Hello! I'm your financial analysis assistant. This in-depth report on Y Combinator's (YC) investment trends for 2026 is packed with a lot of information. To help you understand it easily, I'll translate this article, which is full of jargon and data, into plain language with a structured breakdown.
---
📝 Summary of Key Points: Where Did YC's Money Go?
In one sentence:
YC invested in 630 companies this year, and the money wasn't mainly spent on apps for ordinary people. Instead, it was invested in building infrastructure for AI and transforming the real world (factories, defense, healthcare).
Three key conclusions:
1. Business-to-B (B2B) is the absolute mainstream: Over 90% of the companies were sold to businesses or organizations, with less than 6% sold to individual consumers.
2. Shift from building apps to building the infrastructure: Previously, companies were competing to create AI chatbots. Now, they're competing to build the infrastructure that enables AI to function, such as identity verification, payment systems, memory solutions, and optimized computing power.
3. The bar has risen: Having a good idea is no longer enough. YC now places more emphasis on how much money you've already earned (Annual Recurring Revenue, ARR) and the background of your team (experience from large companies, not just students).
---
🔍 In-Depth Analysis: Five Dimensions of the Report
1. Where Did the Money Go? – A Panoramic View of the Top Ten Trends
The 630 companies YC invested in this year can be categorized into the following ten areas:
First Category: The “logistical support” for AI (the most competitive and core):
- Agent infrastructure (creating identities and bank cards for AI):
- In plain language: AI employees need identities (emails, phones), the ability to spend money (payments), and memory. These companies provide the necessary services for AI.
- Examples: Inkbox provides email services for AI, Agentcard handles payments for AI.
- Optimizing推理 costs (helping AI save money):
- In plain language: Running AI is expensive. These companies use technology to make tasks more cost-effective or allocate them to the most suitable computing resources, saving customers money.
- Internet services for AI (interpreters):
- In plain language: Web pages and APIs are designed for humans. These companies convert them into formats that AI can read and use directly.
Second Category: AI directly replacing human jobs:
- AI employees (taking over jobs directly):
- In plain language: Instead of just improving efficiency, these products claim to replace specific roles, such as accountants, lawyers, or chefs.
- Examples: Rational creates AI accountants, Truffle manages restaurant kitchens.
- Healthcare (getting their own licenses):
- In plain language: Previously, they sold software to hospitals; now, they open their own clinics, get their own licenses, and even sell products like personalized peptides.
- Finance and insurance (full automation):
- In plain language: From underwriting to claims processing, everything is automated to reduce manual verification and auditing.
Third Category: Moving beyond the screen into the physical world (a return to hard technology):
- Robots and embodied intelligence (prices have dropped):
- In plain language: Robots used to be expensive, but now there are affordable (e.g., a $1688 home robot). The focus is on affordability and customizability.
- Factories, logistics, and supply chains (industrial operating systems):
- In plain language: These systems are installed in factories, warehouses, and logistics to optimize operations. This was the largest category this year (about 41 companies).
- Defense and space (government funding):
- In plain language: Developing drone swarms, anti-drone technologies, and precision ammunition. The main customer is the U.S. military.
- Data centers and energy (finding places for computing power):
- In plain language: With limited land and strict regulations, some companies are building data centers at sea to use seawater for cooling.
2. Who Are the Customers? – Why Are Consumer-Oriented (ToC) Projects Almost Nonexistent?
Observation:
In the latest Summer 2026 batch, only 5.5% of the companies were aimed at individual consumers, and there was only 1 company in the education sector. Over 90% of the companies were targeted at businesses or organizations.
Why?
- Predictable revenue: Businesses have clear budgets. For example, if an AI can replace an accountant, they are willing to pay. The logic is simple: it's about calculating the cost savings.
- Clear customer acquisition path: Identify specific roles, calculate costs, compare prices, and make a sale. Although competition is fierce, the revenue is certain.
- Challenges with ToC: Individual consumers are less willing to pay, and although YC talks about creating AI products for a billion people, little money is actually invested in these projects. This shows a big gap between what they want to do and what they are willing to invest in.
Implications for entrepreneurs:
If you want to target consumers, think about why YC wouldn't invest in your idea. Is it because the market is too small, or is it because monetization is too difficult? For now, the B2B market is a safer bet.
3. Trend Change: From Building Apps to Building Infrastructure
Data comparison (Spring 2026 vs. Summer 2026):
- Companies building AI apps: Decreased by 30%.
- Companies building AI infrastructure: Increased by 1.5 times.
- Industrial/hardware companies: Doubled.
Logical explanation:
This is a natural trend in industry development:
- First phase: Everyone was building apps (e.g., chatbots).
- Second phase: With so many apps, it became clear that companies were repeating the same basic tasks (identity verification, payment processing, computing power optimization). So, some companies focused on providing these standard services.
- Analogy: In the SaaS era, the first companies to make money weren't the ones that developed the software; it was companies like Heroku, Twilio, and Stripe that provided cloud services and payment interfaces.
Conclusion:
It's harder to build top-level apps now because the underlying infrastructure has been standardized. The opportunity lies in the services that every AI company needs but doesn't want to build from scratch.
4. Founder Profiles: Who Was Selected?
Team composition:
- Two-person teams are the norm: 64% of the companies have two co-founders. The proportion of solo entrepreneurs is increasing, indicating that the old rule of “you need partners” is becoming less strict, but two-person teams are still the most stable.
- Experience: The average team has 5.8 years of experience. These are not recent graduates or retired veterans; they are practical people who have worked in large companies and see how inefficient certain processes are.
Background and sources:
- Top universities and large companies: Alumni from Berkeley, Stanford, and Harvard account for 16%. Former employers include Amazon, Apple, Meta, and McKinsey.
- Internationalization: 50% of founders have international backgrounds. However, China is not among the top six sources of founders.
- Rise of Indian founders: Indian founders now account for 29% of YC startups in the U.S. (compared to 7% in 2008).
Key restrictions:
- Must move to the U.S.: 91% of international founders set up their companies in the U.S. (76% in San Francisco). YC is not a remote-friendly program; it's like a “ticket that requires you to move there.”
- Warning for Chinese entrepreneurs: If you're in China, with a team and customers in China, you'll have limited access to YC's resources (networks, financing, brand).
5. The Deal: YC Gives You $500,000 and Takes 7%, But What After?
Terms and conditions:
- Financial investment: $500,000.
- Equity: A fixed 7% stake plus the right to convert shares in future financings.
- Hidden conditions: The terms haven’t changed, but the entry requirements have. The new standard is a pre-investment valuation of $40 million and $1 million in annual recurring revenue (ARR) before the pitch.
What does this mean?
- Just having the right idea is no longer enough: You need to have a proven business model and actual revenue.
- YC's business model: It invests in 630 companies, allowing 620 to fail. As long as 10 succeed (with a valuation of several billion dollars), it covers all costs and makes a profit.
- Your risk: YC can make many mistakes, but you can only make one. So, bet on certainty (revenue) rather than the direction of your product.
---
💡 Five Practical Tips for Entrepreneurs and Observers
1. Don’t be the 90th Agent; be the one who sells the tools: If every AI company has to build its own identity verification, payment, and memory systems, consider selling these services separately. This is the most certain opportunity.
2. Name your products after specific roles: Instead of saying “improve team efficiency,” say “replace an accountant” or “replace a lawyer.” Pricing should also reflect this: charge by the number of users or by the results you provide.
3. Focus on unmet needs: Look at the areas mentioned in YC’s RFS (Request for Feedback) but not heavily invested in. For example, consumer products, education, and cryptocurrency. Even if YC invests little, it indicates a gap in the market that may be addressed in the next round.
4. When developing for the physical world, don’t rely solely on models: If you’re making robots or industrial software, your core advantage should be more than just a good model. You need data, access to real-world settings, and knowledge of procurement processes. Models are universal; context-specific barriers are the real competitive advantage.
5. Be cautious with the defense sector: This sector is growing rapidly but is heavily tied to the U.S. government. You can learn from the technology, but it’s difficult to replicate the business model directly.
❄️ Three Cautionary Notes
1. The list may be outdated: By the time you hear YC talking about investing in agent infrastructure, the first batch of companies has already been selected, and the competition in those fields is fierce.
2. Fierce competition: The first two categories (agent infrastructure and AI employees) account for a third of this year’s new companies. Many others are also seeing these opportunities.
3. YC can still make mistakes: Out of 630 investments, 620 may fail. Projects like offshore data centers and affordable robots may not be successful in three years. This is about probability, not prophecy.
📌 In One Sentence
YC’s investment in 630 companies shows that the definition of software is changing. It’s no longer just apps on screens; it’s the “infrastructure behind AI” and the “physical devices beyond the screens.”
If you’re thinking about starting a business, first look at what YC is saying (its RFS) and then at what it’s actually investing in (the data). Make decisions where there’s a gap between the two, and make sure you have actual revenue, not just an idea.