Conference Presentation, Keynote, Panel
AI's Trillion-Dollar Opportunity: Sequoia AI Ascent 2025 Keynote
- The total addressable market for AI services is projected to expand massively over the next 10 to 20 years, driven by a transition occurring faster than prior waves with the software and services profit pools significantly up for grabs.
- AI adoption faces no barriers to entry due to fundamental changes in distribution physics, with a "tremendous sucking sound" in the market expected to overwhelm macroeconomic volatility and a "run-like-heck" business velocity requirement.
- The opportunity at the application layer remains wide open despite some consolidation, as value is predicted to ultimately accrue there rather than at the foundation model level, which lacks the ability to secure deep customer relationships.
- Startup success hinges on 95% of effort devoted to building the company itself, solving important problems, securing top talent, and earning customer trust, which is currently considered more critical than product features alone.
- Token costs have decreased by 99% over the last 12 to 18 months with a continued downward trajectory, while companies shifting from selling tools to selling outcomes can expect price point increases and a path to healthy gross margins over time.
- A data flywheel requires a direct tie to a business metric to be valid, while "vibe revenue" lacking durable behavior change poses a significant risk to long-term viability.
- The market is moving toward "agents first" architectures, with 2025 characterized by vertical agents capable of outperforming human experts in specific domains and an eventual maturation into an agent economy involving resource and transaction transfers.
- Current breakthroughs in reasoning, synthetic data, tool use, and agentic scaffolding are scaling intelligence as pre-training slows due to the depletion of low-hanging fruit.
- Coding is identified as a breakout category signaling an abundance era where labor is cheap, making taste the scarce asset, while voice generation in 2024 crossed the uncanny valley.
- The future economic model will operate under a "stochastic mindset" requiring new management approaches to handle less certainty with higher leverage, as organizational functions merge and processes are completed by AI agents.
- Challenges for persistent agents include achieving consistent memory and self-learning, necessitating seamless communication protocols for the transfer of information, value, and trust.
- The agent economy will foster a cottage industry focused on trust and security, operating in a human-centric ecosystem where agents and people collaborate rather than simply replacing one another.
- Specific early "killer apps" including ChatGPT, Harvey, Glean, Sierra, Cursor, and Abridged have emerged, with engagement ratios for chat applications approaching levels seen on platforms like Reddit.
- Long-term predictions suggest the integration of processes and agents into complex neural networks that will reinvent individual work, rewire companies, and fundamentally recreate the economy.