The AI market is bigger than anything that came before it — here's the case
Sequoia Capital partners Pat Grady, Sonya Huang, and Konstantine Buhler presented their analysis of the AI landscape at a firm event. The core argument is not that AI is important — that's table stakes — but that the size of the market, the speed of adoption, and the structural changes to business models are each larger than anything previous technology transitions produced.
Pat Grady: reframing the market size
Grady opened with the framework attributed to legendary Sequoia investor Don Valentine: "What is it? So what? Why now? What now?" Applied to AI, the "so what" is the market size.
The cloud market — Grady's benchmark — currently generates approximately $400 billion in annual revenue, exceeding the size of the entire global software market when the cloud transition began. That is the reference point.
AI's addressable market starts at least one order of magnitude larger than cloud's starting point. The reason: AI targets not just software spend but the service sector as well. As AI evolves from software tool to AI agent (copilot) to fully automated operator (autopilot), companies stop selling software licenses and start selling outcomes — and eventually sell labor itself, billed against workforce budgets rather than software budgets.
This is the structural shift: the two addressable markets — software and services — are now being redefined simultaneously by AI.
Why now? Grady identifies the "layer cake" of prerequisites:
- Compute capacity (dramatically increased)
- Network ubiquity (broadband and mobile globally accessible)
- Data availability (explosion of training data)
- Distribution infrastructure (social platforms with 1.2–1.8B monthly active users)
- Human capital (talent concentrated in the field)
The contrast with the cloud era: Salesforce's founder Marc Benioff had to run guerrilla marketing campaigns to get anyone to notice the product in 2000. ChatGPT's November 2022 launch required no marketing — Reddit, X (formerly Twitter), and global internet access at 5.6 billion users (vs. ~200 million at cloud's start) did it instantly. The "rails" were already in place.
Where startups can win: Grady argues the application layer will generate the most value, as it has in every prior technology transition. But the competitive dynamics are intensifying — foundation models are pushing capabilities further into the application layer. Startups that aren't building vertical integrations need three specific strengths:
- Start from the customer: understand their actual problem; build backward from there
- Go vertical-specific: deep domain expertise that a general foundation model can't replicate
- Target complex problems requiring human-in-the-loop: the cases AI alone doesn't solve
The "Leone Merchandising Cycle" (Doug Leone's 40-year framework): the full value chain from idea → product → engineering → go-to-market → sales → support. Building moats across this chain — especially through customer relationships that create proprietary data — is how durable businesses are built in the AI era.
Sonya Huang: reading the current state of AI
Huang reviewed the customer and technical landscape with notably specific observations.
On customer engagement: AI-native applications now show DAU/MAU ratios approaching those of Reddit. This is qualitatively different from early-wave hype: people are building daily habits around AI tools. The evidence is visible across advertising copy generation, educational concept visualization, and healthcare diagnostic support (OpenEvidence's clinical decision tool).
Voice AI specifically: Huang described 2025 as the "voice her moment" — a reference to the 2013 Spike Jonze film depicting natural human-AI conversation. The gap between what voice AI could do in 2023 and what it does now is the difference between a demo and a product.
Coding remains the single most dramatic breakthrough application. Models like Claude 3.5 Sonnet have changed software development so substantially that non-programmers are shipping functional software. One user built their own DocSend replacement using AI. The accessibility, speed, and economics of software creation have been fundamentally altered.
On the technical frontier: Pre-training scaling has decelerated somewhat, but the research ecosystem is generating new breakthroughs through other routes:
- OpenAI's reasoning capability improvements
- Synthetic data utilization
- Tool use integration
- Agentic scaffolding (MCP and equivalents enabling agents to orchestrate tools and other agents)
The combination of large base models + reasoning + tool use enables qualitatively new task categories. Huang noted that talking to developers about what o3, Operator, Deep Research, and Sonnet can do that previous models could not conveys the magnitude of progress better than benchmark numbers.
Products like Deep Research and NotebookLM (now called Nu) were explicitly called out as representing the blurring of research and product development — artifacts from that boundary zone.
Pat Grady: what Sequoia looks for in AI companies
On investment criteria, 95% is universal: important problem, strong team, execution evidence. The remaining 5% is AI-specific:
Revenue quality: "Vibe Revenue" — revenue that looks real but reflects customer enthusiasm rather than embedded behavior — is a genuine risk. What matters is adoption rate, engagement, and retention. Trust in a company's commitment to keep improving the product can be worth more than the product itself in early cycles.
Margin trajectory: Current gross margins for AI products are often low due to inference costs. That cost has dropped 99% per token over 12–18 months and will continue declining. Simultaneously, companies moving from tool sales to outcome sales and then to labor sales can raise prices. The combination produces a path to healthy margins — what matters is whether that path is visible.
Data flywheel specificity: Many companies claim data flywheel advantages. Grady's test: "Which specific business metric does your flywheel move?" If the answer is vague, the flywheel likely doesn't exist in any meaningful sense.
Konstantine Buhler: the Agent Economy
Buhler's section addressed the 5–10 year horizon: the emergence of an Agent Economy where AI agents don't just exchange information but transfer resources, execute transactions, track each other, assess reliability and creditworthiness, and form their own economic structures.
This is not a world without humans — it's a world where humans and agents collaborate. Realizing it requires solving three technical challenges:
- Persistent identity: agents maintaining consistency and memory of user context across interactions
- Seamless communication protocols: standards enabling transfer of information, value, and trust between agents — the TCP/IP of the agent layer
- Security: trust mechanisms for interactions with agents whose provenance and intent aren't immediately visible
Buhler argues these technical changes also require mindset shifts:
- Stochastic thinking: computers have historically been deterministic. AI systems are probabilistic. Users and operators need to internalize that AI outputs vary and build accordingly.
- Management thinking: directing AI agents requires the same skills as managing humans — understanding capabilities, giving clear instructions, providing feedback. This is a new literacy.
The result: "way more leverage with significantly less certainty." Buhler noted that the "one-person unicorn" prediction from the previous year's event hadn't materialized yet, but companies are scaling faster with fewer people than at any prior point in history. The leverage continues to increase.
The long-term vision: individual agents and processes coalesce into large, complex neural-network-equivalent collectives that reconstitute how individual work, organizations, and economies function.
Summary
Sequoia's collective message: the AI transition is larger, faster, and more structurally disruptive than any prior technology wave. The application layer is where durable value will be created — but only for companies that build defensible positions through customer relationships, domain depth, and proprietary data loops. The emerging Agent Economy is not speculative; its technical prerequisites are being assembled now. Companies that build with those prerequisites in mind will have structural advantages when the infrastructure matures.
Looking to optimize community management?
We have prepared materials on BASE best practices and success stories.
Streamline event operations with AI | TIMEWELL Base
Struggling to manage large-scale events?
TIMEWELL Base is an AI-powered event management platform.
Proven Track Record
- Adventure World: Managed Dream Day with 4,272 participants
- TechGALA 2026: Centrally managed 110 side events
Key Features
| Feature | Impact |
|---|---|
| AI Page Generation | Event page ready in 30 seconds |
| Low-cost payments | 4.8% fee — industry's lowest |
| Community features | 65% of attendees continue networking after events |
Ready to make your events more efficient? Let's talk.