Watch the full broadcast here: Where is the value in AI?
We’re in the midst of an AI infrastructure boom. Companies like NVIDIA dominate today’s spending, much like Cisco and Dell did during the cloud’s early days. But history is repeating itself: just as SaaS outpaced cloud infrastructure, AI applications will soon dwarf the current hardware frenzy. Here’s why the real value lies ahead—and where to find it.
The Shift from Infrastructure to Applications
Today, 85% of AI investments flow into infrastructure. But as costs plummet (token prices dropped 97% in 2 years, with recent models slashing expenses further), the door opens for scalable, high-impact applications. In 5–10 years, AI apps could be 3–10x larger than infrastructure. The key? Solving previously unsolvable problems.
Where AI Applications Will Thrive
- Unlocking Unstructured Data
AI excels at synthesizing fragmented data—PDFs, emails, city council meeting transcripts—into actionable insights. Example: Savi uses AI to parse local governments’ purchasing data from 180k+ sources, creating a trillion-dollar procurement database once deemed too costly to build. - Automating Repetitive Work
From call centers to sales coaching, AI handles repetitive tasks faster and cheaper. Grow AI cut sales coaching costs by 95% while outperforming human consultants by leveraging real-time context from emails, calls, and CRM data. - Vertical-Specific Solutions
Generic AI tools won’t dominate industries like agriculture or healthcare. Newell AI optimizes fertilizer supply chains using weather data and logistics APIs—use cases too niche for broad models like ChatGPT. - Voice as the New UI
Voice interfaces (e.g., Domino’s pizza ordering) are replacing clunky apps for tasks where speaking is faster than typing. - Democratizing Startup Growth
AI “copilots” accelerate development, letting founders build faster and cheaper—mirroring how cloud computing slashed startup costs from $5M to $100.
What Won’t Work
- Thin Wrappers: Basic prompt engineering atop existing models lacks defensibility.
- Middleware: Providers like OpenAI will absorb tools (e.g., voice transcription) into their platforms.
- AI as a Feature: Adding AI to existing SaaS apps invites disruption by incumbents (e.g., Zoom crushing third-party transcription tools).
- Generic Generative AI: Most image/text generators lack clear monetization beyond novelty.
The Bottom Line
The winners will solve previously intractable problems by leveraging AI’s unique capabilities—not just adding “AI” to old ideas. Focus on:
- Vertical expertise (industry-specific data and workflows),
- Proprietary data integrations,
- Clear ROI (e.g., doubling sales quotas or unlocking new revenue streams).
As infrastructure costs near zero, the battleground shifts to applications. The time to build is now.
— Inspired by insights from Martin Casado, Andreessen Horowitz
TL;DR: AI’s infrastructure phase is peaking. The trillion-dollar opportunity lies in vertical apps that solve hard problems, automate workflows, and democratize access—not in wrappers or features. Build where others can’t follow.*