AI's Trillion-Dollar Opportunity: Sequoia AI Ascent 2025 Keynote

The Sequoia AI Ascent 2025 keynote brought together partners to discuss the massive potential of artificial intelligence. They shared insights on why AI is poised to be a market opportunity significantly larger than cloud computing, highlighting key areas where startups can succeed and how the rise of AI agents will create a new economic landscape. Founders were encouraged to adopt a “stochastic mindset” and move with maximum speed.

The Immense Opportunity of AI

Sequoia Capital partners believe AI presents an opportunity at least ten times larger than cloud computing. The cloud transition, with its $400 billion in revenue, was already huge. If we compare that to AI, the starting point for AI services is an order of magnitude bigger. This means the end point, ten or twenty years from now, could be absolutely massive. AI is not just impacting services; it's also changing software. This means both profit pools are up for grabs. Companies are moving from selling tools to selling outcomes, and even selling work, shifting from software budgets to labor budgets.

Key Takeaways

• AI's market potential is significantly larger than cloud computing.

• AI is transforming both services and software markets.

• Companies are shifting from selling tools to selling outcomes.

The Speed of AI Adoption

AI is not just inevitable; it's happening now. All the necessary conditions are in place: compute power, networks, data, distribution, and talent. These technological waves tend to build on each other, making the current opportunity much bigger and faster than previous ones. Things are moving at an unprecedented pace. The physics of distribution have changed. For a product to succeed, people need to know about it, want it, and be able to buy it.

When ChatGPT launched on November 30, 2022, the whole world started paying attention to AI. Platforms like Reddit and the former Twitter, which didn't exist during the early cloud transition, now have billions of users. This makes it much easier for new technologies to spread. Today, 5.6 billion people are connected to the internet, essentially every household and business. This means the infrastructure for widespread adoption is already there, and when AI took off, there were no barriers.

Where to Play to Win

There's still a lot of open space in the AI market. While some companies are emerging, the opportunity remains wide open. Historically, companies that reached over a billion dollars in revenue during previous transitions were mostly at the application layer. Sequoia believes the same will be true for AI; the value is at the application layer.

However, there's competition. Foundation models are becoming very capable and are moving into the application layer. For startups not building vertically integrated businesses, the advice is to start with the customer. Focus on specific industries or functions and tackle complex problems that might still need human involvement. This is where the real value will be created.

Building an AI Company

Building an AI company is mostly like building any other company: solve an important problem in a unique way and attract great people. The 5% that's specific to AI involves a few key considerations:

• Customer-Back Approach: Instead of focusing on technology first, understand what the customer needs. Provide end-to-end solutions rather than just tools. This helps build strong customer relationships.

• Data Flywheels: Use product usage data to create a data flywheel. This unique data can be a significant advantage that competitors won't have.

• Industry Focus: Be deeply embedded in the industry you serve. Speak their language and understand their specific challenges. This allows for a strong connection with customers that foundation models might not achieve.

What Sequoia Looks for in AI Companies

When evaluating AI companies, 95% of the criteria are the same as for any company. The remaining 5% are AI-specific:

1. Real Revenue: Don't confuse “vibe revenue” (initial excitement) with actual, durable behavior change. Track adoption, engagement, and retention to ensure people are truly using and getting value from your product.

2. Margins: While gross margins might not be high initially, there should be a clear path to healthy margins over time. The cost of AI (cost per token) is decreasing, and as companies move from selling tools to selling outcomes, their pricing power should increase.

3. Data Flywheel Impact: If you have a data flywheel, it must tie to a specific business metric. If it doesn't, it either isn't a true data flywheel or it doesn't matter. A strong data flywheel is one of the best competitive advantages.

The Agent Economy

The concept of AI agents has evolved significantly. A year ago, agents were just starting to form into businesses. Now, these machine assistants are coming together as “agent swarms,” playing a critical role in the AI stack. Agents are working together, collaborating, and reasoning with each other. In the coming years, this will mature into an “agent economy.”

In an agent economy, agents don't just share information; they transfer resources, make transactions, and track each other. They understand trust and reliability. This economy will not replace humans but will involve agents working with people, and people working with agents.

Technical Challenges for the Agent Economy

To achieve this agent economy, several technical challenges need to be addressed:

• Persistent Identity: Agents need to maintain a consistent personality and understanding. They also need to remember and understand individual users to build trust.

• Seamless Communication Protocols: Just like TCP/IP enabled the internet, new protocols are needed for agents to communicate and transfer information, value, and trust. There's a lot of focus on this area now.

• Security: As interactions become less face-to-face, the importance of security and trust in agent-to-agent and human-to-agent interactions will grow even more.

A New Mindset for the AI Era

The rise of AI will change our mindsets in several ways:

• Stochastic Mindset: We need to move away from deterministic thinking. Unlike traditional computer programs that always produce the same output, AI systems can be more unpredictable. This requires managing uncertainty and risk.

• Management Mindset: Managing AI agents will be different from managing human teams. It will involve understanding what agents can and cannot do, blocking processes, and providing feedback.

• Increased Leverage with Less Certainty: The AI era will offer unprecedented leverage, allowing individuals and small teams to achieve more. However, this comes with increased uncertainty that needs to be managed.

A year ago, it was predicted that individual functions would have AI agents, eventually merging into processes completed by agents. While the “one-person unicorn” hasn't fully materialized, companies are scaling faster with fewer people. The highest level of leverage ever seen is expected. Eventually, these processes and agents will merge into complex networks, reinventing individual work, rewiring companies, and recreating the economy.