Focus your AI portfolio on "Wave Two" pick-and-shovel infrastructure plays that benefit from ecosystem growth without taking on single-model risk. Target private or upcoming public market opportunities in intelligent API routing layers like OpenRouter, which optimize enterprise AI spending across multiple models. Invest in high-growth AI application layer companies like Lovable that bridge the gap for non-technical users while maintaining distinct product differentiation. Prioritize agile foundational model providers such as Anthropic that demonstrate capital efficiency and multi-cloud distribution partnerships over bloated neo-labs. Exercise caution by avoiding undifferentiated foundation models, as over 60 funded neo-labs face inevitable industry consolidation.

Anthropic (Private)

• Led by venture capitalist Matt Murphy at Menlo, Anthropic was recognized early on for its performant models and technical efficiency, requiring a fraction of the capital of competitors while matching or exceeding their performance benchmarks.
• The company successfully positioned itself as a major multi-cloud provider through partnerships with Amazon and Google, utilizing Bedrock and Vertex for distribution and technical collaboration.
• Anthropic operates in a massive and expanding market where major tech players utilize its models to increase customer retention and drive platform engagement.
• Despite rising costs of computing inference, Anthropic maintains high performance standards, allowing for a family of models (Sonnet, Opus, Fable) that cater to different use cases.

Takeaways

• For private market investors, backing foundational AI models early requires high risk tolerance regarding dilution and high initial valuations, but can yield outsized outlier returns.
• Investors should monitor how foundational model providers balance high-end reasoning capabilities with cost-efficiency to defend against open-source alternatives.

Lovable (Private)

• Highlighted as a standout AI application company going from zero to hundreds of millions in ARR rapidly, focusing on empowering non-programmers to become creators.
• The company leverages the broader AI ecosystem, utilizing multiple models including open-source options to optimize its margin structure, which is projected to reach healthy gross margins over time.

Takeaways

• Application layer companies that bridge the gap between complex AI models and non-technical end-users represent some of the highest-growth opportunities in the current tech landscape.
• Investors should look for application companies that maintain clear product differentiation rather than relying solely on underlying model capabilities.

OpenRouter (Private)

• Described as an intelligent layer that intercepts API calls from applications to determine the best model based on efficiency, price, reasoning, and performance.
• The company has experienced massive organic growth from developers seeking a reliable inference marketplace to optimize their AI spend.
• Highlighted as being wildly profitable at a scale that shocks most people, acting as a core infrastructure beneficiary of the multi-model AI era.

Takeaways

• Infrastructure and routing tools that help enterprises optimize costs and select the best model among multiple options stand to benefit significantly as companies mature past the initial "test phase" of AI adoption.

AI Infrastructure & Developer Stack (Sector Theme)

• Matt Murphy notes that the market is moving from "Wave One" (getting AI running simply) to "Wave Two" (sophisticated multi-model routing, cost optimization, and developer tooling).
• Infrastructure plays such as agent frameworks, observability solutions, and specialized developer stacks are seeing renewed demand as enterprises look to manage inference spend and navigate multi-cloud environments.
• Neo-labs and specialized foundation model companies face intense competition, with over 60 neo-labs currently funded, suggesting a forthcoming consolidation where only domain-specific or exceptionally differentiated players will survive.

Takeaways

• Investors looking at the AI sector should look beyond headline foundation models and consider pick-and-shovel infrastructure plays—such as model routers, observability platforms, and specialized vertical models—which benefit from overall ecosystem growth without taking on single-model risk.