Executive Summary
After two years of infrastructure arms race dominated by foundation model development, AI is entering what we call the "deep water zone"—a phase where value creation shifts decisively from the model layer to the application layer. The winners of the next decade won't be those with the biggest models, but those who build AI-native applications that fundamentally reimagine how work gets done.
Our analysis shows that while $100B+ has been invested in AI infrastructure and foundation models, less than 5% has gone to application layer companies. This imbalance is about to reverse dramatically. We predict the AI application layer will create $1 trillion in market value by 2030, with early movers capturing disproportionate returns.
The Evolution of AI: From Research to Revenue
AI Development Phases
The AI Stack: Where Value Accrues
Understanding the AI Value Chain
Chips & Infrastructure
NVIDIA dominates with 90%+ market share. Limited investment opportunities due to high barriers and winner-take-all dynamics.
Foundation Models
Commoditizing rapidly. Open-source models approaching proprietary performance. Differentiation increasingly difficult.
AI-Native Applications
Massive opportunity. 99% of value creation ahead. Winners will build complete workflows, not features.
Why Applications Win: The Four Laws of AI Value Creation
1. The Workflow Law
AI features are commodities; AI workflows are moats. Companies that reimagine entire processes—not just add chatbots—will capture the value. Example: An AI that writes emails is a feature. An AI that manages your entire communication workflow, prioritizes responses, schedules meetings, and maintains relationships is a product.
2. The Data Compound Law
Application companies accumulate proprietary data through usage, creating a compounding advantage. Every interaction improves the product. Foundation model companies train once on public data; application companies train continuously on private, task-specific data.
3. The Distribution Law
In AI, distribution beats technology. The best model with no users loses to a good-enough model with millions of users. Application companies own the customer relationship and can swap underlying models as needed.
4. The Verticalization Law
Horizontal AI tools will lose to vertical AI solutions. A generic AI assistant is less valuable than an AI specifically trained for legal contracts, medical diagnosis, or financial analysis. Depth beats breadth.
The AI-Native Application Opportunity Map
Knowledge Work
AI analysts, researchers, and decision support systems that augment human intelligence
Software Development
AI pair programmers, automated testing, and code generation platforms
Creative Industries
AI-powered design tools, content creation, and creative collaboration platforms
Healthcare
Diagnostic AI, treatment planning, and personalized medicine applications
Education
Personalized tutors, adaptive learning systems, and skill assessment platforms
Financial Services
AI advisors, risk assessment, fraud detection, and automated trading systems
What Makes an AI Application Investment-Grade?
Our AI Application Investment Framework
10x Better, Not 10% Better
Proprietary Data Moat
Workflow Integration
Network Effects Potential
The Risks: Navigating the AI Deep Waters
1. The Commoditization Trap
As foundation models improve, features that seem defensible today may become commoditized tomorrow. Companies must build deep moats through data, workflows, and network effects—not just AI capabilities.
2. The Incumbent Awakening
Large enterprises are beginning to integrate AI aggressively. Startups need to move fast and focus on areas where incumbents' innovator's dilemma creates opportunities.
3. The Regulation Wave
AI regulation is coming. Companies building in sensitive areas (healthcare, finance, hiring) must design with compliance in mind from day one.
4. The Talent War
Competition for AI talent is fierce. Companies need compelling missions and equity packages to attract top researchers and engineers.
Our Investment Strategy for the AI Application Era
- Vertical Focus: We're prioritizing vertical AI applications over horizontal tools
- Workflow Transformation: Looking for companies that reimagine entire processes, not add AI features
- Data Accumulation: Investing in applications that get smarter with every user interaction
- Distribution Advantages: Backing teams with unique go-to-market strategies or existing customer relationships
- Regulatory Awareness: Partnering with founders who understand the coming regulatory landscape
The Next Five Years: Our Predictions
By 2030, we believe:
- 1 AI applications will be a $1 trillion market, larger than the current SaaS market
- 2 Every knowledge worker will use 5+ AI applications daily
- 3 AI-native companies will displace 30% of current software leaders
- 4 Vertical AI solutions will outperform horizontal platforms 10:1
- 5 The best AI applications will be indistinguishable from human experts
- 6 New job categories will emerge to work alongside AI systems
- 7 AI regulation will create moats for compliant early movers
Final Thoughts
We stand at an inflection point. The infrastructure has been built. The models are capable. The market is ready. Now comes the hard work of building applications that deliver real value to real users solving real problems. The companies that succeed won't be those with the best technology, but those with the deepest understanding of user needs and the ability to reimagine entire workflows around AI capabilities. The deep water zone is where simple demos become complex products, where features become platforms, and where the real value in AI will be created. The time to dive in is now.
Building the Future of AI Applications?
We're actively investing in AI-native applications that transform industries. Let's explore how we can help you scale.
Connect With Us