20VC: The AI Bubble Will Burst: Half the Neoclouds Will Die |...

Host discusses Jerry's prediction of an AI bubble bursting between October 2026 and March 2027 due to potential global conflicts and credit market disruptions.

Jerry explains his concerns about credit market disruptions, referencing past cycles and the current complacency despite red flags.

Jerry identifies complacency as a primary sign of cracking credit markets, drawing parallels to the 2008 crisis.

Jerry discusses Japan's economic situation and its holdings of US treasuries as a potential global risk factor.

Jerry elaborates on how prior credit market cycles differed, emphasizing the impact of debt on asset value during dislocations.

Jerry discusses how hyperscalers are better positioned to survive dislocations, leading to potential acquisition opportunities.

Jerry explains his prediction that at least half of neoclouds will disappear within 36 months, exacerbated by economic disruptions.

Jerry discusses the factors that will separate successful neoclouds from those that fail, emphasizing leadership and operational efficiency.

Jerry compares inference providers, favoring Fireworks over Base 10 due to capital efficiency and profitability.

Jerry discusses the future of AI models, suggesting a world of specialized models alongside frontier providers, driven by customization and cost efficiency of open-source models.

Jerry addresses the divergence between token traffic towards open models and dollar traffic towards frontier models, suggesting frontier models will continue to lead in innovation but open-source will fill the cost-efficiency gap.

Jerry disagrees with the notion that "a token is a token," arguing that customization changes a token's value and efficiency.

Jerry discusses how specialization and customization of open-source models can offer more efficient solutions for specific tasks compared to frontier models.

Jerry explains why the cannibalization of frontier models by open source is unlikely in the short term due to continuous innovation in frontier models.

Jerry discusses Alex Karp's perspective on enterprise customers' fear of working with frontier model providers, emphasizing data security and discernment.

Jerry discusses the increasing cybersecurity threats and the underestimation of security needs, particularly the importance of sandboxes.

Jerry explains his investment thesis, favoring companies with unique teams and a strong sense of imperative, like Fireworks in the AI inference space.

Jerry discusses the current generation of AI companies accepting lower margins as a land grab strategy, cautioning against building a culture around low profitability.

Jerry argues that while infrastructure companies can scale rapidly, app-layer companies are not yet compelling investments in the current AI landscape.

Jerry emphasizes the unprecedented nature of frontier model companies like Anthropic and OpenAI, comparing their achievements to foundational inventions.

Jerry believes that only a few neoclouds will dominate, with many others burning through capital and failing.

Jerry questions the long-term viability of model routing layers like OpenRouter, predicting disruption from direct inference exchanges.

Jerry advises investors to focus on impact and founders with a deep commitment to their business, rather than just chasing hype.

Jerry highlights companies like Fireworks and E2B as examples of founders with a strong imperative to build their businesses.

Jerry discusses liquidity timelines and how events like Cursor's sale to Elon Musk are exceptional, not the norm.

Jerry explains why Airtable's inability to IPO doesn't concern him as much as the need for SaaS companies to integrate AI effectively to remain relevant.

Jerry describes the current trend as a "co-work era" driven by autonomous agents, posing a threat to traditional SaaS companies.

Jerry expresses concern about PE firms heavily leveraged in the current market, especially if a financial dislocation occurs.

Jerry discusses the potential impact of market drawdowns on levered assets and the difficulties in handling asset deflation.

Jerry questions the necessity of government ownership in AI development, citing historical precedents and the current scale of companies like OpenAI.

Jerry suggests a need for a clear, strategized approach to technology policy, rather than arbitrary regulation.

Jerry believes concerns about backdoors in Chinese open-source models are overestimated, as current models will likely be obsolete within a decade.

Jerry discusses the concept of continuous learning models and their potential to replace current AI architectures.

Jerry explains that continuous learning models will require new architectures and training methods, fundamentally differing from current static models.

Jerry emphasizes the dynamic and context-dependent value of data, suggesting specialized use cases for enterprise data.

Jerry agrees with the thesis of specialized models for companies, requiring additional data for training and opening up significant market opportunities.

Jerry discusses the need for diversity of intelligence in AI to tackle complex creative problems beyond simple task execution.

Jerry believes continuous learning models are still years away, comparing the progress to the long-term development in cancer research.

Jerry predicts that blockchain will find significant utility in agent payments and other transactional applications, moving beyond its current state.

Jerry expresses confusion about NVIDIA's flat stock price despite strong numbers, attributing it to market plateaus and obfuscated growth due to hyperscaler activities.

Jerry chooses to "marry" Meta, Google, and Microsoft for the long term, citing their massive user bases as stable buffers.

Jerry explains his rationale for holding Microsoft, emphasizing its control over enterprise communication and its stable, monetizable business.

Jerry believes the consumer base is the primary factor keeping Meta, Google, and Microsoft afloat.

Jerry identifies Meta as the most likely "boring" Mag7 company, potentially becoming like a utility, but still stable due to its ad business and user communication control.

Jerry believes the Mag7 companies are safe havens, expected to benefit from positive market cycles over the long term.

Jerry states that Apple's AI strategy is hard to assess, depending on whether they are active innovators or passive consumers of technology.

Jerry believes his firm, Insight, is well-positioned for the future due to its small PE portfolio and intelligent capital deployment.

Jerry highlights firms like Menlo, Benchmark, and Khosla Ventures as venture investors who have fared well in the AI transition.

Jerry predicts that blockchain for agent payments will be a significant, obvious innovation in five years, despite its current "valley of disillusionment."