The Bottom Line
When Eric Schmidt and Jensen Huang emit similar signals about AI Agents within the same week, this isn’t just personal opinion—it’s public confirmation of industry consensus. The focus of AI entrepreneurship in 2026 has shifted from “building large models” to “using Agents to solve specific problems.”
What Happened
Eric Schmidt’s Statement
Former Google CEO and Alphabet chairman Eric Schmidt stated clearly on social media:
“If you really want to make money, it’s actually easy. Found an agentic AI company.”
His core logic:
- The Agent era rewards builders
- The only credential is what you shipped this week
- Traditional degrees and backgrounds no longer matter in Agent entrepreneurship
Data Support
Schmidt’s statement is backed by real market data:
| Metric | Value | Meaning |
|---|---|---|
| AI VC funding (Q1 2026) | 4.5X QoQ growth | Capital flooding in |
| Aggregate AI capex (2026) | ~$820B | Unprecedented infrastructure investment |
| Hugging Face models | 2M+ | Mature open-source toolchain |
| Hugging Face users | 13M+ | Developer ecosystem formed |
Jensen Huang’s Supplement
NVIDIA CEO Jensen Huang also gave his judgment during the same period:
“Every engineer is going to have and manage hundreds of agents.”
This isn’t science fiction—it’s an engineering paradigm shift already happening.
Why Now?
Technology Maturity Reaches Critical Mass
The Agent tech stack in May 2026 is already sufficiently complete:
| Layer | Representative Tools | Maturity |
|---|---|---|
| Foundation Models | Claude, GPT, Gemini, GLM | Mature |
| Agent Frameworks | Hermes, OpenClaw, LangGraph, CrewAI | Usable |
| Tool Protocol | MCP (97M installs) | Standardized |
| Deployment Platforms | HermesOS, Dify, n8n | Commercialized |
| Skill Market | agent-skills (29K stars) | Community-driven |
A developer can build a production-grade Agent application in one week—something impossible six months ago.
Business Model Clarity
The money-making paths for Agent companies are increasingly clear:
- Vertical Industry Agents: Dedicated Agents for legal, healthcare, finance sectors
- Workflow Automation Agents: Replace repetitive work in specific roles (report generation, data organization)
- Agent Infrastructure: Toolchains for Agent development, deployment, monitoring
- Agent-as-a-Service: Usage-based Agent services
Practical Path for Entrepreneurs
If You Want to Start Now
Schmidt’s advice is direct—“whatever you shipped this week.” Specifically:
Week 1: Choose a Specific Pain Point
- Don’t build a “general AI assistant”
- Find a specific, repetitive, high-value workflow in an industry
- Examples: contract review for lawyers, monthly closing for accountants, resume screening for HR
Weeks 2-4: Build MVP with Existing Tools
- Foundation model: Pick a cost-effective one (GLM-5.1, DeepSeek V4 are very cheap)
- Agent framework: Hermes, OpenClaw, LangGraph all work
- Tool integration: Use MCP protocol to connect external systems
- Deployment: HermesOS free tier is enough to start
Weeks 5-8: Find 3-5 Paying Users
- Don’t chase perfection
- Find early users willing to pay for efficiency gains
- Iterate based on feedback
Pitfalls to Avoid
| Pitfall | Explanation |
|---|---|
| Training your own model | Extremely costly, open-source models are sufficient |
| Building a general platform | Too competitive, vertical domains are the opportunity |
| Over-focusing on technology | Users care about results, not which model you use |
| Ignoring compliance | Especially in finance and healthcare |
Investor Perspective
Q1 2026 AI VC funding growing 4.5X means:
- Now is a good window for fundraising: LP enthusiasm for AI projects is high
- But competition is also intensifying: Many teams see the same opportunity
- Execution speed is key: Schmidt says “credential is whatever you shipped this week”—speed is the moat
Action Items
- Start this week: Don’t wait for a “better time,” the Agent window may only be 6-12 months
- Start from pain points, not technology: Find work you or someone you know does weekly that could be Agent-automated
- Leverage open-source ecosystem: Don’t start from scratch—Hermes, OpenClaw, MCP have built the infrastructure
- Focus on monetization path: From day one, think clearly about who will pay for what