odd Lots newsletter in their inbox, plus unlimited access to the site and app. <a href="https://www.bloomberg.com/subscriptions/oddlots?in_source=oddlotspodcast" data-link-cookie="1" data-destination="https://www.bloomberg.com/subscriptions/oddlots?in_source=oddlotspodcast">bloomberg.com/subscriptions/oddlots</a></p><p>See <a href="https://omnystudio.com/listener">omnystudio.com/listener</a> for privacy information.</p>
While equity markets remain overwhelmingly committed to the AI trade, Barclays tactical strategist Alexander Altmann warns that emerging credit strains, capital supply limits, and an unrealistic consumer revenue gap indicate the cycle may be entering a denial phase.
Credit Strains and the Neocloud Financing Divide (7:10)
A critical boundary has emerged between cash-rich hyperscalers like Alphabet and Amazon—which Altmann characterizes as self-funding 'cash volcanoes'—and peripheral 'neoclouds' like CoreWeave and Cipher Mining. While tech bulls often assume massive growth renders the sector immune to interest rates, neoclouds rely almost exclusively on external debt to acquire compute capacity. This reliance leaves them exposed as debt markets tighten. Altmann likens stress in the neocloud periphery to subprime vehicles preceding past financial crises, noting that when Cipher faced closed direct debt markets and issued dilutive convertible notes, CoreWeave bonds fell in sympathy despite issuing no negative news.
Macro Capital Bottlenecks and Finite Corporate Profit Pools (10:00)
The assumption of frictionless capital is colliding with macro constraints. Investment-grade and high-yield net issuance is projected to hit roughly $1.1 trillion in a single year, breaking a three-decade norm by surging despite restrictive benchmark yields. Altmann notes that roughly 60% of surveyed institutional clients anticipate AI financing bottlenecks, warning that data center debt is aggressively competing against other corporate borrowers. Furthermore, market-projected total addressable markets—such as SpaceX's pitched $30 trillion TAM—clash with corporate reality: the entire S&P 500 generates approximately $17 trillion in revenue and only $5 trillion to $5.5 trillion in net profit to fund all AI tokens, cloud spend, and software subscriptions.
The Efficiency Curve and the Missing Consumer Layer (25:00)
A structural shift in technology threatens the economics of massive centralized infrastructure. Altmann observes that open-weight model usage has expanded to an estimated 80% share while token costs have deflated to 1/100th of previous levels. Simultaneously, multi-billion-parameter open-weight models from Meta and Nvidia can run directly on consumer laptops and mobile processors, casting doubt on whether routine administrative tasks require expensive queries to distant Hopper or Vera Rubin clusters. Altmann argues enterprise budgets alone cannot provide the immediate return on investment the market demands, meaning the ultimate financial justification for current hyperscaler capex must come from an unproven consumer application layer.
Pod Shop Mechanics and S&P Concentration (40:00)
Multi-manager hedge fund assets have expanded from around $300 billion in 2020 to over $1 trillion. While these pod shops use risk models like Barra and Axioma to neutralize traditional style factors, the shared hedging techniques have driven intense factor crowding beneath a deceptively calm headline index. Furthermore, forward hyperscaler capex has reached a run-rate near 12% of GDP, with approximately 75% of next year's S&P 500 earnings growth projected to come exclusively from semiconductors and memory. Altmann and the hosts observe that this extreme hardware dependence has effectively converted the American benchmark into South Korea's tech-hardware-heavy KOSPI index.
The central unresolved dilemma is whether external enterprise customers and end-consumers will realize genuine productivity gains quickly enough to fund the unprecedented capital expenditure pipeline. While evangelists dismiss concerns of compute overcapacity, Altmann suggests that any future deceleration in hyperscaler capex announcements represents the primary catalyst that could rupture the market's bullish consensus.
[12:35] Alexander Altmann: AI is not issuing debt in a vacuum. There are other verticals that need to tap financing, and there is a finite amount, right?
[40:00] Alexander Altmann: You're either long AI or you're short AI. And let's face it, the past three years, being short AI probably hasn't been a particularly lucrative trade.
Original transcript
[0:00] There are some market stories where you want every detail. Joe and I have made quite a few podcasts on that basis, but sometimes you've only got 10 minutes and just want to know what's moving markets fast. That's the Barclays Brief podcast. Every week, experts from Barclays Markets and Research get you up to speed on what's happening and what to watch next. Concise, focused, and brief. Search Barclays Brief wherever you get your podcasts. This is the Bloomberg Tech Minute I'm Carol Masser. Getting hired at SpaceX requires exceptional ability and may include having Elon Musk review a candidate's resume. So writes Bloomberg Businessweek's Lauren Grush. She reports that the company culture is defined by Musk and job candidates must understand that a commitment to SpaceX means adhering to this culture and exhibiting loyalty to Elon Musk. To get hired, candidates must demonstrate their achievements from past jobs degrees, with the company looking for people with raw, innate intelligence. A spokesperson for SpaceX didn't respond to a request for comment on its recruiting and hiring practices, but according to recruiters, advisors, and those who've interviewed for jobs there, it's the space exploration dream that draws people to the company, and Musk invokes that dream regularly to remind workers why they're there. Prospective hires are told that the job will require stamina, including in job ads that hours and weekends as needed. One major perk for all that hard work, employees access to equity, either the right to buy shares or payment in the form of restricted stock units that invest over time. That's the Bloomberg Tech Minute brought to you by ChachiPT. Get started at ChachiPT.com. you I'm here with Dr. Steven Rothenberg, a leader in minimally invasive and robotic pediatric surgery at HCA Health One's Rocky Mountain Children's, which is known for this expertise. Dr. Rothenberg, how does this approach differ from traditional surgery? So instead of making large incisions, we make tiny incisions and use small scopes and small instruments to affect the same procedures, children. How do you help parents understand the risks and the benefits? So we talk about the specific surgery they need and then the benefits of not having giant incisions that then grow with the children can cause musculoskeletal deformity and increase their pain postoperatively. Learn more about pediatric surgery at HCA Health One Rocky Mountain Children's. Visit hcahealthone.com forward slash Rocky Mountain Children's. Bloomberg Audio Studios. Podcasts. Radio. News. Hello and welcome to another episode of the All Thoughts Podcast. I'm Tracy Allaway. And I'm Joe Weisenthal. Joe, interesting times in markets. You know what? I kind of feel like I lost the plot. I know. I was like, I had this realization. I don't know what anything's going on anymore. We're just talking markets. And obviously I lost the plot with life in general. Like no one's disputed that. But just like, what are the consensus trades these days? Like what are people actually worried about with some of these data center stuff? Like what core we have, what, et cetera. I have no idea what anyone thinks anymore. Like I need a reality check badly. So we're recording this on October 28th. in terms of market reality checks. We're actually doing better on this one because we often end up recording these right before a Fed decision or right before an S1 comes out or right before earnings guidance or something. This time, we've timed it pretty well. So we just had the FOMC decision. Earlier this week, we had the S1 filing from Anthropic. During the Fed presser, Oracle came out with some additional guidance. There's a lot of stuff going on, even though There's more stuff coming up later this week. And so I feel like it's a really good moment in time to kind of stop and figure out what's going on. What's going on? Seriously. So – You mentioned Oracle had some Investor Day data coming out. The stock is up 12%. I know. I would say this was a bigger story today than the FOMC. Yeah. Right? Like the questions about... Capital expenditures, how much debt the big hyperscalers and so forth are going to take on. Clearly, I would say in the minds of many participants, much bigger questions than like, oh, did they're going to hold in October or whatever? Not a high bar given Kevin Warsh's ability to deflect. I tuned in for a few minutes to get press conference. But you're right. That guy's good at nonsense.
[5:00] All right. Well, we do, in fact, have the perfect guest. We're going to be speaking with someone who we've had on the podcast before. Last time we were talking about levered ETFs. This time we're going to get a general sort of market overview. It's Alex Altman, Global Head of Equities Tactical Strategies over at Barclays. So, Alex, thank you so much for coming back on All Thoughts. Thank you very much for having me again. What's going on? What's going on? No, I mean, is there like a consensus AI outlook at the moment? yeah there is the consensus ai outlook is is that ai is still going to this train is in motion yeah we're not stopping this train that's still the consensus like to be very clear and Let's say if you were to sort of take a sample of... and invest a bunch of investors across different cohorts, whether it's in the US, Europe, Asia, and ask them gun to their head, do you want to still be bullish or bearish the ai story you're going to get an overwhelming positive answer yeah right and we can debate like where people are on the positioning scale but like that's the general view right and this is why i think that if you you know you've probably seen that chart of where we are in like the bubble kind of thing and then that as it begins to roll over the first dip down is the denial one yeah of course i'm not going to say we're at kind of this bubbles popped i know there's commentary out there that actually has said that but i definitely if you were to sort of juxtapose those two narratives we're in the denial phase right now because everyone is still extremely bullish yeah yeah you got you got guys saying like oh semiconductors you got hynix or whatever or some of these korean memory stocks they're on three times earnings i'm like you can on a Bloomberg terminal. That requires no work at all, right? So everyone is out there. Thank you for plugging the utility of the Bloomberg terminal, by the way. What do you mean there are 3X earnings? There are memory stocks that are... They're traded with an EPS of... Not an EPS, a PE of three. Correct. Because people... There is... And because the perception is that we're at such a peak and we know how cyclical of an industry it is, that when the earnings are this good, you're not going to pay up for them. Exactly. The old mantra is you don't buy cyclical companies, especially ones as historically cyclical as semiconductors, when they're at peak earnings. I didn't realize they were that low. Okay. So now. So let's put some pieces together. So we have the fact that the PE is actually quite low. People do not expect exponential growth from the memory companies in terms of their EPS anytime soon, as implied by the PE ratios. What is going on with the neoclouds? Why is it that despite, as you say, the consensus is bullish, still everyone wants to be long, the AI trade, suddenly there's all this anxiety about, you know, the people are quoting core weave CDS charts. to take a step back for a second right because i want to talk about neoclouds in the context of the broader ai ecosystem right so obviously neoclouds are a group of businesses that are effectively in some way shape or form mandated to build compute which they will then sell to predetermined customers some some which predetermines some of them will be speculative regardless of what you may believe in that business model or not these are tip of the spear these businesses don't have operating cash flow today or minimal operating cash flow today in many cases and so they represent uh the riskier part of the financing aspect of the ai spectrum right and i think that's that's an irrefutable fact you know people can talk about well actually their skies and actually it they're not they're nowhere near as risky as people think they are sure let's have that debate but compared to say the hyperscalers whether it's you know the googles and the amazons who have got huge up in their cash volcano businesses they are different different sort of category and and the reason i want to make that distinction is because if you think of neoclouds as periphery and hyperscalers as core yeah and i want you to draw a parallel not directly because right the problem doesn't didn't start with gc banks right it started with subprime vehicles 12 18 24 months beforehand right and so what i find most interesting is that there is evidently a financing problem within neoclouds right that is evident cypher tried to they did their convert nuked their own share price in doing so because they clearly couldn't tap the debt markets directly A company you've never heard of, and yet the reason why CoreWeave Bond sold off was not because of CoreWeave. They didn't say anything that week. It was because Cypher, another neocloud, tried to get tried to tap financing and in doing so materially hurt their own share price and then all of a
[10:00] sudden a bunch of investors, Quick off the mark, put two and two together. right and then it spilled over but like quite frankly if you look at the data center bonds in isolation whether like the beignets of this world that you know that's the meta one exactly yeah these had already been widening for months prior to that right and we'd seen a creep in ig hyperscale cypher is one of these ones that pivoted from bitcoin mining to getting into ai yeah yeah all right so i guess the point i'm trying to make is is that There's ripple effects in financing that start in... in the periphery and we're seeing that so what i find amazing about the original point in terms of you asked what's going on are people bullish or bearish and we literally just conducted a survey with you know group of investors yeah and and the only i mean you could make this argument this is a high number but from the responses that we've got so far around about 60 of investors we asked them a question do you think financing is going to be a problem ai financing is going to be a problem for the physical foreseeable future only 60 said yes By definition, what's happened over the past couple of weeks means that answer should be 100%, right? And yet, let's just call it just over half thing or just under half thing, there is no financing issue, but there is. So what's happening in AI to me is evidence of like almost a denial of a reality that's already here. To be fair, one of the reasons CoreWeave started selling off was because Brandon McBee gave an interview to the FT where he was talking literally about unlimited amounts of borrowing, which I think ties into what you're discussing when it comes to capital concerns. On the financing side, I mean, we've seen we've seen some differentiation in terms of lending money to different AI entities out there. Is the overarching point that we've been in a period where it seemed like there was, in fact, unlimited money and people could tap the IG markets or the loan markets for as much as they needed. for. That is exactly it. Right. Because... Part of the job that my team does is we travel the world. We talk to a lot of different investors. We ask questions. We collate the kind of feedback that informs our decisions. And there was this almost universal view that, and you guys would have heard it as well, AI was immune to high interest rates. Yeah. evidently is not true. right and so there is a spectrum and even i think that's where i think the meta results were pretty instructive in the 14c in the sense that they highlighted the fact that financing does constrain that financing does have limitations and we wrote a load of we did a load of work on this in the summer where we were looking not specifically at ai we were just looking at the holistic net issuance profile of the market and maybe you guys saw this and we looked at just The entire corporate market is IG high yield. We looked at treasuries and we looked at equity. And we went back to about 1995, collated all the data and looked at each market, the amount of net issuance in any given year as a percentage of each market's sort of market cap in that given year. And then we plotted it over time. And what we found was there was unsurprisingly a very strong relationship when interest rates went down, overall issuance went up and vice versa, as you'd expect. The laws of the economy still exist. Well, apparently not so much in 2025 and 2026. Yeah, here's the punchline. Yeah, exactly. This is the punchline. 2026 is completely shattered what you've seen in the thing in the norms. I think, Tracy, you looked at that chart. Yeah, so you had this great chart plotting the relationship between the two-year yield and net issuance. And you can see the relationship really breaks down in 2026. So we are issuing a lot more. a lot more debt related to benchmark rates than... pretty much like ever in history it's it's kind of biblical it's it's certainly we have data back to the 90s so you know if you compare it to say dot com and the interesting about dot com is of course there was a lot of equity issuance but and interest rates were high but the u.s government was basically running a surplus and all this kind of stuff so today is very unique and the reason why i bring it back to the sort of ai financing is because there is this competition for capital it is real ai is not issuing debt in a vacuum there are other verticals that need to tap financing And there is a finite amount, right? So IG issuance and a high yield this year on a net basis, we forecast to be $1.1 trillion, right? And we just throw around trillions as though it's just this casual number that everyone's got. But, like, there is limitations, right?
[15:00] *music* There are some market stories where you want every detail. Tracy and I have made quite a few podcasts on that basis, but sometimes you've only got 10 minutes and need to know what's moving markets and why. That's the Barclays Brief Podcast. Every week, experts from Barclays markets and research get you up to speed on what matters and what to watch next. About the time it takes to grab a coffee, so search Barclays Brief wherever you get your podcasts. This is the Bloomberg Tech Minute brought to you by Chachi PT. I'm Carol Masser. Getting hired at SpaceX requires exceptional ability and may include having Elon Musk review a candidate's resume. So writes Bloomberg Businessweek's Lauren Grush. Job candidates must understand that a commitment to SpaceX means adhering to this culture and exhibiting loyalty to Elon Musk. To get hired, candidates must demonstrate their achievements from past jobs. And hands-on experience matters more to recruiters than advanced degrees. With the company looking for people with raw, innate intelligence. A spokesperson for SpaceX didn't respond to a request for comment on its recruiting and hiring practices. It's the space exploration dream that draws people to the company, and Musk invokes that dream regularly to remind workers why they're there. Prospective hires are told that the job will require stamina, including in job ads that candidates, quote, must be willing to work extended hours and weekends as needed. One major perk for all that hard work? Employees' access to equity, either the right to buy shares or payment in the form of restricted stock units that vest over time. Chachi PT. Get started at ChachiPT.com. If you listen to financial news, you know a lot of time is spent thinking about what's next, the next opportunity, the next investment, the next move. But sometimes what matters most is being ready for what you never saw coming. A storm damages a second home. A pipe bursts above a collection you spent decades building. An unexpected loss puts a growing business on pause. Those are the days no one sees coming. But for more than 75 years, Cincinnati insurance has been ready when they do. With deep expertise, coverage for businesses, homes, valuables, and more, Cincinnati Insurance works with independent agents who take the time to understand what matters to you. Together, they help protect more than what you own. They help protect what you're building toward. And when a bad day does come, Cincinnati Insurance has real people there to help make things right. Because planning for the future isn't only about knowing what's next, it's about making sure you're ready for what you can't predict. Let Cincinnati Insurance make your bad day better. Find an independent agent at CINFIN.com. Running a business means you're always on the move. The Delta Sky Miles Reserve Business American Express card is made for frequent travelers who want premium benefits that works as hard as they do. For a limited time, new Delta Sky Miles Reserve Business card members can earn 200,000 bonus miles, the highest miles offered this year. Precious productive time or restful reset. You can enjoy 15 visits each medallion year to the Delta Sky Club when flying Delta and unlock unlimited Delta Sky Club access after spending $75,000 in purchases on your card during a calendar year. Plus, receive four one-time guest passes each medallion year so you can bring a colleague along. That's 15 Sky Club visits plus four guest passes for an annual value of up to $950. Value estimate based on $50 per visit rate to purchase additional visits or bring a guest to the Delta Sky Club. Apply today at www.delta.com slash podcast. Offer ends 11-426. Minimum spending requirements and terms apply. Reconcile this. If there is this perception that People are still bullish. The train is still moving at full speed. Why should capital be hesitant? You can see on the numbers, the core weaves and the irons and the cipher and a bunch of other ones that I had never heard of. Yes, clearly the capital markets are telling them fear about these companies that are at the tip of the sphere? Wouldn't the bull say, look, someone is going to pay for these GPUs. Like, The demand is going up there if they buy or rent or whatever, if they have electricity.
[20:00] They have GPUs. They bought them, whatever. They're not going to zero. They're going to be worth a lot of money some way or another. Yeah. So that is the bull argument. Yeah. And what's wrong with that? There's nothing wrong with it in terms of build it and they will come. I would say that the challenge will be, two things, to the earlier point, tip of the spear. So when you start to see cracks or canaries in coal mines or however you want to characterize it, you should probably pay attention. Yeah. That's why we have you on. And the second point is, is that, thank you. The second point is we need to think about revenues at the end of the day. And this is something which we addressed at the end of August in a note called placing your AI chips more carefully. And the idea was, is that we basically decomposed. It was around the time of the SpaceX TAM. If you remember, that was around about $30 trillion. It was a total addressable market of like the entire universe, not just Earth. Something like that. Yeah. 30 trillion dollars, which is obviously around about the GDP of the United States. And so we want, again, to the point about people throwing around trillions. Fun fact for you Since 1990 I can actually take credit for this stat by the way This actually belongs to Cameron Kreese at who also works Our yeah and so you know give him a give him the the the necessary plug on this uh looking at the cumulative revenue of the two most successful uh tech companies in modern history it was microsoft and apple and it was the cumulative tam of those two was i think inflation adjusted not even 10 trillion dollars since 1990 right so but you know spacex is addressable is talking three times that number right so the reason i so that kind of prompted us where's all this money going to come from yeah so punchline again 17 trillion dollars of s&p top line revenue right and they've got around about 11 or 12 trillion dollars of expenses and so that means that the net profit approximately of the of the s&p is somewhere between five and five and a half trillion dollars right so it's like all right We can make some bold claims about the amount of money that's going to get allocated to rent to AI expenditures and all the rest of it. But last time I checked, like we all look at the ramp data or the open router data, and we can clearly see that that token expenditure has kind of begun to level off a bit, right? At least near term. So it like all right our company is going to spend this much money And that where the other point to answer your question joe is well you know your company barclays is uh i know managing the anthropic ipo so or as you know involved in this we're not going to ask any anthropic questions to you but i will say to tracy okay no i won't make you call you know we understand the rules here but i will say to tracy that anthropic numbers on the top line looked great. And I don't think we learned that much about open router data because that captures a specific slice of people who are usually looking for cheap Chinese models, which they're doing well, but the top line numbers for anthropic look unbelievably good. That's the truth, Tracy. Thank you. There's earnings progress there in the S1. You can see the earnings progression. However, like billions in earnings versus valuations at trillions, like I think that is the gap that people are debating. I'm just pointing out that there's lots of charts. There's ramp data that's floating around. Lots of people look at the ramp data and lots of people look at the statistics on OpenRouter and they start to say oh the American Frontier Labs are they in trouble And they might be but I just saying if you look at the data that filed with the SEC as pretty recently at least one of the companies the one that is filed in S1 that number is going straight up still Why don't we talk about the frontier models more generally? So, you know, the argument has been that they can come up with a better model, a better agent, a better technology, and that people are ultimately going to pay for it. And the persistent, I guess, doubt in the market has been, well, you know, open source models are kind of catching up very quickly. They're worse on their benchmark test, but not by that much anymore. And they're a lot cheaper to operate. What are people saying about the open source versus frontier debate? I would say here this is weirdly is a bit of a contradiction versus that bullish narrative, because the data is the data you can evidently see on a whole bunch of different sources. is that open weight has kind of gone from a 20% market share to anything up to about an 80% market share. And then you'll also see that the token cost is a fraction, what, one one-hundredth versus? Incredible token deflation has happened every year. It might be one of the fastest deflating assets in history. It makes those NFTs back in 2021 too. But GPU rental rates have been going up. They have.
[25:00] Yes, absolutely. And so There is clearly a two-way debate, and I think it again goes back to the evolution of this broader AI narrative. It's not just necessarily a one-way train, rising tide lifts all boats. There's going to be degrees of differentiation. But you asked about open weight models, and I think where one of the leading-edge questions right now is – not even just open weight, but combine that with locally run open weight. And this is something which, I'd like to call being long or short human creativity, right? In so far that, So if when the iPhone came out in 2007, stay with me here, that if I had if someone from the future had said 20 years from now, you're going to have whatever the orders of magnitude compute differential and the battery is going to be the same size, that 2007 person would have said impossible. you'd need an iphone the size of this table right so clearly we got there and amongst many other things and so the idea that we necessarily need to ping a bunch of hoppers or vera rubens in an ohio data center just so that we can sort of get our you know agent manage my email for me yeah or to you know to send an angry uh response letter to the hotel to do that with the chip on your phone by next year Right, if not already. So we've all kind of got supercomputers in our pocket. And the ability to run, say, multi-billion parameter models locally using whether an open-weight model or a closed-weight model, but that's provided by NVIDIA's Nemetron or, you know, whatever Meta's putting out. Which Meta's already put out a multi-billion parameter model that's run locally on your laptop. It's crazy. Right. But it's not because that's technology, right? Yeah. debate is it's not even so much about open weight it's about the fact that you're moving to along this efficiency curve and that was also bringing in the revenue question as well which you know ties in with those frontier models as well frontier models we're not going to be producing you and i with the greatest respect to everyone in this room we're probably not going to be discovering the new biotechnological drugs Thank you. This is the Bloomberg Tech Minute brought to you by ChatGPT. I'm Carol Masser. Getting hired at SpaceX requires exceptional ability and may include having Elon Musk review a candidate's resume. So writes Bloomberg Businessweek's Lauren Grush. She reports that the company culture is defined by Musk and job candidates must understand that a commitment to SpaceX means adhering to this culture and exhibiting loyalty to Elon Musk. from past jobs, and hands-on experience matters more to recruiters than advanced degrees, with the company looking for people with raw, innate intelligence. A spokesperson for SpaceX didn't respond to a request for comment on its recruiting and hiring practices, but according to recruiters, advisors, and those who've interviewed for jobs there, it's the space exploration dream that draws people to the company, and Musk invokes that dream regularly to remind workers why they're there. Prospective hires are told that the job will require including in job ads that candidates, quote, must be willing to work extended hours and weekends as needed. One major perk for all that hard work? Employees' access to equity, either the right to buy shares or payment in the form of restricted stock units that vest over time. That's the Bloomberg Tech Minute brought to you by ChachiPT. Get started at ChachiPT.com. If you listen to financial news, you know a lot of time is spent thinking about what's next. The next opportunity. The next investment. The next move. But sometimes what matters most is being ready for what you never saw coming. A storm damages a second home. A pipe bursts above a collection you spent decades building. An unexpected loss puts a growing business on pause. Those are the days no one sees coming. But for more than 75 years, Cincinnati Insurance has been ready when they do. With deep expertise, coverage for businesses, homes, valuables, and more, Cincinnati Insurance works with independent agents who take the time to understand what matters to you. Together, they help protect more than what you own. They help protect what you're building toward. And when a bad day does come, Cincinnati Insurance has real people there to help make things right. Because planning for the future isn't only about knowing what's next, it's about making sure you're ready for what you can't predict. Let Cincinnati Insurance make your bad day better. Find an independent agent at CINFIN.com. Running a business means you're always on the move. The Delta SkyMiles
[30:00] Your business American Express card is made for frequent travelers who want premium benefits that works as hard as they do. For a limited time, new Delta Sky Miles Reserve Business card members can earn 200,000 bonus miles, the highest miles offered this year. And when you need to recharge or find a quiet place to work, the Delta Sky Club can turn a layover into precious productive time or a restful reset. You can enjoy 15 visits each medallion year to the Delta Sky Club when flying Delta and unlock unlimited Delta Sky Club access after spending $75,000 in purchases on your card during a calendar year. Plus, receive four one-time guest passes each medallion year so you can bring a colleague along. That's 15 Sky Club visits plus four guest passes for an annual value of up to $950. Value estimate based on $50 per visit rate to purchase additional visits or bring a guest to the Delta Sky Club. Apply today at www.delta.com slash podcast. Offer ends 11-426. Minimum spending requirements and terms apply. Your team is on a deadline, but moving all your enterprise data will take time your business simply doesn't have. IBM technology lets AI work with your data where it already lives, with an open hybrid cloud approach designed to build on your existing investments. It's another way IBM helps you put your data to work. Let's create smarter business. IBM. Visit ibm.com to learn more. I was actually, I was talking to someone when I was out in San Francisco recently, and I actually want to ask you about it. Thank you. East and West Coast views on the technology, maybe. But I was talking to someone in San Francisco who had this term. He he called it the frontier sink. And it's like, what are the industries that actually need Ph.D. level and beyond technology? And there's a few that we could come up with pretty easily. One of them is probably biotech. Right. You can probably never have too much intelligence when it comes to medicine. Another one actually is probably semiconductor design itself. You probably never be too smart, etc. for that. - Yeah. Maybe another one of them has to do with quant trading, where you can never be too smart. I'm not sure. Who knows? But it actually does get to be hard to think about. There might just be a lot of industry for which... there is a certain level of intelligence that is enough yeah i mean for the moment again do you remember when the i i like going back to the iphone as an analogy because it really was a technological revolution at the time and do you remember when we thought like peak iphone in 07 08 was when you had that app which had like the pint of beer in it right yes and you could drink and we thought that was so cool i didn't but no everyone is like oh this is so cool you could pretend i Small things. No? This looks like whimsy to me. I don't get this. Oh, yeah. Joe hates whimsy. I hate fun. But keep going. But I remember how, yo, why was everyone going crazy about the fact that you could, like, This is a very good point. And so and I just don't think we know what the future of AI really looks like. So yes, you're right. Today, we probably say we just don't need more than 100 IQ person sitting in our pocket, managing our emails, sending angry messages to the car rental company because they charged me $2,000 for a ding. I do think the consumer layer AI is going to be people ask me all the time like what will get you really constructive on ai again i said like when we figure out that killer consumer application layer because at the end of the day the consumer is everything it's the economy it's where to the point about revenues like enterprise revenues i just don't think is going to plug that gap really i just don't see how like we yeah no you said you're right the numbers i don't see it can probably get there eventually but it can't get there in the time frame that the market wants to see the roi And so like back to the point about Frontier, The consumer probably doesn't need Frontier for the time being. to do all of these things and so i one i'm my big mantra is is that one of the best technological changes that we experience as humans is actually not necessarily some new advanced tech it's it's changing the application of existing tech and so it's very possible that we take the 100 iq agentic workflow that still kind of becomes almost ubiquitous within the next 12 to 18 months and we really going to get the next upswing in AI. Can I just say, by the way, Tracy, last several days, my wife was traveling out of town. My kids are at a very easy age. It's not, you know, it's not a big challenge or anything to solo parent for a few days. But it happened to be a very busy schedule for them, birthday parties and a few other things. And it was very nice having ChatGPT being able to
[35:00] where should I be in the next hour? Where should I be the hour? Cause there were a lot of things going on and, um, It's very easy. It's very nice. I saw your tweet on this show. No, you saw the tweet. This is part of the problem. You tweet everything before you tell it to me. And so I can't react to it in real time. And now I'm really upset that you said that because you saw the tweet from my private alt. I did not tweet that from my main account. And now a bunch of people are going to know that I have an alt and they're going to ask me for it. You're not going to get it. So I wish you hadn't said that. Sorry. Sorry. I can't I can't differentiate the two. Can I can I can I jump in? Yeah, of course. So what's that worth to you a month? What? Well, just to have that ability, let's just say almost like an account executive or something or like an administrator. Yeah. Because to answer that question, it matters. It matters for the revenue profile, right? I mean, also, if you just had the right user interface on a calendar, this would be solved too, wouldn't it? If you know you have to be somewhere at 3 p.m., 4 p.m., 5 p.m.? Well, there's a lot of things that are in so many different emails. So, for example, my daughter goes to – I take her to play rehearsal. And some days she – it's from 1 to 6 in the afternoon on Sundays. But some days they don't do the full cast for the five hours. So somewhere in the email it says, oh, these students or these participants, they can leave at 3 and stuff. And it's scattered across so many emails. which ones they're in. And so being able to say, Chad, it's Sunday, October 25th or whatever. What time am I supposed to pick up my daughter? Three or six? That's very easy and nice to have. Fair enough. I just want to go back to the beginning of the conversation. You talked about neoclouds as the sort of tip of the spear and maybe being in the beginning of the denial phase. What do you think would spook investors? Like what should we look out for next in terms of a degradation in the AI bull belief. That's a great question. Off the top of my head... I would say it would be some level of capex being reined in. by one of the hyperscalers in fact on the walk over here i was asked by one of my clients who was saying whether questioning whether that would actually even come up potentially in the results this week and famous last words It's probably still premature. Joe, to answer your question about West Coast, East Coast, if you think about West Coast, the further you are closer to the epicenter of this technology, sorry, the closer you are to the epicenter, the more I would say that both the companies and investors are working. bought in on that narrative and there's definitely kind of almost like an echo chamber within that certain ecosystem you think about the the rooms that people are sitting in so we're we're um My suspicion is that it's not going to be the set of results and that maybe we have to wait till the new year that we get some kind of potential temperament of the enthusiasm. But like Google, Google kind of set this all off again for the next wave, because on the CapEx side, it was actually the last set of results. where they said it's going to be, I forget the exact wording they used, but they were talking about capex means significantly higher. And that got everyone enthused once again and got them all G'd up. Well, Oracle just said it's investor day that it was going to limit its additional fundraising to, I think, 20 billion through May. So we might be getting close to that. But again, like the reaction so far is this is a good thing and the build out can continue as long as people reasonable amount of capital, I guess. Yeah. And that kind of goes back to maybe just a slightly more holistic market view, which is this almost S&P does nothing because what's good for hyperscalers will be bad for memory. So you just get this constant violent rotation. And I call it like a knife fight in a phone booth, you know, for all these funds that have to run long, short portfolios. It's a nightmare. I'm glad you asked this because this is a question that I wanted to ask you. So volatility. We're seeing volatility everywhere except our measures of volatility. So if we look at realized volatility, it's pretty low. If we look at the VIX, it's not that high. Can you explain why it is that there's all this anxiety about certain parts of the most important trade in the entire world? And yet when we actually look at the way we measure volatility, it looks very benign. Yeah, I've got I've been asked. This is probably the most asked question. that I get in some way, shape or form. And there's, there's two parts of the answer, right? There's a structural part. And then there's a really like technical part. Yeah, let's hear it. All right. So the structural part is got to do with the fact with the way that money's deployed. these days so obviously equities are meant to be a long only product really
[40:00] But in this day and age, they're not, right? We've got... 2020. The AUM deployed in multi-manager funds was about $300 billion. And today it's in excess of a trillion dollars. right and if you think about the leverage yeah i mean not not just pods but obviously let's just think about like that sort of ilk yeah And it makes perfect sense, right? Because if, especially in a higher interest rate environment, high interest rate environment is perfect if you're running a long short strategy because I deploy money on my gross long side and on my gross short side. And actually I'm not really deploying I'm deploying units of risk, but the actual capital being deployed is really only what I'm actually posting as margin on my long and my short. Remember, I can leverage that too through the prime prime businesses. So actually what I'm doing is I'm taking a large chunk of capital and I can put that into, say, a high interest bearing business. a high interest instrument clip a very nice coupon and then i try and generate alpha on that gross capital deployed that's the business model and it's for the most part a very good business model right the problem is is that you've effectively tripled more than tripled the aum plus obviously leverage and the number of securities you can trade has stayed roughly the same or actually shrunk over time So that creates crowding out. So how do I organize or how do I mitigate that crowding out effect? Well, I've got all kind of cool risk parameters. So, you know, Barra, I'm sure you've heard, or Axioma, just, you know, all these different style factors, quality, value, size, growth, you know, all these different things. You would have heard us talk about them quite a lot. They've become ubiquitous. And obviously, funds spend an inordinate amount of time and computing power trying to neutralize those factors. So that's cool. Thank you. Now, The problem is, is that At the end of the day, everyone's still doing this, which means all you're creating is different kinds of crowding. And so that's why the volatility isn't showing up so much at an index level, right? Because ultimately a lot of this AUM is kind of index agnostic, so to speak. And you're seeing the violent factor rotations happening because ultimately you've got a huge amount of crowding sitting there. And the one thing that leads me into sort of the technical aspect, which is the one thing that I think a lot of funds have not yet been able to solve is the AI factor. Right. Yeah, because you could be everything either is or isn't AI now. Right. Exactly. And this is a problem. Right. It is sort of it's dawned on me recently because I've been asked a series of questions by investors specifically. Uh... Along the lines of, have you managed to map the AI factor? And the answer is no, because the AI factor is changing all of the time. The beaters are changing all of the time. Remember the beginning of the year, you had a lot of people that wanted to be short software. Yeah, that's right. Now software is trading close to 12 month highs. And so the idea of being short software, it was part of the AI factor. And now is it so much? Well, not necessarily unclear. So because the narrative is shifting, the factor of AI is shifting too. And that in turn complicates the ability to funds to actually try and hedge that. And bottom line is, is you're either long AI or you're short AI. And let's face it, the past three years, being short AI probably hasn't been a particularly lucrative trade. By the way, Tracy, in the time that Alex was giving us, that Alty was giving us this great answer. You sorted out your entire weekend schedule using ChatGPT. No, in that time, we just got earnings from Alphabet and Microsoft. And it's very interesting. Alphabet down 7.9% revenue XTAC misses estimates. Microsoft jumped 6.4% after Q1 cloud revenue Tops estimate. stocks they both have very violent reactions but if you're the indexed investor on net looks like it all kind of washes out yeah it does indeed but man some of these like these numbers are incredible microsoft q1 asure revenue x asure this is their cloud platform as you what do you think what's your guess you you're not looking at a terminal the way no so what do you think you'd get this is fun the bogey was i think 47 percent Growth? Yeah. Man, you're really good at your job. So the official estimate that we had collated at Bloomberg was 44.8%. They hit 48%. 48% growth for these gigantic companies. That's a good number. That's unbelievable. Cloud revenue at Google was $30 billion. I'm not saying what the growth number was. But these are unreal. Yeah. I mean, they are. And this goes back to the point about kind of AI winners and losers. Fantastic legacy business. And you're a cash volcano. From your operating. Legacy operating business. Then if you want to go. And embark upon. The largest. you know, CapEx experiment really in the history of America...
[45:00] Total non sequitur, by the way. I was trying to work out, because we're spending, if you believe the CapEx numbers in terms of the forward, and I'm sure that they're going to change from what the consensus looks like now. But on face value, it's about 12% of GDP. Right? It's like two Netherlands. Right. And I was thinking... Not in modern history, like the Louisiana Purchase, the railroads, it's blown all out of the water. What have we done as a human society where we've spent that much money? Certainly in a short period of time. And I had to start getting like really creative. I'm sure like Joe, you can you'll be awake at night thinking about this and trying to find like some historical work on it. Pyramids, maybe, but they took hundreds of years, if not thousands. Great Wall of China, maybe. I just can't I can't think of a period where we've spent. What do you think? How fast do you think Google Google Cloud revenue grew? Again, I'm turning my laptop away from you. I actually don't even – I can't even believe this number is real. I'm giving you a hint. 30-something percent? I don't really understand. I almost... They had 103% Google Cloud Revenue grew 103%. It seems great. I don't know. All these numbers are insane. Anyway, that's what it says. I mean, these are impressive businesses. And this is why a lot of people say S&P undefeated, right? By the way, funny, the concentration of the S&P, basically, given, Tracy, to your point, everything's kind of AI. Concentration of S&P, next year, 75% of your earnings growth in the S&P is going to come from memory, from semiconductors. What percent? Sorry, a bigger part. gets you to around about 75%. Wow. So you basically, S&P has become Cosby. Yeah. Oh, Cospification. That would be a good... Cospification, yeah. Yeah. All right, well, Alex, we're going to leave it there because we want to get this episode out pretty quickly. But thank you so much for coming back on All Thoughts. Really appreciate it. Thank you. Thank you so much. See you next week. Thank you so much. Joe, it was great catching up with Alex. I think he does a great job of sort of synthesizing a lot of different things going on in the market right now. But I do think the key thing that's emerging is this differentiation idea and the idea that we're moving from this sort of capital abundant environment into an area where people are perhaps a little bit more selective and what that means Thank you. You know, the classic is different. The time is, oh, interest rates don't matter anymore. Right. And this is a classic thing. Oh, this is not going to be a rate sensitive industry because the numbers are so unbelievably high. If you're growing 44 percent in a year, what do you care about? Well, the rates are five percent or five point two five. It's completely irrelevant. You're just going to keep building, etc. But that is a lie that investors tell themselves. you spear at least for some of these leveraged companies like the neoclouds or whatever uh For one thing, it looks like a range matter. Yeah. And also, I mean, just the costification idea of the market, it is really quite stunning because a lot of the earnings power of the tech companies, the hyperscalers is predicated on the fact that, well, other companies are going to buy the services so that they can boost their own productivity and get the earnings up. it won't just be about AI. This is the thing that, like, When you talked about the pod shops, for example, the whole premise of a pod shop is that everyone is supposed to find their own undifferentiated alpha. Yeah. Where is that right now? So I think basically if you're an investor – And you made money in the last year at scale. I mean, of course, like you might be able to like, oh, I opened, you know, a new like pasta restaurant in the East Village. That's a separate question. But if you're at scale and you made money in the last year, in some way it was connected to AI. You have all these different pods and some of them are chips and some of them are industrials, etc. But it's hard for me to believe that the ones who are making money aren't just in some way finding a way to belong to AI factor. long AI factor, I would imagine. The, uh, the AI factor, you know, I was talking to, um, It's so funny, you know, we try to... Alex has such a good read on the industry, but also it's great to talk to him because he talks to class at the Wall Street View and so forth. I was talking to this computer science student at Stanford recently, and they were,
[50:00] Wall Street is an AI pill. And I said, what are you talking about? Wall Street is an AI pill. Look at all these stocks that are going to the moon. The numbers on my screen tell a different story. What are you talking about? But... That person's view was that if you're even having a discussion about the possibility of overabundance of GPUs, you just don't get it. Like, what do you like? Oh, you don't you don't get A.I. sufficiently. It's so funny. It's yeah, definitely an interesting time in markets. Shall we leave it there? Alphabet... They do 32 billion tokens per minute. That was up 40% Q over Q. I don't even know what these words mean. Let's leave it there. All right. This has been another episode of the All Thoughts Podcast. I'm Tracy Allaway. You can follow me at Tracy Allaway. And I'm Jill Weisenthal. You can follow me at The Stalwart. Follow our producers, Carmen Rodriguez at CarmenArmand, Dashiell Bennett at Dashbot, Kale Brooks at Kale Brooks, and Kevin Lozano at Kevin Lloyd Lozano. You can share about all these topics 24-7 in our Discord. I shot up. For more Odd Lots content, check out our daily newsletter, which you can find at Bloomberg.com. And if you enjoyed this conversation, if you like it when we check in on the markets, then please leave us a positive review on your favorite podcast platform. And if you're watching on video, then please like or leave a comment or better yet, subscribe. And remember, if you're a Bloomberg subscriber, you can listen to all of our episodes absolutely ad free. All you need to do is find the Bloomberg channel on Apple Podcasts and find the instructions there. Thanks for joining. Thank you. empower is all about helping you invest well so you can go out and live a little and their free empower personal dashboard shows you your full financial life in one place You'll be able to see your net worth, monitor budgets, analyze your portfolio, set retirement goals, and more. Whether you're saving for something big, want to treat yourself, friends or family, or meet long-term goals, Empower helps you see where you stand. You've worked hard for your money, and Empower can help you make the most of it. Download the Empower Personal dashboard or visit Empower.com. Not an Empower client paid or sponsored. This is Troy Millings from Earn Your Leisure. If you run a small business, you know how important it is to stay on top of your numbers. That's why I want to tell you about Xero. That's X-E-R-O. Xero brings accounting, payments, payroll, and analytics together in one platform, giving you a clear picture of your business. With reports, visual dashboards, forecasts, and AI insights, Xero can help you better understand your cash flow and what's coming in and going out. It's a simple but powerful platform built for small business owners and designed to make managing your business finances easier from day one. Join 5 million customers worldwide who use Xero to manage their finances. Learn more at Xero.com. With ChatGPT, you can do more than get answers. You can get things done. Whether you're leading a big project, planning an elaborate trip, or trying to balance work and life, ChatGPT can help. Give it a goal, plus things like files, notes, or project plans, develop a trip itinerary with daily updates, or create a prioritized schedule, so you can spend less energy sorting out the details and more on bringing your biggest ideas to life. Get started at ChatGPT.com. Dear America, Toyota's save bucks on trucks event is on. Haul it all with a powerful Tundra, or play hard in the rugged Tacoma. Toyota trucks are built for weekends that don't end when the pavement does. Boat ramps, campsites, trailheads, and everything in between. So don't wait. Haul home heavy duty savings and save bucks on trucks with Toyota. Toyota, let's go places.
