In-depth Analysis of the 630 Companies YC Invested in This Year: The Top 10 Directions It Is Most Optimistic About

Author: Leo Luosha, Luosha in AI

From 2026 to now, YC has invested in three batches: Winter 199 companies, Spring 196 companies, Summer 235 companies, totaling 630 companies. At the same time, it has updated its Request for Startups twice.

The list is being circulated widely, but it serves as an enrollment advertisement, indicating what YC wants to receive. The real answers lie within the 630 companies it has already funded.

I reviewed the public descriptions, industry classifications, and founder information from these three batches to answer four specific questions: What directions is it investing in? How crowded are these directions? Who are these companies selling to? What do the selected individuals look like?

Here’s the conclusion: Over 90% of the money YC invested this year is not for To C, and more than half is for AI infrastructure.

Let’s first list the directions directly, then look at the data, and finally, I will share a few points that I believe are truly useful for entrepreneurs.

01 The 10 Directions YC Invested in with Real Money

Arranged from most to least by the number of companies, after merging similar items, these are the ten directions. Each entry includes an approximate number from the three batches and a specific example.

01 Agent Infrastructure: The Layer That Issues IDs to AI Employees

There are 55 companies in the Spring batch, and a total of 89 companies in S26 related to agents and AI infrastructure, making this the most densely populated area this year. They are doing very specific dirty work: Inkbox issues emails, phone numbers, and 2FA to agents so they can be recognized by the system like employees; Agentcard handles payments for agents; Glen does memory; Hyperprobe does runtime debugging; Archal verifies what agents have done and automatically opens PRs for fixes.

02 AI Employees: Taking Over a Position Directly Instead of Making Tools

There are 56 companies in Spring and 52 in S26. Rational describes itself as an "AI employee" for accounting firms, while Last Accounting Company and Billow AI fall into the same category; Osmaura and Erinys are in legal; Locke handles government affairs; Truffle takes over restaurant kitchen operations. The commonality is that the product pages list job titles, not function names.

03 Inference Costs: Turning Bills Themselves into a Business

At least four companies in S26 are handling different stages of this chain. Understudy Labs captures production traffic, fine-tunes cheaper open-source models, and uses automatic quality gates to decide which requests can be routed; Conifer does inference routing; Stoa and OpenRelay are on the GPU side.

04 Robotics and Embodied Intelligence: Price Points Have Dropped

There are about 23 companies in S26. Nori is making a dual-arm home robot priced at $1,688, with the selling point being price rather than performance, and it allows users to train skills at home and share them; Salem Robotics inspects nuclear facilities; Libra Robotics does photovoltaic installations; Manifold and Proprio handle warehouse logistics. Emerging companies like Robocurve, Instance, and Markov are also providing evaluations for robots.

05 Operating Systems for Factories, Logistics, and Supply Chains: The Largest Group in S26

About 41 companies, making it the largest single cluster in this batch. Pango creates an agentic OS for e-commerce logistics, Control Seat adds a layer of intelligence to industrial SCADA systems, Alloovium interprets construction documents, and Torus acts as an engineering assistant for physical infrastructure projects.

06 Defense and Space: Procurement Has Personally Stepped In to Make Requests

Among the 26 hardware companies in Spring, 13 are in defense; S26 has grouped companies focusing on precision munitions, drone swarms, anti-drone technology, and battlefield situational awareness (ISENGARD, Earendil Robotics, Guild, Hop Aero). The entry "The Future of American Defense" in RFS is signed by U.S. Army Secretary Daniel P. Driscoll, not a YC partner.

07 Data Centers and Energy: Computing Power is Looking for Places to Settle

Atomarine is creating floating data centers at sea, using seawater for cooling; Pacific is developing modular data centers that can be quickly deployed to unfinished infrastructure. YC itself wrote in RFS about "offshore computing power," citing that land-based electricity and approvals have become bottlenecks.

08 Healthcare: Transitioning from Selling Software to Holding Licenses

There are about 20 to 24 companies in S26. Taiga handles AI medical billing for independent clinics, Care GP provides operational suites for general practices in Australia, and RonanRX has become a full-stack pharmaceutical company—delivering personalized peptides and GLP-1 directly to individuals. A notable change in this area is that many companies are no longer selling software to licensed institutions but are obtaining licenses themselves.

09 Finance and Insurance: Underwriting, Settlement, and Reconciliation

There are 17 to 26 companies in S26 (with slight variations in figures). Klaimee offers agentic insurance, Vestris handles mortgage settlement processes, and Spectre Intelligence focuses on intelligence and risk control. Note that the proportion in this area is declining, which will be discussed later.

10 The Internet for Agents: Redesigning Web Pages

Today's web pages, APIs, and documentation are designed for humans. Context.dev and Rindler are working on translating them into forms usable by agents; the "self-maintaining API" written by Harsha Gaddipati in RFS is another aspect of the same issue—automatically updating client codebases when interface changes occur.

Among the ten directions, five ultimately lead to the world beyond the screen.

02 Over 90% of These 630 Companies Do Not Plan to Go To C

First, let’s look at the industry distribution of the Summer 2026 batch, according to YC's own classification:

Figure 1 Industry Distribution of YC Summer 2026 Batch (n=235)

B2B ███████████████████████52.3%

Industrial ███████████23.0%

Healthcare ███8.5%

Fintech ███6.8%

Consumer ██5.5%

Real Estate █2.6%

Breaking down the industrial, healthcare, finance, and real estate categories, their clients are also enterprises and institutions. Therefore, a more realistic interpretation is that over 90% of this batch consists of companies targeting enterprises and institutions, with only 5.5% truly aimed at individual consumers, and only one in education.

This has two direct implications for entrepreneurs.

One implication is the customer acquisition path. Over 90% of companies will follow the same route—finding a specific position, calculating how much it spends in a year, and then comparing it with this cost. This path is crowded, but it has defined payers.

Over 90% of companies are competing for the same group of payers; in the remaining less than 10% of categories, almost no one is competing.

The other implication is its opposite. There are only a few dozen companies in the consumer category, and only one in education, while YC simultaneously wrote in RFS about "AI consumer products for 1 billion people" and children's AI tutoring. What it wants does not align with what it has invested in; this gap will be discussed separately.

03 In Half a Year, Money Has Shifted from "Making Applications" to "Fixing Infrastructure"

If we only look at four percentages, this year's changes can be summarized. This shows the movement between the Spring and Summer batches:

Figure 2 Four Displacements in Half a Year (Spring 2026 → Summer 2026)

AI Infrastructure 8% ↑ 20% █████████

AI Application Products 55% ↓ 39% ███████████████████

Industrial 12.8% ↑ 23.0% ███████████

Fintech 10.2% ↓ 6.8% ███

The number of companies making AI applications has decreased by 30%, while the number of companies working on AI infrastructure has increased by one and a half times. This is not a sudden shift in YC's preferences, but rather a normal sequence after a category reaches scale: first, there are a bunch of applications, and then someone extracts the dirty work of having to redo all applications and sells it.

A similar situation occurred when SaaS began to rise. The first batch to make money were not those making SaaS, but Heroku, Twilio, and Stripe.

When a category's billing itself is worth creating a company to optimize, it indicates that it has reached scale.

The doubling of industrial companies is the other half. To put it more intuitively: for every four companies YC invests in, one’s value ultimately needs to fall on a piece of equipment, a production line, or a warehouse, whereas a year ago this ratio was close to one in eight.

04 The Selected Individuals: Two People, 5.8 Years of Experience, Half Are Not Americans

This data mainly comes from organizing the founder information of the W26 and Spring batches, with the criteria being third-party statistics, not officially released by YC.

Team Size. Two co-founders are the absolute mainstream—129 out of 199 companies in W26 are two-person teams, accounting for 64%; single-person startups are 22, accounting for 11%; and about a quarter have three or more members. The number of single-person startups has noticeably increased compared to a few years ago, and the old rule of "must have a co-founder" is loosening.

Experience. The average work experience of founders before entering YC is 5.8 years; those in the agent direction are younger, with a median of 4.8 years. They are not students nor veterans, but rather in the age group of "having worked at a big company and realizing how bad a specific process is."

Schools and Previous Employers. Among W26 founders, there are 30 from Berkeley, 22 from Stanford, and 18 from Harvard, collectively accounting for about 16%. Among previous employers, Amazon appears in 14 companies, Apple in 12, and Meta and McKinsey have a significant presence in the agent direction.

Nationality. In the Spring batch, founders with international backgrounds account for half, with the top six sources in public statistics being:

Figure 3 Top 6 Sources of International Founders in YC Spring 2026 (Number of People)

UK ███████████████████████33

India ████████████████████29

France ███████████████21

Canada ████████████17

Germany ██████████14

Switzerland ████████10

Two things are worth noting for Chinese readers. First, there are no Chinese individuals in this top six list. Second, another set of data shows that among domestic YC founders in the U.S., 29% are of Indian descent, up from 7% in 2008 and only about 15% in 2019.

I did not find authoritative statistics on the proportion of Chinese founders, so I will not draw conclusions. However, it can be confirmed that 91% of international founders ultimately set up their company headquarters in the U.S., with 76% of Spring batch companies headquartered in San Francisco.

YC is not a remote project. It is a ticket that requires you to relocate.

If you are in China, your team is in China, and your customers are also in China, the cost-effectiveness of the YC path needs to be recalculated—not that you cannot enter, but that the value of that resource network after entering cannot be fully realized.

05 YC Gives You $500,000 for 7%, Then What?

The terms have not changed in recent years, but many people only remember half of them.

Figure 4 Two Payments in YC's Standard Agreement

█████$125,000 for 7% equity, post-money SAFE, priced immediately

███████████████$375,000 with no valuation cap MFN SAFE, converting to equity based on your next round's terms

Total: $500,000. The first payment truly affects dilution: 7% is fixed, regardless of your previous valuation. The second payment has no cap; the more expensive your next round is, the smaller the proportion it receives when converting to equity—so it is essentially betting on your ability to raise good prices after graduation.

So what is the price after graduation? The benchmarks given by investors tracking these two batches are $4 million in financing, with a pre-money valuation of $40 million, and having $1 million ARR before roadshows is no longer unusual. This is an estimate from the investor side, not data published by YC, but the direction is clear:

Choosing the right direction is no longer an advantage; having chosen the right direction and already generating revenue constitutes an advantage.

Putting these two numbers together, YC's business model is actually quite straightforward: buy 7% for $500,000 plus a subsequent conversion right, invest in over 600 companies, and only a few need to reach tens of billions to cover everything. This machine allows for a high error rate—but you cannot afford to make mistakes.

06 Five Most Useful Points from These Numbers

First, don’t be the 90th agent; go for that layer that all 89 others need to redo. The judgment criterion is simple: if every agent company has to build it themselves and they are all reluctant to do so, this matter is worth extracting and selling separately. Identity, payment, memory, evaluation, and inference costs are all emerging from this.

Second, write job titles on product pages, not function names. These two batches of companies collectively completed a shift in wording: they no longer say "help teams improve efficiency," but rather "replace this position." The pricing model also changes accordingly—charging per seat assumes your employees are using my software; once the narrative shifts to replacement, pricing can only be based on workload, results, or directly benchmarked against the annual salary of that position.

Third, the categories that YC repeatedly mentions but still have empty spots in the ledger are the true directions it is recruiting for. The consumer category only has 5.5%, education has only one company, and crypto is hardly visible in the batch structure, while all three are specifically mentioned in RFS. RFS is essentially an enrollment advertisement to fill gaps—it indicates what is missing in the applications we received.

First, clarify how much the client spends on the position you want to replace in a year. The rest is just packaging.

Fourth, if you are working in the physical world, first confirm that your advantage is not the model. The threshold in the industrial category lies in whether you can access that batch of data, enter that site, and navigate that procurement process. If your answer is "we have a good model," you are likely not in a competitive position in this track.

Fifth, the defense category needs to be judged separately for Chinese entrepreneurs. It has the fastest growth this year, but the dividends are highly tied to the U.S. procurement system. The direction can be referenced, but the path cannot be copied.

07 A Dose of Cold Water: Lists Lag, Tracks Are Crowded, and YC Can Be Wrong

Lists can sometimes lag. The Summer RFS mentioned "Software for Agents," but by the time this was publicly written, there were already 89 companies in the S26 batch. There is usually a gap of one to two batches between the appearance of RFS and the corresponding companies landing in large numbers. By the time you read the list, the first wave is already present.

The tracks are already very crowded. The first two directions above account for over a third of the new companies this year. You are not the only one seeing the opportunity.

YC can also be wrong. It invests in over 600 companies a year; it is a probability machine, not a prophet. The batch structure reflects what applications it has received and its current preferences, which do not equate to market conclusions. Offshore data centers, $1,688 home humanoid robots, and companies issuing IDs to agents may not survive three years from now—this is entirely acceptable within YC's model.

YC can be wrong 620 times; you can only afford to be wrong once.

There is also a question that this batch of companies collectively has not answered: when an agent truly replaces a position, who is responsible for its mistakes? Compliance, auditing, and insurance layers have appeared in RFS, but are still thin in the batches. This is both a risk and one of those empty spots mentioned earlier.

08 One-Sentence Summary of These 630 Companies

What they truly indicate is not which direction will win, but that the boundaries of the word "software" are shifting outward: part of it is moving to the invisible infrastructure beneath agents, and another part is moving to devices, production lines, and warehouses beyond the screen.

The entries on RFS will change with the next issue, but this movement will not.

If you take away only one thing from this article: first look at what YC says, then look at what it has invested in, and only make choices where the two do not align.

This content is provided for general informational purposes only and doesn't constitute financial, investment, legal, or tax advice. Any events, rewards, online promotions, or related information mentioned herein should not be considered a recommendation, solicitation, or invitation to purchase, sell, trade, or otherwise deal in any crypto assets. Crypto assets are highly volatile and may result in loss. The availability of WEEX services, products, and related events may vary by region. You are responsible for ensuring that your participation is in accordance with applicable local laws and regulations.