• NVIDIA (NVDA) has strong near-term chip demand, but weigh that against potential power and cooling bottlenecks and the risk that AI hardware needs may evolve.
• Favor AI companies with useful, specialized products, customer access, and proprietary feedback data; ElevenLabs is cited as an example, but it is private and no investment terms are provided.
• Be selective with SaaS stocks such as HubSpot (HUBS): AI could make some products easier to replace, while enterprise complexity and differentiated data may offer protection.

Artificial Intelligence (AI)

• Ben Horowitz sees AI as a major technology shift—potentially bigger than the internet—and says falling software-development barriers could let people with deep knowledge of a problem turn ideas into products.
• He describes rapid consumer adoption and improving technology, but says investor sentiment can swing between overexcitement and pessimism.
• He expects substantial opportunity as AI infrastructure, models, and supporting systems evolve. He says much of today’s infrastructure may need to be rebuilt, similar to the shift to cloud computing.
• He cautions that a temporary glut of GPUs could emerge if constraints such as power or cooling limit deployment.

Takeaways

• Look for businesses that solve specific problems and can use AI to build useful products—not just companies that invoke AI as a theme.
• The opportunity may extend beyond AI applications to infrastructure and model development, but the pace of change and potential infrastructure bottlenecks make the sector risky.
• Horowitz does not say that AI is currently in a bubble; he does warn that one could develop.

NVIDIA (NVDA)

• Horowitz says demand for NVIDIA chips is currently “unlimited” and that the company cannot make enough chips.
• He rejects the claim that chip demand is merely artificial because AI companies invest in one another.
• At the same time, he questions whether today’s GPUs—originally designed for video games—are the right long-term infrastructure for AI. He also points to possible power and cooling constraints.

Takeaways

• The discussion is positive on near-term demand for NVIDIA chips, but raises questions about whether current hardware and infrastructure will remain the best fit as AI workloads evolve.
• Consider both demand growth and the risk of changing technology or a temporary GPU supply glut; the podcast gives no price target or buy recommendation.

ElevenLabs (Private)

• Horowitz cites ElevenLabs as an example of an AI company that may be more defensible than expected, even if large technology companies develop similar capabilities.
• He says the company built a strong developer product and entered the developer stack, helping it attract customers and gather real-world data to improve its models.
• He also points to product features such as expressive, multilingual speech, and mentions deals with Netflix and Televisa as evidence of commercial traction.

Takeaways

• The discussion favors AI companies that combine a useful product with customer access, proprietary feedback data, and specialized capabilities.
• Horowitz’s example suggests that a model’s basic capability may be less important than how well a company serves specific use cases. ElevenLabs is private, and the podcast gives no valuation or investment terms.

Software-as-a-Service (SaaS), Including HubSpot (HUBS) and Affinity

• Horowitz says software and SaaS companies are trading at low multiples amid concern that customers may use AI to build their own software.
• The host describes using AI tools to build a personal CRM instead of HubSpot and a fundraising platform instead of Affinity. This illustrates the possibility of customers replacing some software with custom-built tools.
• Horowitz notes that an individual’s needs differ from those of a large enterprise, where replacing complex systems may be more difficult.

Takeaways

• The discussion is cautious on SaaS businesses whose products customers could readily recreate for their own needs.
• When assessing a software company, consider how difficult its product is to replace and whether it offers advantages beyond basic functionality, such as valuable data or enterprise-scale capabilities.
• The transcript does not recommend selling HubSpot, Affinity, or SaaS stocks; it describes a sector-wide concern and the potential for company-specific differences.

AI Companies and Competition: OpenAI, Anthropic, and Other Model Developers

• Horowitz says OpenAI reached $10 billion in revenue and describes another major AI effort led by Elon Musk as approaching product parity in a few years; the transcript does not name that effort.
• He mentions Anthropic in the context of providing computing resources, and says AI-model and infrastructure development may change significantly.
• He argues that falling computing costs and the ability to spend heavily on a problem can accelerate competition, while a company’s data and focus on valuable use cases may help it stand out.

Takeaways

• The discussion suggests that AI leaders face rapid competition and that current product advantages may not last.
• For AI companies, consider whether they have differentiated data, distribution, or customer-specific capabilities—not just access to computing power.
• These are private companies or businesses discussed without investment terms; no valuation, target, or recommendation is provided.

Cryptocurrency and Crypto-Related Businesses

• Horowitz says the Biden administration created difficulties for crypto entrepreneurs by restricting access to banking and taking enforcement actions while, in his view, failing to provide clear rules.
• He says a16z sought a relationship with the incoming administration and wanted clearer regulation for the industry.
• No specific cryptocurrency, token, or crypto investment is discussed.

Takeaways

• The main investment insight is regulatory risk: policy and access to financial services can materially affect crypto businesses.
• The podcast offers no recommendation on buying cryptocurrency or crypto-related stocks.