Just how frothy is the AI Bubble anyway? | E2195

Key Concepts

• AI Bubble Debate: The ongoing discussion about whether the current surge in AI investment constitutes a speculative bubble.
• Economists' Predictions: The historical tendency of economists to predict more recessions than actually occur.
• Total Addressable Market (TAM) for AI: The potential revenue that can be generated by AI technologies, encompassing software and labor replacement.
• Financial Manipulation: The recurring theme of finance professionals exacerbating market downturns through speculative practices.
• Wealth Effect: The phenomenon where a decline in stock prices reduces consumer spending due to decreased perceived wealth.
• Multi-Cloud Strategy: The practice of utilizing services from multiple cloud providers to mitigate risks associated with single-provider outages.
• AI-Generated Content and Watermarking: The challenge of identifying and authenticating AI-generated images and videos, and the effectiveness of watermarking.
• Post-Truth Era: The current cultural climate where factual accuracy is often secondary to ideological alignment and narrative reinforcement.
• AI Trading Bots: The development and performance of AI models designed to trade in financial markets, particularly cryptocurrencies.
• "The Social Reckoning": The upcoming sequel to "The Social Network," focusing on Facebook's internal issues and featuring Jeremy Strong as an aged Mark Zuckerberg.

AI Bubble and Market Dynamics

The discussion begins with the prevalent debate surrounding an "AI bubble." Economists Jared Bernstein and Ryan Cummings, former members of the Council of Economic Advisers, are cited for their op-ed "Warning! Our stock market is looking like a bubble." Their core arguments include:

• Lofty AI Investment Plans: They find OpenAI's projected $1 trillion spend to be disproportionate to its current revenue.
• Overvalued Stocks: Historical price-to-earnings ratios suggest current stock valuations are too high, with AI being a primary driver of these gains.
• Wealth Effect Risk: If the market corrects, wealthy individuals, who own a significant portion of stocks, will have less disposable income, potentially leading to a broader economic impact.

The hosts acknowledge the accuracy of economists' predictions, humorously noting they "predicted 11 of the last seven recessions and corrections," implying that corrections are inevitable.

Total Addressable Market (TAM) for AI

A counterpoint to the bubble narrative is presented by analyzing the TAM for AI. The current revenue for AI companies is approaching $100 billion. Projections suggest that within the short to midterm (1-5 years), the TAM could reach $1-2 trillion in revenue, with an estimated consistent annual growth of $50 billion for several years. This TAM is not limited to the software industry but also includes the cost of labor that AI automation aims to replace, effectively encompassing a significant portion of the economy.

Historical Bubble Bursts and Finance's Role

The conversation delves into past market corrections, identifying a pattern where financial institutions play a significant role in exacerbating downturns:

• Great Recession (2008): Attributed to financial manipulation by Wall Street, not tech or real estate directly.
• Dot-com Era: Also driven by financial manipulation, with Wall Street taking companies public while simultaneously devaluing them internally. The case of Henry Blodget is mentioned as an example.
• Silicon Valley Bank (SVB) and First Republic Bank Collapse (2023): Caused by a finance-driven decision to reprice debt amidst changing interest rates, leading to a confidence crisis and bank run.

The argument is made that finance professionals, driven by fees from deals, often "f everything up" by accelerating trends too far, akin to "Wolf of Wall Street" scenarios. The question is raised whether this pattern is repeating in the current AI boom.

Tiger Global and "Finance Genre" Investors

The role of large investment firms like Tiger Global in the 2021 venture capital boom is discussed. The distinction is made between genuine capital allocators and those in the "finance genre" who are primarily focused on making quick money. The influx of capital with "big chip stacks" and rapid deal-making is seen as a potential indicator of a market top. Speculative activities like "round-tripping" and aggressive financial maneuvers by these entities raise red flags.

Sam Altman's Strategy and Unnatural Financial Acts

Sam Altman's strategy of raising substantial funds for AI infrastructure, including deals with AMD and Broadcom, is highlighted as a potential sign of a building bubble. The scale of these investments, particularly the $1 trillion target for infrastructure, coupled with the "round-tripping nature" of some deals, is described as "unnatural financial acts."

Consumer vs. Enterprise AI Adoption

A key point of contention regarding the AI bubble perception is the lack of widespread consumer adoption and payment for AI applications. While ChatGPT has 800 million global users, only 5% are paying. The current AI boom is largely fueled by corporate and enterprise spending. This disconnect leads to a public perception that AI is a "dead technology" because everyday users aren't paying for it.

Corporate Spend and Productivity Gains

In contrast, the corporate spend on AI is deemed "very real." Corporations invest when there's a clear path to ROI. Evidence suggests AI is making professionals like CPAs 10-20% more efficient. Startups are seeing low double-digit productivity gains for their customers within 3-6 months. This real-world impact on the ground is acknowledged, but the concern remains that finance professionals are accelerating this adoption beyond its natural pace.

The Greatest Investment Cycle vs. Frothy Moments

A strong conviction is expressed that this AI cycle will be the "greatest investment cycle in the history of technology," dwarfing previous revolutions (PC, smartphone, internet). However, this certainty is tempered by the acknowledgment that some investments might be poorly timed or overvalued. The potential downside is not an economic crash but a delay in technology adoption by 3-4 years, leading to companies pausing or cutting aggressive spending. The analogy of the fiber optic build-out is used, where companies over-invested and then scaled back. The current situation is described as "frothy," like a "cappuccino," expected to become a "flat white" with less foam.

Trigger Points for a Market Correction

The potential trigger for a market correction is identified as a significant failure in a major data center project or a drastic valuation drop for companies heavily involved in data center build-outs. Companies like CoreWeave, with substantial debt and ongoing projects, could face pressure to pause or sell assets if they struggle with loan repayments. This would lead investors to re-evaluate their price-to-sales ratios (currently 14.81 for CoreWeave) and debt levels. A slowdown in hyperscaler spending, such as Meta's potential reduction in its $600 billion investment projection, could also impact companies like Nvidia. Concerns about energy availability for data centers are also mentioned as a potential bottleneck.

Cloud Infrastructure and Outages

AWS Outage and its Impact

A significant outage in AWS's US East-1 region is discussed, affecting major services like Coinbase, Robinhood, Snap, and Signal. This event highlights the critical reliance on cloud providers and the potential for widespread disruption.

Founder Strategies for Cloud Outages

Founders are advised to consider multi-cloud strategies to mitigate risks associated with single-provider outages. While implementing multi-cloud can be complex and costly, it offers resilience. For early-stage startups, the focus should be on backups rather than full multi-cloud infrastructure, which requires dedicated teams. The question of whether any affected services successfully failed over to other clouds is raised.

Decentralization and Single Points of Failure

The AWS outage serves as a reminder that even seemingly decentralized systems can have single points of failure. While the internet itself is designed for decentralization, specific services tied to a single cloud provider are vulnerable. The US East-1 region's importance is explained by its role as a routing hub for many other AWS services, making its failure have a cascading effect.

Acceptable Outage Frequency

The frequency of major cloud outages is discussed, with an average of once every six months across all providers. Amazon's outages are noted to occur roughly every two years. This frequency is considered "acceptable" given the rapid pace of infrastructure build-out and increasing utilization, especially in the context of AI development. The electrical grid's reliability is presented as a greater concern than data center outages.

AI-Generated Content and the "Post-Truth" Era

AI Image Generation and Watermarking Issues

The widespread sharing of an AI-generated image on Blue Sky, depicting protests, is used as a case study for the challenges of AI content authenticity. The image contained numerous tells of AI generation (bent flagpoles, misaligned flags, odd creature poses, illegible signs), yet many users, including a friend of the pod, shared it without realizing it was AI-generated. The presence of a Gemini watermark was largely ineffective.

The Ineffectiveness of Watermarks Against Open Source AI

The debate arises whether watermarks are a viable solution, especially with powerful open-source AI models available. The concern is that these models can replicate the capabilities of proprietary ones, making it difficult to track origins.

"Disclosure" and the Foresight of Deepfakes

Michael Crichton's 1994 novel "Disclosure" is referenced for its prescient exploration of manipulated video evidence and its potential to alter reality and cause significant harm, including job loss and societal unrest. This highlights that the issue of fabricated visual media is not new, but the technology has advanced dramatically.

The Erosion of Trust in Social Media and the Rise of Fact-Checking

The conclusion is that trust in social media will continue to erode, and rightfully so. The value of established journalistic institutions with insurance and accountability will increase. The need for real-time fact-checking services, akin to Snopes or community notes, is emphasized. The tendency for individuals to retweet content that aligns with their narrative, regardless of its veracity, is identified as a major problem.

The "Slop Issue" and Incentives for Misinformation

The "slop issue" is described as the problem of people sharing unverified or misleading content, often driven by incentives like follower growth and revenue sharing on social media platforms. The focus on trending content over accuracy perpetuates the spread of misinformation.

USAFacts.org as a Ground Truth Resource

USAFacts.org is presented as a non-partisan resource founded by Steve Ballmer, providing reliable data on government spending and other topics. While valuable, its reliance on government data means it's still dependent on an external authority for ground truth.

The "Post-Truth Senior Voting Block"

A speculative concern is raised about the potential for a "post-truth senior voting block" due to the challenges of discerning truth in the current media landscape.

AI in Finance and Trading

AI Trading Bots and Market Beaters

The pursuit of an AI that can consistently beat the market is a long-standing goal in finance. A new game by "in of one" pits leading AI models against each other in a crypto perpetuals trading challenge with a $10,000 starting capital.

Performance of Leading AI Models in Trading

Initial results (after a couple of days) show GPT-5, Claude Sonnet 4.5, Gemini 2.5 Pro, Grok 4, DeepSeek Chat 3.1, and Quinn 3 Max competing against Bitcoin's performance. Grok 4, DeepSeek 3.1, and Sonnet 4.5 are currently in the black and outperforming Bitcoin, while Google, OpenAI, and Quinn are losing money.

The Future of AI Trading Challenges

The potential for users to enter their own AI trading "horses" with proprietary instruction sets is discussed, creating a competitive race to double money within a time-bound challenge. This incentivizes risk-taking and highlights the importance of prompt engineering and data context.

Entertainment and Media

"The Social Reckoning" and Jeremy Strong's Portrayal of Zuckerberg

The upcoming sequel to "The Social Network," titled "The Social Reckoning," is announced. Aaron Sorkin is directing, and Jeremy Strong will star as an aged Mark Zuckerberg. Strong's dedication to character, even on the red carpet, is noted, with his interview responses mimicking Zuckerberg's speech patterns.

Jeremy Strong's Acting Methodology

The discussion touches on Jeremy Strong's intense acting method, which can make him challenging to work with, as he often stays in character. This contrasts with actors who can easily switch between character and their personal lives.

Jeremy Allen White's Role in "The Social Reckoning"

Jeremy Allen White will play a Wall Street Journal reporter named Jeff Horowitz, who teams up with a whistleblower (played by Mikey Madison) to expose Facebook's secrets.

Aaron Sorkin's Dialogue Style

Aaron Sorkin's signature dialogue, characterized by its theatricality and heightened realism, is discussed. While praised for its intelligence, it's noted that conversations rarely reach such dramatic heights in real life. The comparison is made to "The West Wing," which presents an aspirational version of Washington.

Sponsorships and Calls to Action

The video includes several sponsored segments:

• Quo (formerly OpenPhone): A business phone system offering streamlined customer communications, with a free trial and 20% off the first six months.
• Vanta: A compliance and security platform that simplifies SOC 2 reports and other compliance needs, offering $1,000 off for a limited time.
• LinkedIn Ads: A platform for B2B advertising, offering a $250 credit for new campaigns.
• Claude.ai/twist: A discount for Claude Pro, an AI assistant.

The hosts also encourage listeners to leave reviews on Apple Podcasts and Spotify, offering shout-outs for thoughtful comments.