Picture this: A 17-year-old founder from Asia casually turns down a seven-figure buyout offer. Two rows over, another team pitches AI-guided counter-drone missiles to a room of investors sitting slack-jawed. Around them, 158 other companies deliver 60-second pitches that almost all start the same way: “Our AI does ___.”
Welcome to Y Combinator’s Summer 2025 batch. It felt less like a demo day and more like a time machine, showing us what tech will look like in three years.
The raw numbers tell part of the story. About 1,500 investors packed the room to watch roughly 150 founders (many barely past their teens) pitch their startups. But here’s the stat that hits you: 92% of these companies called themselves AI companies. That’s the highest concentration in YC’s 20-year history. If 2023 was the year some startups used AI, Summer 2025 was the batch where almost every startup was an AI startup.
But something more interesting was happening beneath the surface.
The Shift from Buzzword to Backbone
Founders weren’t just slapping “AI-powered” onto an app and hoping for checks. They were doubling down on solving real, unsexy problems that AI could finally crack. One team built an agent that crawls through code and fixes bugs while you sleep. Another trained voice models to answer restaurant phones so human staff don’t have to field calls during the lunch rush. A pair of sisters tackled loneliness by matching women for friendships using what they call an “AI bestie.”
Even the wilder ideas (yes, those mini-missiles) were grounded in actual needs. This wasn’t science fiction. It was a snapshot of where tech is headed right now, equal parts serious innovation and almost self-aware futurism.
So what did we do in this frenzy? We met with dozens of founders (over coffee, over Zoom, sometimes while they were literally debugging on stage) and placed bets on 13 startups we think are truly special. Below, we’ll walk through each investment, who they are and why we backed them, then zoom out to the bigger patterns we saw across the entire batch.
Think of this as your insider’s guide to Summer 2025: the companies we’re betting on and what they tell us about where things are going.
Our Batch Bets: A Quick Tour
We invested in roughly 8% of the cohort. These companies span enterprise AI tools, industry-specific platforms, and one breakout consumer app. Each one attacked a painful problem with a fresh solution, often one only possible now because of recent tech breakthroughs. And in every case, the founders showed that rare mix: deep understanding of the problem plus scrappy execution.
Here’s who we backed and why.
Autosana: The AI QA Engineer for Mobile
Imagine having a robot colleague who works 24/7 catching bugs in your mobile app before your users do. That’s Autosana.
Their vision-language agents act like tireless software testers, auto-generating and running tests that don’t break whenever you change the UI. (If you’ve ever maintained test scripts, you know the pain of brittle tests that fail the moment you move a button.)
Why we invested: The founder-market fit is off the charts. This team has built four mobile startups together before, so they know the territory cold. Timing matters too. New vision-language models finally make resilient, self-healing UI tests viable. And the early traction was real: even YC’s own mobile team started using it. By building their own device farm for testing, they’re gaining a cost and performance edge over alternatives.
For any mobile dev team that’s pushed an update on Friday night and lived to regret it, Autosana is like insurance that actually works.
Certus: Voice AI for Restaurants
If you’ve ever called a restaurant during dinner rush and gotten endless ringing, you know the problem Certus solves.
They replace restaurant phone lines with an AI that answers calls, takes orders, books reservations, and pipes everything into the restaurant’s existing systems. Restaurants lose an estimated $100K+ per year in missed calls. It’s a huge pain point.
Why we invested: The founders grew up in restaurant families and deeply understand this problem. They’ve already signed 1,000+ restaurant locations tired of losing customers when things get busy. We loved the vertical focus too. Rather than a generic voice assistant, they built a food-industry specialist that knows a pizza order from a pasta one.
Voice AI tech has finally crossed the usability threshold (no more awkward “sorry, can you repeat that?”), and restaurant owners are desperate for staff help post-COVID. The timing was perfect.
In short, Certus is giving every mom-and-pop restaurant its own AI receptionist who never puts you on hold.
Datafruit: DevOps on Autopilot
Datafruit is building an AI DevOps engineer to automate cloud infrastructure management. Provisioning servers, configuring security, optimizing costs, all handled by AI so startups never have to hire a DevOps team.
Why we invested: DevOps has become the new bottleneck in software. Code generation and AI coding tools have sped up building features, but deploying and managing them is still complex and human-intensive. Datafruit’s AI agent tackles that pain directly.
The team of four founders blew us away. Deeply technical, with backgrounds from Berkeley, Amazon, Meta, and Georgia Tech, plus years of building together. Their bold vision (”what if no startup ever needed to hire a DevOps engineer again?”) is matched by early proof: they already have enterprise pilots who really want this.
It’s ambitious, but if it works, it’s like having a seasoned DevOps lead living inside your AWS console, working lightning-fast for a fraction of the cost.
Floot: AI-Native No-Code Web Apps
Floot lets non-coders build full-blown web applications using natural language and AI, and crucially, debug and deploy them too. Think Webflow or Bubble, but reimagined from the ground up with AI assistants to hand-hold you through every “I’m stuck” moment.
Why we invested: The team has deep compiler and infrastructure expertise (not your typical no-code hobbyists), which means they’ve built a truly robust platform under the hood. They already had paying users and strong revenue growth even before Demo Day, showing that non-developers will pay for a smoother way to build software.
The AI-first approach is the differentiator. The platform doesn’t just give you drag-and-drop components; it actually helps write and fix code when you hit a wall. This dramatically lowers the barrier for a non-tech founder to create a real app.
With strong early traction and credible product-led growth loops, we see Floot becoming the go-to toolkit for the next wave of makers who have ideas but not coding chops.
Golpo: Instant Explainer Videos from Your Docs
Golpo turns boring documentation (specs, wiki pages, SOPs) into bite-sized AI-generated videos that explain the content in plain language. Feed it a dense HR policy and get back a friendly 2-minute explainer that feels like Cartoon Network meets corporate training.
Why we backed them: The Kar brothers behind it have world-class AI research credentials and a knack for practical applications. They’re not chasing Hollywood CGI; they focused the AI on clarity over flash, which businesses actually want for training and onboarding.
Early users loved it (some went viral internally, a good sign of product love), and they quickly hit meaningful revenue. The timing lines up with a trend we’ve noticed: companies moving toward asynchronous training. Let people watch a short video rather than attend yet another Zoom workshop.
Golpo is building an “explainer layer” for all the dry textual stuff in a company. Right idea, right time, right team.
Halluminate: Training Data & Sandboxes for AI Agents
If someone is building an AI that navigates a computer to complete tasks (think an AI executive assistant or an AI customer support rep), Halluminate gives them a safe playground plus high-quality human demonstrations to learn from.
Think of it as “Westworld for AI interns.”
Why we invested: One big bottleneck in the AI agents space is exactly this: lack of good training data and realistic environments. Halluminate attacks it from both sides. Simulated environments to let agents practice, and curated human workflow datasets to teach them.
The founder’s technical edge in this area is serious, and early customers (the kind who obsess over AI agents) were already knocking on the door. It’s a picks-and-shovels play in the AI gold rush, and the early traction with serious AI companies convinced us this wasn’t just a cool research project but something companies will pay for.
Humoniq: AI Agents for Airline Customer Support
Anyone who’s been on hold with an airline for hours (or tried to rebook a flight during a snowstorm) can appreciate Humoniq.
They build AI voice and text agents that handle airline and travel agency support calls end-to-end. “My flight got cancelled.” “I need to change my reservation.” No human agent needed in many cases.
The scale here is massive. Airlines field over a billion support calls a year, and it’s costly and frustrating all around.
Why we invested: The Humoniq founders previously built a successful travel business (Flightfox), so they know airline systems and where the bodies are buried. (Hello, GDS and SABRE.) Real-time AI speech tech has finally gotten fast and accurate enough to not annoy customers, and airlines post-COVID are desperate to cut costs in support.
Early pilots showed the AI resolving real customer calls with impressive success. It’s a classic vertical AI wedge: start with a niche (airline help desks) where the pain is acute, nail it, then expand.
If it works, next time your flight is cancelled, an AI might handle your rebooking and refunds before you’ve finished cursing at the weather.
IronGrid: AI Insurance for New Hardware
IronGrid is an AI-driven insurer focused on hardware companies, starting with batteries and other high-risk energy hardware.
If that sounds niche, consider this: when a startup invents a new battery or robot, traditional insurers shy away or overcharge because there’s little historical data on failure rates. IronGrid uses physics simulations plus machine learning to underwrite these policies more accurately, meaning hardware innovators can get insured at fair rates (and investors can sleep at night).
Why we invested: The founder has an exceptional background. Stints at Apple, cutting-edge battery companies, and Stanford, all focused on the science of reliability. In short, he deeply understands both the tech and the risk.
The wedge is smart: start with batteries (where he has data and expertise), then expand to other hardware like hydrogen systems, robotics, etc., essentially becoming the insurer for the clean-tech and deep-tech renaissance. With trillions in new energy infrastructure rolling out (and often underinsured), timing couldn’t be better.
IronGrid feels like a classic YC story: a boring-sounding problem (insurance) in a cutting-edge industry, tackled by a founder who lives and breathes it.
Minimal AI: Lightweight AI Tools for Enterprise Workflows
Minimal is building a suite of super-focused AI tools to simplify various enterprise tasks. Data cleansing, onboarding flows, internal dashboard automation, the boring glue work that every company has.
Why we backed them: We saw a strong founding team turning AI hype into tangible utility. They’d already landed some early enterprise customers and showed surprisingly strong revenue signals even at the pre-seed stage.
The philosophy of minimal AI (hence the name) appealed to us. Instead of boiling the ocean with a giant platform, they deliver small modules that slot into a company’s existing processes and make things just work better.
In an environment where every big company is experimenting with AI but many are overwhelmed by complexity, Minimal’s “do one thing well” approach is refreshing. We also liked that they focus on less glamorous problems (like cleaning up messy data) where there’s clear pain that hasn’t been overrun by competitors.
It’s early days, but Minimal could ride the wave of enterprise AI adoption by being the easy button for a lot of common needs.
Nixo: Ops Platform for Forward-Deployed Engineers
Nixo is building the first ops and knowledge-sharing platform specifically for forward-deployed engineers (FDEs), those technical teams that parachute into Fortune 500s to implement AI solutions on-site.
These folks are the unsung heroes of AI adoption in big companies, and they currently hack together their own tools to track issues, re-use scripts, and show ROI. Nixo gives them a centralized hub: scope issues faster, find if someone in the org has solved a similar problem, share fixes, all powered by AI to surface relevant info.
Why we invested: FDEs are mission-critical for enterprises embracing AI, yet no one else was building tools just for them. The founders have perfect backgrounds (Stanford AI lab, deep infra experience) to understand both the AI and enterprise sides.
Despite being a very early product, they showed unusually strong pilot traction. It turns out FDE teams at multiple Fortune 500s were already begging for something like this.
We love these kinds of under-the-radar opportunities: a new role (AI deployment engineers) emerging in big companies, with no dedicated software until now. Nixo could quietly become the default platform that powers the AI rollout army in every large enterprise.
RealRoots: AI Social Matchmaker for Newcomer Women
RealRoots is one of the rare consumer/social startups in this batch, and its premise is heartwarming. A mobile app that guarantees women new friendships when they move to a new city or stage of life.
How? By matching them in small groups through an “AI bestie” that learns their interests and schedules, then organizing fun real-life meetups. Think a coffee or a hike with 3-4 compatible people.
We don’t often invest in social apps, but RealRoots blew us away. The founders (two women who experienced this problem firsthand) managed to crack the hardest part of any social app: engagement and retention. In trials, women kept meeting up regularly and formed lasting friend circles, not just one-off meetups.
The traction numbers turned heads. They pulled in $782,000 in revenue last month from 9,000 paying users via subscriptions and events. If those figures hold, that’s phenomenal early product-market fit. (Indeed, RealRoots was cited as one of the batch’s highest-valued startups.)
Why we invested: Loneliness and social isolation (especially among young women moving for jobs or school) is a gigantic, under-addressed market. RealRoots has a clever solution that blends AI with real human connection.
Plus, it’s worth noting this was one of the few female-founded companies in the batch, and they’re exactly the kind of mission-driven yet sharp founders we love to support.
Picture a friendly AI concierge that introduces you to your future best friends. That’s RealRoots, and given their growth, it’s striking a chord.
Slashy: Agentic AI Copilot for Your Work Apps
Slashy builds an AI “coworker” that lives across your work tools (Google Workspace, Slack, Notion, CRM, you name it) and can actually take actions on your behalf, not just chat.
For example, it could notice your calendar is double-booked and move a meeting, or draft responses to Slack threads, or update a Notion page after a team standup. All the little digital chores that eat up time.
Why we invested: They represent what we think is the next evolution of workplace AI. Moving from passive assistants (a chatbot that suggests things) to active agents (a bot that just does it for you in the background).
Slashy’s approach of embedding into the tools people already use (rather than asking you to use a new interface) is smart. It lives inside Slack, Google Docs, etc., feeling like a natural extension.
The founders impressed us by how fast they ship and how well they understand bottoms-up adoption. Very YC-ish DNA: build something employees start using, which forces the boss to pay for it. Early traction was strong with small teams, and revenue was already coming in, showing that people will pay for a little automated “intern” that makes their workday easier.
We also saw a credible path for Slashy to become a broader “AI OS for work” if the simple use cases hook teams in. We’re in an era where everyone has too many apps and messages. An AI that lives across them to lighten the load sounds pretty heavenly.
Veritus: Compliance-First AI for Debt Collection
Veritus is bringing AI agents to the consumer lending and debt collections industry, but with a twist. They built a vertical SaaS platform and launched their own licensed collections agency in-house.
This means their AI can actually contact borrowers (via text, email, calls) to negotiate and service loans under full regulatory compliance, because Veritus itself is a regulated entity.
Why we invested: Debt collection is a massive, unloved problem ripe for AI disruption. In the US alone, lenders spend billions on it, and much of it is inefficient or hamstrung by legal rules that generic chatbots would quickly violate.
The founder is a force of nature. A repeat entrepreneur with prior exits, who saw firsthand how archaic this industry is and decided to build the “compliant AI” that others weren’t.
The combination of software platform plus in-house agency gives them a data moat (they’ll amass the proprietary interactions and outcomes data) and credibility with big lenders who might otherwise say “eh, we can’t trust an AI with this.”
Early pilot customers saw enough improved recovery rates that they’re coming back for more. For us, this is a bet that regulation-heavy sectors (like finance) need tailored AI solutions, not one-size-fits-all ones. Veritus’s early progress validated that hypothesis.
It’s not glamorous, but if they help banks recoup even a few extra percentage points of loans, that’s a multi-billion-dollar impact. And the best part? Borrowers might actually have a better experience too. No one enjoys talking to human debt collectors.
Takeaway: Thirteen companies is a lot to digest, but you might notice some common threads. We gravitate toward startups that use new technology to solve age-old problems, the kind that always existed but maybe weren’t solvable until recently. We also look for founder-story fit: people who have lived the problem or have exceptional insight about it.
Certus’s founders literally grew up in restaurants. IronGrid’s founder is a battery PhD who dealt with insurers firsthand. RealRoots’ founders built communities before. When those ingredients line up with a big market need and a timely tech enabler, you get that tingle that something special is forming.
And perhaps most importantly, these founders are relentlessly resourceful. As investor Gabriel Jarrosson put it on our podcast: “The best founders never stop. Close the door, they come in through the window. Close the window, they come through the chimney.”
We saw that grit in this group. Several had already turned down acquisition offers or big-job opportunities to bet on themselves. Many were coding through the night (sometimes literally debugging on stage at Demo Day). That kind of tenacity matters, because startups are never easy, but these are the folks who find a way through the wall, not around it.
Zooming Out: What the Batch Tells Us
Beyond our picks, a few big patterns emerged in YC Summer 2025.
AI Ubiquity (and Maturation)
It can’t be overstated: this was the AI batch. Roughly 9 out of 10 companies were building with AI in some form. That’s up from maybe 30-40% just a couple years ago. A seismic shift.
And it wasn’t just how many but how they were using AI. We noticed a shift from gimmicky “AI-powered X” apps to more substantive AI solutions. Founders targeted core business functions (QA testing, customer support, billing, logistics) and even built developer infrastructure (billing platforms, agent deployment tools) rather than just novelty chatbots.
In other words, the batch moved beyond the “AI hype for pitch points” phase into a “picks and shovels for the AI gold rush” phase.
To borrow an insight from Eric Ries (author of Lean Startup), this wave feels like past tech booms: yes, there’s some bubbly excitement, but underneath, it’s a genuine revolution in how software gets built and businesses run. AI today is both a bubble and the real deal, like the telecom boom where irrational exuberance and profound innovation happened together.
The trick, for founders and investors alike, is figuring out which AI startups are solving real problems versus merely riding the hype. We think the winners in this batch (and we hope, our investments) fall on the side of solving real problems with AI as the enabling tool.
Enterprise-Focused, Developer-Focused
YC used to be known for consumer apps a decade ago, but that era is definitively over. Nearly 85-90% of S25 startups were B2B (selling to businesses), and a full 30% were building tools for developers or engineers.
In fact, “developer tools” was the single largest category by count. To put it in perspective, 94 startups across YC’s 2025 batches built dev tools. That’s one new dev tool startup popping up every 4 days on average. In the S25 group alone, roughly a quarter of the companies were dev/infrastructure tools, from code assistants to testing platforms.
Why so many? It reflects the huge opportunity as software development itself gets a reboot in the AI era. Everyone is racing to provide the picks and shovels to this new wave of builders.
Meanwhile, consumer startups were rare, only around 9% of the batch. And those few that did exist often had an AI twist (language learning with AI tutors, social networking with AI matchmaking, etc.).
This heavy enterprise tilt is a continuation of recent batches but has accelerated even more. YC has clearly been prioritizing B2B ideas, which tend to have clearer monetization and arguably more predictable early traction.
From our perspective as investors, this trend is both a blessing and a challenge. There are more high-quality enterprise startups than ever, but picking the eventual winners (when many are somewhat overlapping, dozens doing AI dev tools for example) requires extra careful pattern-matching and diligence.
As Gabriel Jarrosson quipped on our podcast: “If you don’t have traction, what are we even talking about?”
In crowded fields like dev tools, we paid close attention to who had early customer love or revenue. Traction became the great differentiator amidst the enterprise glut, and the batch’s best definitely had it. Several companies were already at $100K-500K ARR by Demo Day.
Vertical Solutions & “Agentic” Platforms
A striking theme was startups either going deep into specific industries (vertical AI), or building the underlying agentic platforms to make AI more powerful.
On the vertical side, we saw AI tailored to restaurants (Certus), airlines (Humoniq), real estate and property management, finance/collections (Veritus), healthcare, and so on. These founders aren’t trying to make a general AI that does everything. They’re doing what you might call the “Iron Man suit” strategy, custom-fitting AI to each domain.
It makes sense. AI’s value often comes from understanding context, and context lives in domains.
Meanwhile, another set of startups focused on the infrastructure and platforms. We saw “Stripe for AI billing” (one called Autumn, which wasn’t in our portfolio but had a lot of buzz), a “Vercel for AI agents” (Dedalus Labs), tools for deploying and monitoring AI models, etc.
The clear evolution from even six months prior is that instead of 100 chatbots with GPT-4 under the hood, now we had companies building the plumbing and scaffolding to deploy dozens of specialized AI agents, or to manage AI workflows inside big companies.
In other words, the conversation shifted from “Hey, we have a cool AI demo” to “Here’s how we productize and scale AI in the real world.”
This is a sign of a maturing sector. It’s also exactly where we positioned many of our investments (Autosana, Datafruit, Halluminate, Slashy, etc., are very much about making AI practical at scale).
Batch Size & Selectivity
You might recall that a few years ago YC batches were gigantic. The Winter 2022 batch had 402 companies. In Summer 2022, YC deliberately shrunk the class to around 240 due to the market downturn.
Now in Summer 2025, the batch is about 160 startups (give or take, depending on how you count those that dropped or merged). YC has kept batches smaller and more focused in the post-2022 era, and it shows.
As an investor in the audience, Demo Day felt less like a firehose and more like a curated snapshot of the startup landscape. This also means getting into YC has become harder. The acceptance rate is rumored to be lower now than it was at peak batch size.
You could feel a bit of extra polish in many of the pitches and a bit more traction on average, likely because YC could afford to be picky.
For our Fund, it meant we had fewer companies to sift through than in the 2021 era, but each one probably deserved to be there. We still did dozens of meetings, but at least we weren’t trying to meet 300 companies in one batch. (A welcome relief for our calendars and sanity.)
Founder Profile: Young, Technical, and Global (But Still Few Women)
This batch’s founder demographics were fascinating.
First, it skewed very young. Many teams were just out of undergrad or even current students. It’s estimated roughly 15-20% of S25 founders were college dropouts or fresh graduates. We definitely saw that “hacker kid” energy. Think 19-year-olds with prototypes built in dorm rooms, and very few MBAs or suit-and-tie types. (In fact, one observer at Demo Day noted he could count the MBAs in the batch on one hand.)
These founders grew up with Github, learned AI not in theory but by playing with models, and they prioritize building over PowerPointing.
There was also a shared DNA among many founders in terms of prior affiliations. Google (especially DeepMind) was prominently on a lot of résumés. At least 30 founders had worked at Google or DeepMind, and a dozen had come out of MIT’s AI research labs.
So, in a sense, the new “startup mafias” feeding YC aren’t just ex-Facebook or ex-PayPal folks (like in the old days). They’re ex-DeepMind researchers, ex-OpenAI engineers, and alumni of top AI labs. This speaks to how much cutting-edge AI research has bled into startups. Those who pioneered techniques at big labs are now spinning out to apply them commercially.
We even saw some repeat founders in the mix (YC alumni coming back for a second startup), which reinforces that YC is now seen as valuable even if you’ve done it before. The network and platform are that strong.
On the geographic front, YC remained very international. Founders hailed from 30+ countries (the batch included startups from Pakistan, India, Europe, Latin America, you name it). That global spread is something YC has emphasized in recent years, and it continued, though interestingly, many of those international founders still relocate to SF for the program duration. So the center of gravity remains Silicon Valley even as talent comes from everywhere.
A special mention on diversity: the batch still had a significant gender gap. By our count, only around 15% of the companies had a woman founder, and roughly 9% of total founders were female (similar to global tech averages, unfortunately). This is a slight dip from some previous batches that inched closer to 18% female-founded.
It’s not a great statistic, and YC knows it has work to do on that front.
On the bright side, some of the most impressive traction we saw did come from female-led teams (for example, RealRoots with two female founders was absolutely killing it). It’s a reminder that while the playing field isn’t equal yet, the outliers can come from anywhere. And we as investors should actively seek them out, not miss the next Canva or Airbnb because it doesn’t fit the stereotype.
Open Problems & Surprises
For all the areas covered by 160 startups, there were still some big sectors left almost untouched.
An analysis of the batch data revealed that government tech had just one startup in the entire cohort, despite government being a $600+ billion market. Insurance (a gigantic industry) had almost no startups aside from a couple like IronGrid or one doing claims automation. Construction tech, despite trillions wasted in delays and inefficiencies, saw very few plays in the batch. Education and legal were also sparsely represented.
This hints at potential “blue ocean” opportunities for founders not following the herd. In a batch where dozens chased AI for software dev or customer support, almost no one tackled AI for construction project management, for instance.
For our fund’s strategy, this is interesting. It suggests areas we might proactively target or encourage teams to explore outside of YC. The flip side is some markets might be underrepresented because YC didn’t pick them due to difficulty or sales cycles. (Selling to governments is hard for a 3-month sprint.)
Still, it’s a useful radar: where are there billion-dollar problems and zero YC startups? That’s often where the next wave of founders can make a mark. As investors, keeping an eye on these gaps is as important as tracking the hot trends. After all, in venture, sometimes the real opportunity is where everyone isn’t looking.
Takeaway: As Marc Seitz put it on our podcast: “You can build anything now, so the question isn’t can you, but do you understand the problem deeply enough to solve it well?”
This quote nicely sums up the spirit of S25. With AI and modern tools, building technology is easier and faster than ever. Many YC teams shipped MVPs in days or weeks. The differentiator was founder insight and empathy.
The strongest companies were those where the founders really understood the customer’s pain in a way 10 AI copycats could not. It showed in their traction and in how they talked about their product. It’s a lesson we take to heart. Even in a batch overflowing with AI, the winners aren’t the ones with the fanciest algorithms, but the ones with the best grasp of a real problem.
Wrapping Up
YC Summer 2025 gave us a front-row seat to the next chapter of tech. We saw AI move from buzzword to backbone, industries once considered “boring” get turbocharged by automation, and a new generation of founders step up, some barely old enough to drink, others veterans of AI labs, all with outsized ambition.
For our Fund, it was a thrilling and busy season. We’re proud of the 13 bets we made. Each one in a way is a bet on people we believe in as much as the ideas they’re pursuing. And we’re happy to report many of them are already raising strong follow-on rounds (some oversubscribed within days of Demo Day).
For you, our LPs, the takeaway is that the pipeline of innovation remains very rich. Yes, valuations and markets will ebb and flow, but seeing this batch convinces us that the startup ecosystem is as vibrant as ever. It’s just evolving.
We have more technical founders solving more fundamental problems, and that’s a recipe for big impact. Our plan is to continue to stay close to YC (and other top accelerators and founder communities) to back the best of these teams early.
If anything, S25 reaffirmed our strategy of focusing on founder-market fit, early evidence of pull, and picks/shovels in booming areas. That approach led us to this portfolio, and we’re excited to see how these companies grow from here.
We’ll keep you posted as they progress (expect some deeper dives on a few of them in upcoming updates). In the meantime, thanks for reading this rather long newsletter. Hopefully it gave you not just a recap of what we’ve been up to, but also a sense of the broader tech currents at play.
YC Demo Days can indeed feel like “where the future gets built” (as one attendee put it), and we’re grateful to have a stake in that future through our founders.
Onward and see you next quarter, perhaps with fewer buzzwords, but no fewer reasons to be excited.
After all, if an AI can book our flights and build our apps, maybe it can also free up some time for us humans to do what we do best: imagine the next big thing.