Newcomer VC Sentiment Report

Contents

0. Introduction

Introduction

Venture investors are on a knife’s edge as the artificial intelligence boom roars on — optimistic about the outlook and flush with cash, but increasingly jittery about nosebleed valuations and whipsawed by the unprecedented pace of change in the business.

That’s the excited but anxious mood that came through in numerous ways in our first-ever series of conversations for the Newcomer VC Sentiment Report.

Our inaugural circle includes twenty-five venture investors, including former and current Midas List members representing many of the best-known firms and emerging managers who’ve started their own funds. Speaking with the promise of anonymity, they answered a set of questions and shared their unvarnished views on a range of industry topics. The conversations were fascinating.

Some of the takeaways come as no surprise: VCs love Anthropic, for example. But we were struck by how people view the success of its anticipated IPO as crucial to keeping the AI financial flywheel humming. Nearly 40% of the panelists expect a big pullback in valuations, if not necessarily an outright bubble-bursting, within the next year.

We asked about dual-valuation rounds, which we wrote about recently, and got an earful. A clear majority disliked the practice, for various reasons, with Sequoia singled out as the firm driving the trend.

Investors had all kinds of ideas on which early stage company is most likely to be a generational startup: only Generalist and Quince garnered as many as two votes, while 31 different companies got one apiece.

On the flip side, the group was pretty united in their judgments of their peers. Founders Fund was the runaway winner on the question of which multi-stage fund investors would pick for their own money. Among emerging managers, Conviction was the clear choice.

This is our first report in what will be a quarterly series and reflects sentiment at the midpoint of Q3. We have the complete survey results, many choice quotes, and further discussion of each topic below. Together, we hope they will convey the flavor of our 25 fascinating conversations and serve as a benchmark as we continue to collect information the old-fashioned way: talking to the people in the trenches of the VC economy.

VC Sentiment Report

Compared with the previous quarter, how aggressively are you currently deploying capital?

VC SENTIMENT REPORT: A confidential survey of insider VCsGet access to the report at newcomer.co

Sept. 2026

1. Compared with the previous quarter, how aggressively are you currently deploying capital?

Deployment pace

There were no major shifts in how aggressively VCs were deploying capital during Q3, with a clear majority maintaining the stepped-up pace that’s become the baseline during the AI boom. Given the normal seasonal slowdown — several respondents noted only a very modest summer pause and didn’t see it as part of any trend — activity was quite robust. “The market has been absolutely insane,” one investor told us. “It's so nuts how busy it is.”

Still, a handful of folks said they were getting a little less aggressive due to less attractive opportunities, plus the large slugs into OpenAI and Anthropic earlier in the year that made it hard to match the dollar pace.

We expect this first question to yield interesting results over time as we see the trends in capital deployment.

Unlike some of these other people that are YOLOing into the Etched round at $20 billion, that's not what we do.

[Deployment is] slower than I'd like. But I think that's probably a good thing, given where valuations are. It feels like 2022, 2021 all over again.

As measured by new companies, less aggressively so far. We actually may not even make an investment in Q3, which is interesting.

We are behind pacing from where we would like to be in terms of deployment. But actually I see that as a positive... We don't love valuations right now. So we definitely are deploying more slowly than we did this time last year.

Nobody knows how to make money... So we're not deploying as much.

Pretty consistent pace for probably three quarters now.

I'm pretty sensitive to deploying too fast or even too slow. It's never good to be too fast, and it's not necessarily going to be too slow either. So I'm trying to stay on pace of a three-year investment pace, and we're pretty much doing that right now.

I'd say more aggressively, though I think it is very tied to a handful of investments. We've had some bulkier investments. So it's not a higher volume of checks. It's just the checks are bigger.

I don't think there's any seasonality in the business.

The obvious trade over the last three years was plowing money into these great late-stage platform companies, and then OpenAI, Anthropic. And our view is basically after that, and after maybe wave one of these app or harness or whatever companies you want to call them, you're left with a lot of companies with very speculative business models that historically have not done very well, are deeply competitive with each other, and are not a great risk-reward relative to what's coming for them.

VC Sentiment Report

Compared with the previous quarter, how is your interest in defense tech?

VC SENTIMENT REPORT: A confidential survey of insider VCsGet access to the report at newcomer.co

Sept. 2026

2a. Compared with the previous quarter, how is your interest in defense tech?

Defense tech

We got some pretty clear signals on which subsectors VCs think are hot, or not.

On the AI-adjacent categories, defense tech garnered the most enthusiasm, perhaps unsurprisingly given the urgent necessities of drone warfare and the drip drip of headlines about dwindling US weapons stockpiles. But not everyone is comfortable with it: it’s a crowded trade now, for one, and one person raised the issue of corruption in light of Palantir and Anduril’s exceptionally close ties to the Trump administration.

Capex-intensive hard tech stuff, be that data centers, chips, defense tech, satellites, that stuff is certainly significantly up for us. The run-of-the-mill applications, UI-centric, either infrastructure or applications, is down for us.

As a firm, I would say we're very bullish on defense. We think defense is amazing. But are we going to do a net new deal? I think unlikely.

There was a wave, or a bunch of it. We think some of those are going to be market leaders. There's going to be consolidation. But from here, for a net new company to be started today, I think that feels less likely to me than an Anduril or Saronic or Chaos or whoever it is consolidating.

For a long time I felt like especially the early-stage part of the market in defense has been just totally inflated and is completely bloated. There's been so much capital going to pre-seed and seed-stage funds in defense, and frankly, the TAM might be big, but it doesn't grow a lot. The NDAA is what it is, and it's allocated.

I think a lot of people's excitement has been skewed by Anduril and maybe one or two other companies that have really outperformed.

Unlike AI, which I would call an untapped opportunity where there's just a whole bunch of net new things to be done, defense is kind of a known landscape with a known problem set.

VC Sentiment Report

Compared with the previous quarter, how is your interest in AI applications?

VC SENTIMENT REPORT: A confidential survey of insider VCsGet access to the report at newcomer.co

Sept. 2026

2b. Compared with the previous quarter, how is your interest in AI applications?

AI applications

The biggest split in opinion came on AI applications. Some were dismissive on the grounds that the big labs were ultimately going to eat a lot of the app space. Others noted that major verticals including coding, legal, and customer service already had several established incumbents who would be hard to displace.

“There's just no moat, and people are going to want to pay less and less for software. They're going to be like, 'Oh, if it's not super dirt cheap, I'm just going to have it coded up by Claude or coded up by Kimi.'”

But others point out that applications have been central in every prior computing wave and there’s no reason to think Anthropic and OpenAI can service every vertical well.

“Being an investor is sometimes just going to where other people aren't,” said one confidant. “And right now people are avoiding AI app-layer companies, so that makes me think there's opportunity there.”

Being an investor is sometimes just going to where other people aren't. And right now people are avoiding AI app-layer companies, so that makes me think there's opportunity there.

I've been high conviction and focused there throughout. So actually maybe I'd say even higher, because I think we're starting to have a moment.

If you're building something for the office of the CFO or the general counsel, and it's deeply ingrained into a very archaic workflow in an organization, that's more interesting than a new PowerPoint competitor.

I think all the easy categories where you could have had breakout distribution before the labs got good at the use case have all been taken.

I think Google should just [build a personal agent], because they already have [the users]. Microsoft will do this, and people probably just don't care that their product is a little bit worse.

For more horizontal tooling, we're less interested. For more vertical and functional tooling, probably equally or more.

There's just no moat, and people are going to want to pay less and less for software. They're going to be like, 'Oh, if it's not super dirt cheap, I'm just going to have it coded up by Claude or coded up by Kimi.'

People like to take an intellectual point of view of wanting to know which one is going to lose. And I just think the honest answer is both are going to do very well, and the pie will grow quite a bit.

We've kind of moved away from DevOps or historical monitoring tools. A lot of the stuff that we would have invested in five years ago, the AI bent on those things are not interesting to us.

[OpenClaw] was like the Homebrew Computer Club of personal agents. And now we're getting the Macintosh or something.

Open source, namely distilled models from China that vertical applications can fine-tune or post-train against, has created an opening for companies to do really good work where they might have been cost- or compute-disadvantaged to compete against OpenAI, Anthropic before.

If you just look at [application] company fundamentals, the fundamentals are all strong. So despite the narrative that the labs are going to do everything, if you just look at, are apps growing revenues at astounding clips? The answer is yes.

I'm not super long a lot of the existing AI applications, which is why I feel good about investing in them early stage today.

I think the people who think the app layer is dead are the people at the labs and the people close to the people at the labs... I love the labs, but I think there's a form of psychosis inside the labs where there are researchers who've never worked in a company. And they're like, 'Well, obviously I can just one-shot this piece of software and it's done.'

Customer support is kind of like, you're Sierra, Decagon, or nobody.

The thing that would make me most bearish on apps, other than if the revenue started slowing down, would be if I really thought we were on an exponential on model progress, and I'm less convinced we are. I don't have data to point you to, but my vibe is we're going to make progress [and] Fable 6 will be better than 5, but I don't think we're getting to the point of one-shotting everything in the world.

I think it's less defensible than it's ever been.

VC Sentiment Report

Compared with the previous quarter, how is your interest in AI infrastructure?

VC SENTIMENT REPORT: A confidential survey of insider VCsGet access to the report at newcomer.co

Sept. 2026

2c. Compared with the previous quarter, how is your interest in AI infrastructure?

AI infrastructure

In the AI world, infrastructure including data centers, chips, and everything in between is seen as brimming with opportunity still: 14 respondents said they were increasing their focus on the subsector. This category covers a lot of different kinds of businesses — as one investor noted, he liked chips but not data centers — but they’re all linked to similar assumptions about an aggressive build-out of AI compute capacity.

“There's another wave of super interesting companies coming that are real infra companies that are reinventing the stack,” one investor told us. “Reinventing TSMC, reinventing ASML, reinventing some of the accelerators. Stuff like that is pretty interesting.”

It just feels more and more obvious that those are going to be the biggest companies in the world.

I think the winners have already been started, and the winners will keep getting larger. But I think it's less likely a new winner starts today.

You see so many of the infra companies, whether it's memory, whether it's chips, whether it's any derivative of those, they'll hit or even exceed their estimates and the stock will still go down.

Maybe slightly more, because the open-weight models and open models seem to be rumbling a bit.

There's been such a massive infrastructure build-out for the last three years, and it's been completely tremendous. Everything all the way down to silicon. We haven't had ambitions to rethink silicon at industry scale since the '50s, '60s, '70s.

Everything across security, everything across the stack. There's got to be so many opportunities in the next few years.

The level of sophistication of good tech that's in the infrastructure space is something that can't be vibe-coded. So there is still a moat, there's still defensibility.

If you looked at even the pitch to our LPs early in the AI wave, our view was that basically the infrastructure wasn't there yet to build great applications. So a lot of our investments were in infra... We've definitely shifted towards more application-layer things in the last 12 months, with increased intensity.

Chips, we've spent more time on. We've done optical networking stuff. We've done inference providers in the last bit.

There's another wave of super interesting companies coming that are real infra companies that are reinventing the stack. Reinventing TSMC, reinventing ASML, reinventing some of the accelerators. Stuff like that is pretty interesting.

There's a set of companies that are benefiting from the rise of agents and the increase in the number of agents. You look at Temporal raising at $12 billion, which was a company that, I wouldn't say was left for dead, but was kind of not growing as fast, but you're seeing a huge tailwind. Temporal, Vercel, Tailscale, Railway, Render, they're all companies that are really benefiting.

The part of AI infrastructure where there's been a demand boom is real estate, anything that's derivative of a data center, data center ops.

It's easier to use Vercel than to recreate a whole hosting platform. It's easier to use Resend than to recreate an email delivery platform. So there are tools that AI agents are using that are not specifically AI infrastructure that are interesting.

It's like during the SaaS boom, people didn't use less infra. They used more. So AWS became a thing. So we think that trend is going to continue.

If open source becomes a thing, then vertical apps become a thing. Vertical apps and open-source-native places to run, like the neoclouds and the inference providers, and then the infrastructure that enables it all, I think are competitively advantaged in this world.

Look at companies like Databricks. That's infra. So we're looking at Databricks as a proxy. And when you see the usage that's happening there, people are looking for a lack of vendor lock-in.

Spending a lot more time in compute and chips, robotics, scientific discovery. Those are probably the three more frontier areas we spend more time in.

Infrastructure is such a picks-and-shovels business, and you kind of can't lose. You're like bartending in Cancun on spring break. You can't miss. [But] compared to how enthusiastically you were entering headfirst into that six months ago, less. And it's less only based on how hot it got.

We haven't done [a neocloud deal]. We've thought about it probably more than we ever have in the past. So I would say increasing interest, but we've had a hard time getting over the finish line on those.

VC Sentiment Report

Compared with the previous quarter, how is your interest in AI labs & foundation models?

VC SENTIMENT REPORT: A confidential survey of insider VCsGet access to the report at newcomer.co

Sept. 2026

2d. Compared with the previous quarter, how is your interest in AI labs & foundation models?

AI labs

On the flip side, the foundation model trade seems to be almost played out: Only four of our confidants said they were increasing their focus on AI labs and foundation models, while 13 were less interested.

They are all small variants on the transformer. They're not reinventions. I would be very interested if I saw something that is a new architecture.

We could go do Kevin Weil's new thing, which is raising 200 at 650 pre on nothing. We could go do this neolab thing that we don't know much about. There's a couple of Stanford PhDs raising 200 at $3 billion pre, that is nothing except for some research. Or we could do Rillet, which actually is a business, growing, at a billion. If we were a growth-stage book, which one of those would we do? I actually would say probably Rillet. But it's wacky right now. It's crazy.

There are a few that we did earlier because we thought, 'Oh, maybe this is interesting.' But there are so many of them now. We're not touching them. I can't tell.

Short-term bullish, long-term bearish.

I don't want to go do these neolabs right now. I am shit scared at the capital requirements.

The more specialized labs, like science, bio, probably more interesting. But then general labs, less interesting.

I'm very positive on all of the scaled frontier labs except Anthropic, which is very contrarian.

I compare them in some way to social gaming companies. Zynga was only as good as the next FarmVille. The same goes for these model vendors. They're only as good as their next model.

What everybody's missing about the labs is I think they're going to turn out to be wildly profitable, and we'll see when Anthropic's S-1 gets revealed.

Just too many crazy-priced rounds.

There's no entry price of, like, 100 post. You're looking at entry prices in the billions. It's just nuts. It's nutty. And so if you're going to do those deals, you better be right.

I do think we're going to start getting to a place of diminishing returns on frontier models vis-à-vis the state-of-the-art open source that's available.

Even though there's all this discussion about open weight, it really has not cast any shadow on the labs' businesses, which seem better than they've ever been.

VC Sentiment Report

Compared with the previous quarter, how is your interest in fintech?

VC SENTIMENT REPORT: A confidential survey of insider VCsGet access to the report at newcomer.co

Sept. 2026

2e. Compared with the previous quarter, how is your interest in fintech?

Fintech

Fintech and healthtech were viewed as a bit ho-hum, with important exceptions: stablecoins were cited by one confidant as a major fintech opportunity, and AI healthtech is garnering real interest too.

Some stablecoin infrastructure and next-gen blockchain stuff we find really interesting on the transaction-processing and commerce side. But generalized neobanks and fintech, we feel like that cycle's over.

There's no reason why fintech should be less interesting today than it was five years ago.

It's similar to 10 years ago where you had payments and payees, except the payees are agents.

There's just no investor interest in those companies right now, or vanishingly small. And these are companies that are awesome. They're going great, they're doing well, they're growing. But if you aren't obviously an AI company, it's just a much harder market to fundraise in right now.

The last wave was stablecoins. That feels like it got overheated and played out. And then prediction markets as a section of financial services. Crypto is obviously kind of where it is, and stagnant at the moment.

We're increasing there simply because we're bullish that people still need ways to manage their money. That's a very contrarian belief, though.

Fintech is kind of an evergreen space. There was over-exuberance, but there's always going to be opportunities.

Do we want more exposure to fintech, or do we want more exposure to AI applications and infra? And the answer is obviously the latter.

This stuff costs real money. So money is going to have to be charged, which is actually a very good thing for the whole ecosystem and certainly for fintech.

VC Sentiment Report

Compared with the previous quarter, how is your interest in healthtech?

VC SENTIMENT REPORT: A confidential survey of insider VCsGet access to the report at newcomer.co

Sept. 2026

2f. Compared with the previous quarter, how is your interest in healthtech?

Healthtech

For these categories it’s more about specific products or businesses that could be interesting rather than any special interest in the sector.

In healthcare land, I just see a lot of me-too companies. I see a lot of consolidation in the future. I just see the 26th AI-for-billing company. And it's just not very inspiring.

We're increasing there. People still get sick. We're mainly doing more life sciences. But health tech in hospitals, we have a few companies, they're spending time there. So we're still deploying there.

Health tech may mean more programmable biology, different pieces of it, to us.

We're not going to go after something until we have a pretty strong point of view on what we're looking for. So sometimes we have sprints on things, and then we're like, okay, we need to reframe how we're thinking about what health tech means to us right now.

The traditional health tech, like wearables and digital health and software, less time on that. But more on AI-driven health advancements, drug discovery, pharma, things like that.

You have insurance companies that are incentivized to make their claims process really inefficient because they don't want to pay the claims. AI is really good at fixing that.

The healthcare industry is such a mess. And the reason it's a mess is mostly data integration.

There's so much opportunity in healthcare with AI now.

VC Sentiment Report

Are there any verticals not asked about that you're increasingly interested in?

VC SENTIMENT REPORT: A confidential survey of insider VCsGet access to the report at newcomer.co

Sept. 2026

2g. Are there any verticals not asked about that you're increasingly interested in?

Other verticals

Finally, we missed one important category for inclusion: robotics and physical AI, which was the clear winner on the open-ended category question. Biology and materials science, and energy were also cited by multiple respondents.

Cyber is the most valued sector in software right now. Palo Alto's $300 billion in market cap, CrowdStrike is 200, Fortinet's 100. We're going to see many more hundred-billion-dollar cyber companies.

[For consumer AI products,] this is 2010 for the iPhone, three years into the iPhone, which was the year that WhatsApp, Uber, Airbnb, many others were founded.

A year ago, we weren't sure of the role of open source in an AI age... And [today] I think we have more conviction that the pull is going more towards an open economy.

I'm probably the last person out there that is so bullish on certain parts of crypto.

[The next breakout consumer companies] exist today. Maybe we don't know quite which ones they are, but we'll see them fulfill their ambition and destiny.

Consumers don't want to save time. They want to spend time. They don't use email and calendar. They're looking for things that are not important with a capital I. They're looking for things on the spectrum from change the world to lean back and watch TV. They're looking for things closer to the TV end of the spectrum, and nobody's building that yet.

I believe in [AI-enabled services]. I don't know that it will be a venture business. We have not invested in any of those things because we're not in the private equity business.

Right now physical AI is fully consensus. And I think you're going to see some serious deployment into this category in the fall.

I think people will lose so much money in robotics. It's so early right now.

There's a lot of really interesting stuff happening both on the nuclear fusion side of things, with companies like Helion Energy, as well as modular nuclear fission companies that are developing smaller-scale modular power solutions to serve the whole power crunch that the AI data center boom is creating.

[Energy is] the one area that's not created a trillion-dollar outcome or a half-a-trillion-dollar outcome in the decade.

Maybe more contrarian is, hey, we believe the consumer applications are important.

Biology and life sciences, I think that still becomes very big. AI companies like Chai, which we're not doing, but then you have other companies that are using AI to go find better drugs or predict clinical trials or to help pharmas get clinical trials out faster.

Energy, chips, semis, inference. All of these are super boomy.

Mathematics and science, I think, are some of these interesting, somewhat investable categories. And maybe that's the neolabs or the next set of foundation model companies that are chasing those categories.

VC Sentiment Report

Which of the following best characterizes your view of the current state of the venture economy?

VC SENTIMENT REPORT: A confidential survey of insider VCsGet access to the report at newcomer.co

Sept. 2026

3. Which of the following best characterizes your view of the current state of the venture economy?

State of the venture economy

When it comes to peoples’ feelings about the overall state of the venture economy, the mood of the group would best be described as “feeling good, but a little nervous.” A majority is certainly net positive on the state of things, but in the conversations a lot of people who said things were good also said they were trending towards fragile. (We posed the question in a colorful way that prompted some fun conversations.) Only one person thought we were all heading for a big fall. On the other hand, the pace of change is clearly adding some strain: “I think everybody’s very anxious right now,” said one respondent.

Multiple respondents said they thought valuations were ridiculous but that they’d continue to play anyway.

“It's certainly a very full, frothy market, but that doesn't necessarily mean a bad thing,” said one. And regardless of mediocre historical performance measures, there’s no shortage of capital for venture funds. “There's so much LP demand to invest in venture capital, to an insane level,” another confidant added.

Billion-dollar companies are plenty, with not much revenue. Is that a red flag? Is that the new norm, given the size of the outcome that we're looking at with Anthropic and others?

[The stress is] the lack of exits, and the diversity of those exits. I know some people who are in SpaceX are feeling awesome, but there's a much deeper level.

These valuations are just not sustainable. These raises are not sustainable.

The overall fertility of the opportunity set, I think, is really unbelievable. Best it's ever been in my career.

I think everybody's very anxious right now.

Very few people in venture, and especially in very early-stage venture, think about the macro. Even in the last couple weeks, the news coming out of Treasury suggests we're going to see some interest rate changes that are going to have a trickle-down effect that a lot of people are not thinking through.

If you think about where venture was when prices were cheaper a decade ago, it was Slack or Uber or Airbnb, or Box or Dropbox. Not to discredit those companies, but those companies weren't, on weekends, proving fundamental theorems about mathematics or creating in-silicon intelligence.

We're three years into this cycle, four years into this cycle. These things tend not to last that long.

Most of the people who made money on SpaceX were random family offices and rich people, and half of Hollywood boomed off of SpaceX. And they all want to invest back in AI.

I think there's a little pressure on returns. People have gone back to the well with LPs. And people want distributions.

Outcomes are larger. And why are outcomes larger? Because the TAM is larger. There are just more people that use the internet. There's more commerce that flows through the internet than ever before. And AI has accelerated that.

There's so much LP demand to invest in venture capital, to an insane level.

I think people are excited about AI, but also there's so much capital rushing in. Probably the second problem is there's a ton of circular financing going on.

It does feel like 2021 again on several dimensions.

With all this secondary stuff, the scary interpretation is it gets handed off to retail or non-sophisticated high-net-worths or family offices. And it's kind of a hot potato.

It's certainly a very full, frothy market, but that doesn't necessarily mean a bad thing.

There are a lot of tailwinds related to macro that make venture a great place to be right now, if you are willing to be judicious about avoiding overvalued or overpriced stuff.

There's always a stupid hand-wringing. Literally every year, The New York Times calls bubble at least once a year.

Nvidia is investing in companies that can buy their [product]. Last time that happened was dot-com, and it happened in crypto, where you had FTX and Sam Bankman-Fried investing in other crypto companies so they could use [FTX]. It's crazy.

VC Sentiment Report

What do you think are the chances that AI-related valuations (public & private) will be down by 25% or more 12 months from now?

VC SENTIMENT REPORT: A confidential survey of insider VCsGet access to the report at newcomer.co

Sept. 2026

4. What do you think are the chances that AI-related valuations (public & private) will be down by 25% or more 12 months from now?

Bubble odds

On the great bubble question, our confidants were all over the map — just like everyone else. We picked the metric of a 25% or more fall in public and private valuations within a year as signalling the end of the party, and in aggregate, our group judges there’s about a 35% chance of that happening. But few had much conviction. “I'd be an idiot to tell you. This could last 10 years or five. It will pop at some point, but I don't know when.”

Most investors were fairly sanguine about the idea that a big correction of some kind is out there; there’s little choice but to keep playing.

“I think we all, in our sick-pleasure way, enjoy that fall in some kind of way,” said one. “But it will be bad. It will be bad for the market overall when that happens.”

I'd be an idiot to tell you. This could last 10 years or the last five. It will pop at some point, but I don't know when.

How many more credulous primary investors exist to keep subsidizing growth in capex and tokens?

The biggest influence on the AI market right now is the success or failure of OpenAI and Anthropic's IPOs.

The whole ecosystem depends on Anthropic and OpenAI's revenue. That's it. That's literally the only two numbers that matter.

I don't really consider [the big public tech companies] AI companies. Meta is in a lawsuit right now where they ruined the mental health of children. If they lose that lawsuit and they have to pay 100 billion in penalties, or God knows what, it's just too commingled. It's a conglomerate.

Maybe Saudi Arabia comes in and just gives $50 billion, and you're like, okay, that alone probably buys us another year or two of just grinding at it.

Nvidia just announced record earnings today at $52 billion. And I sat there going, well, they're going to get a lot more people investing in them, even though I don't think that's going to last.

I think there'll be mini corrections that might slide through, but I don't think there's going to be a large correction.

I think it's going to oscillate a lot. And so whether or not it'll be down or up in any one quarter is harder to predict.

We do think that there's an air pocket coming in the infrastructure build-out. I don't mean the buy side, like Anthropic and OpenAI. I mean overbuilding on neoclouds, data centers, that piece of the ecosystem.

We've never seen companies with this kind of potential this quickly. This kind of net new ARR and revenue growth this quickly.

We're in for probably some kind of major financial crisis in the coming months that will mostly screw the lower end of the economy. And that will have a drawdown in valuations temporarily. But over the next five years, AI is going to be a monster opportunity.

At some point I think the music will stop playing on [pre-revenue, billion-dollar] rounds in particular.

At some point I think we are going to hit some bottlenecks, whether it's chips or data centers. Or it could be open source takes a real bite out of OpenAI, or something that could change that sentiment. I just don't think it's a year from now. I think it's probably two years out.

I think you'll have volatility, but I don't view a 20 or 25 swing as a bubble burst. A 50 drawdown I would view as bubble bursting. But 25, I think markets will live with just fine.

Insofar as the economic value ascribed to these things is based on their capabilities, and the capabilities are growing exponentially, and maybe even at a faster exponential than 12 months ago, how can they not be dramatically more valuable?

There's so much concentration of capital, the averages might stay where they are, because the great companies are just accumulating so much capital and raising at such high prices. I saw a first-time team raising a $185 million seed yesterday.

[While that euphoria persists,] we'll have waves of companies get shuttered. But the value creation, I think, will outweigh the value destruction.

I don't think 12 months is long enough.

I think we all, in our sick-pleasure way, enjoy that fall in some kind of way. But it will be bad. It will be bad for the market overall when that happens.

If you say three years from now, it's like 80 [percent].

The thing about the public ones is actually they look much more rational than the private ones.

My hottest take is I think there's a good chance that in four years, either OpenAI or Anthropic will not exist. A bunch of people will have made a lot of money along the way. But the asset will not exist, because fundamentally it will be unable to sustain the level of capital it needs to exist.

VC Sentiment Report

What was the most competitive early-stage round over the past 3 months?

VC SENTIMENT REPORT: A confidential survey of insider VCsGet access to the report at newcomer.co

Sept. 2026

5a. What was the most competitive early-stage round over the past 3 months?

Most competitive early-stage rounds

We wanted to know which deals investors were really, truly clamoring to get in on, and in the latter category it’s no surprise that Anthropic came out on top. On early stage rounds, Jeff Dean’s Discovery Loop and the hot personal assistant Instinct were cited the most often.

What was most striking about the responses here, though, was how quickly they scattered after the top few. Investors cited some 25 different companies as contenders for most-competitive early stage round, which obviously suggests that it’s not just a few buzzy companies that are generating intense interest from VCs.

Right now there are definitely consensus founders coming out of the Cursors, the Mercors, that are just like that.

Often what ends up happening is somebody blows out the price on something, and so there's a question of, was the deal competitive or not? Lots of people would have done the deal at price X. Lightspeed paid 2X. At 2X it was not competitive, but at X it would have been competitive.

So many people tripping over themselves to do [Instinct].

[Why is Instinct hot?] Because people love heat. People love heat.

The best companies are ones where they're being preempted from the inside by insiders on the cap table, and maybe one new fund finds a way to sneak in for a smaller allocation.

[Trajectory had] five, six term sheets. They ended up going with Hacker 80. It's obviously a really interesting company.

I'll meet this 25-year-old on a Friday. And they're like, 'Yeah, we're expecting term sheets over the weekend.' I'm like, 'When did you meet the other firms?' They're like, 'No, no, we met them this morning.'

[Source Foundry is] building a new ASML in the U.S. These two guys, they're fucking awesome. It's a total science project, but they're fucking amazing.

The most interesting ones are the most contrarian and noncompetitive right now.

[Competitive rounds are] all just fake. It's never really competitive.

There's generally an absolute dearth of consumer companies. So anything that has legs of any sort of traction, plus a great founding team, will get so much interest.

If you come out of a top 50 company and you're 25 to 32 years old, everybody wants to write your first check.

I do think Rillet and Campfire were both very competitive, as examples. The craziest thing is Rillet, I would term it growth stage, but actually a lower valuation than Instinct.

[Vinod is] taking a load of flesh on the [Discovery Loop] SPV. So I don't know what to say on that one.

VC Sentiment Report

What was the most competitive late-stage round over the past 3 months?

VC SENTIMENT REPORT: A confidential survey of insider VCsGet access to the report at newcomer.co

Sept. 2026

5b. What was the most competitive late-stage round over the past 3 months?

Most competitive late-stage rounds

On the late-stage side interest was a little more concentrated — but not much! Some 15 companies won votes for the most competitive round.

The breadth of the perceived opportunities here might explain some of the nervousness we flagged above. Never have there been so many companies growing like rockets — and never have there been so many VCs chasing so hard after them. Losing out on Anthropic might be a disappointment, but losing out on a Baseten or Cognition might look like not doing your job, even if you were right there.

There might be a new VC skillset at hand. The Databricks round was like the “fucking food fight of all food fights,” said one of our confidants. “The business just really reaccelerated in the last two quarters.”

I think that was [the Serval] Series C that Sequoia did. Oh, I guess that was a Series B. I don't know, what even are these letters anymore? It was a $2 billion valuation.

The late-stage one that no one's talking about that was insanely competitive was Zipline.

Everyone's placing somewhat different bets. Some people are really bullish on the inference clouds and some people hate them, and so they're not competitive rounds at all. They only get one or two term sheets. And then the app companies, like Harvey, it was one term sheet.

[Isomorphic] was such a hot piece of tail that it was really an inside job of how they brought it together.

Fractal was very competitive, the chip company. And I think that also is a tranche one. I know Thrive's in there. I think Founders Fund was first tranche, Thrive was tranche two, Lightspeed tranche three. That was very competitive.

[The Databricks round was] like fucking food fight of all food fights. The business just really reaccelerated in the last two quarters.

If you're talking about Thinking Machines, actually, you can say it's really competitive because nobody can get in. But they're also dying for money. That's like every company.

It reminds us a lot of 2021, in 2026. But I think the competition definitely comes in when there's product-market fit. So more in the stage where it's working. Then everyone wants to be in it.

Ramp is going to be very hot and continues to be hot. They raise, like, quarterly.

The Anthropic round, the $980 [billion] round, was really, really, really competitive too. Near impossible to get access to.

The Chai round, I guess it was a Series C, that Index led. That was extremely competitive.

[The Cognition round is] by far the most excited or consensus round I've seen since Anthropic at 380.

Starcloud, I think, was pretty competitive, and they did the tranched round.

VC Sentiment Report

What would you do with shares in OpenAI?

VC SENTIMENT REPORT: A confidential survey of insider VCsGet access to the report at newcomer.co

Sept. 2026

6a. What would you do with shares in OpenAI?

Buy/Sell/Hold OpenAI

Opinions on OpenAI were much more mixed, with half a dozen of our confidants expressing outright pessimism; two said OpenAI’s current leadership was a concern, and one singled out CEO Sam Altman as the problem, saying they’d be a buyer if he were replaced.

“I don't know if I'd short OpenAI, because when you have sentiment going in your direction, capital concentrates,” said another investor. “So shorting would be a stupid move. But I'd sell it. They're lighting money on fire. I'd sell.” Still, others thought it would inevitably be a winner given the scale of the AI transformation.

Things feel insane, but I felt that in the last five years, and the ceiling has only raised. So what do I know? At this point, I'm a true believer.

[Sell, because of] the ongoing competitive pressure on foundational models paired with their enormous burn. Someone told me the other day, who knows, they burn 138 million dollars per day.

OpenAI is, in my view, a little bit contrarianly undervalued. But again, those two companies are holding up the entire fucking U.S. economy, literally.

Right now I'd buy all of them. And then sell them.

Currently, OpenAI Codex has a slight edge in vibes, and Claude Code feels like it's on the back foot a little bit in vibes. But it's really a tick-tock. And this is secondhand.

As a venture fund, we look for five to 10x returns in what we invest in, and I don't know that that's there for them anymore. As a personal investor, yeah, I would buy.

They're going to get commoditized. It's going to be a race to the bottom. It's easy to switch out. Most of these workloads are not going to be super sophisticated where you need the cutting-edge models. I just don't believe that.

There's a change in CEO could be coming down the road, and so it could be doing better as a result. And the model is definitely better than before. So I wouldn't buy, but I wouldn't sell it. I will hold.

It's just so many departures. Something's wrong with the way that [Sam] manages.

Their medium-term, largest competitor is Google, who has infinite capital and needs to win the space. Google makes about three times [OpenAI's daily burn] a day, and they spend more than that on capital expenditures.

I think what your question should be is, would you buy right now or short? That should be the question. Because if you say sell, it depends when I got in.

There's a strong case that it was just going to fall behind. I just don't find the product offering as compelling. The product builder is compelling, and obviously huge leadership changes, etc.

I don't know if I'd short OpenAI, because when you have sentiment going in your direction, capital concentrates. So shorting would be a stupid move. But I'd sell it. They're lighting money on fire. I'd sell.

Either we believe what we're saying or we don't. Either we believe that we are on the cusp of this incredible economic transformation, and OpenAI is one of the main companies that's going to participate in it.

VC Sentiment Report

What would you do with shares in Anthropic?

VC SENTIMENT REPORT: A confidential survey of insider VCsGet access to the report at newcomer.co

Sept. 2026

6b. What would you do with shares in Anthropic?

Buy/Sell/Hold Anthropic

Anthropic might be a controversial company in some quarters, but among venture investors it’s all but an object of love. Fully 19 of our respondents would buy shares in the company even at its valuation of nearly $1 trillion — a remarkable vote of confidence in the frontier lab’s ability to stay ahead technologically, manage the mind-bendingly complicated safety and security issues associated with advanced AI, and stay on sides with the politics.

It might not be a true VC investment anymore though: “As a venture fund, we look for five to 10x returns in what we invest in, and I don't know that that's there for them anymore,” said one confidant. “As a personal investor, yeah, I would buy.”

The problem that they have is they don't really have enough compute to continue to compete. They're the most compute-constrained of all of the major ones. Because they don't own any data centers. OpenAI, Google, Meta and SpaceX have monopolized all the compute.

Now that surge pricing in the market has taken shape and [Uber is] trying to make a profit, you're no longer getting to JFK for $30. The only reason I bring this up is I have the $200 Max plan. I should be paying $2,000 a month for how much I use Anthropic.

This is the most obvious buy on the planet.

You can't [reduce spend if you're] Anthropic, where they burn $5 billion on a model... and then three months later it's obsolete.

What I understand, the economics are much better than OpenAI.

They're getting more ingrained with the enterprise. There's better management. The company is run much better. OpenAI's not run all that well. Anthropic's run much better. They have the team.

It's like a game of Settlers of Catan. They seem like they're in the lead, so everybody starts to [gang up on] them a little bit. They don't want to trade sheep with them.

We're all being subsidized by people lighting money on fire. And Google, ironically, has the most stingy plan.

I think it's going to be a 10 trillion dollar company.

Anthropic is obviously undervalued in my view. Super obvious.

Early innings. I think it will do a great IPO.

VC Sentiment Report

What would you do with shares in Stripe?

VC SENTIMENT REPORT: A confidential survey of insider VCsGet access to the report at newcomer.co

Sept. 2026

6c. What would you do with shares in Stripe?

Buy/Sell/Hold Stripe

Opinions on Stripe were pretty evenly split, with some seeing more upside and others feeling it had mostly played out. Interestingly, though, the issue of Stripe’s apparent intention to stay private indefinitely was not a problem: investors said there’s such a robust secondary market for Stripe shares that it wasn’t an issue.

[Sell.] Stripe is not an AI company. I think it's probably a slow compounder, is my sense.

There are so many things Stripe should win. I think the OpenRouter buy was a really good buy. If they can win agentic payments, that's a whole other chapter. And they're so good. And they actually have a good org.

OpenRouter, very smart investment acquisition. So I will buy.

It's a payments company. My goodness. They haven't done much since.

[Buy.] If you talk to growth investors, and I'm not growth myself, they will say at their valuation, their cash flow multiples are not crazy.

Everything that's happening is good for Stripe. If you believe that we're going to have a bunch of agents running around the internet buying shit, then you need a trust layer, someone to validate all this stuff, and they're becoming the pipes on that.

It was borderline. I almost said short, but I think it's hold.

Stripe still feels in so many ways early innings in terms of where it's headed and what it wants to displace in terms of Mastercard, Visa, and new payment opportunities.

It's a duopoly market where they're becoming by far the number one. And they're really levered to this trend.

The stock is quite liquid in the secondary market. People expect that the company will continue to do well. So even if there's no IPO, there are always more and more people who think that buying it, they will sell it later at a higher price.

VC Sentiment Report

What seed to Series B company are you most bullish about being the next generational startup?

VC SENTIMENT REPORT: A confidential survey of insider VCsGet access to the report at newcomer.co

Sept. 2026

7. What seed to Series B company are you most bullish about being the next generational startup?

Next generational startup

We were interested to know which specific early-stage companies our confidants thought could be “generational” startups — very large or very important in some other way. As with the question on competitive rounds, the responses showed the remarkable breadth of the perceived opportunity: no fewer than 35 companies were named as possible breakthrough ventures.

Only two, Quince and Generalist, got as many as two votes. Quince is an online retailer and Generalist is building robot software, so we can be confident it’s not the category people are judging.

“Everything's a bet right now,” said one investor. “I have companies going from two million to $150 million ARR, and I don't know if they're going to last. It's kind of crazy.”

We couldn’t identify a lot of commonalities among the 33 companies that received one vote for generational startup. FWIW, we popped the names into Claude and asked for commonalities, and the best it could do was that they “share a distinctive naming style” and “a lot of them are AI-first, even the ones that don't sound like it.” That much we figured!

I think [Modal] has a shot at being a hundred-billion-dollar business.

[Wabi is] one of the only products that is a network. There's just so few network effects, and the lesson of the last few years around moats is that the traditional moats matter more than ever, and networks are the gold standard... And then [Eugenia] is extraordinary. She's a complete force of nature, dedicated to nothing short of dominance.

To be truly generational, it has to be an absolutely massive idea. Simile is a big idea, the synthetic user startup, and a talented team.

Arguably the TAM, the market for [Physical Intelligence, Generalist, and Skild], is even larger than the market for an OpenAI or an Anthropic, because you're potentially displacing or replacing physical labor globally.

I don't know about paying $2.5 billion for a pre-monetization company [like Instinct], but I understand why this category is hot. One of these companies, I think, could be absolutely generational.

Most of these hyper-growth AI application companies are growing the fastest. If you look at even Legora, Lovable, Higgsfield, Wispr Flow, they start off prosumer and then they've expanded into enterprise. It's truly that motion.

Everything's a bet right now. I have companies going from two million to $150 million ARR, and I don't know if they're going to last. It's kind of crazy. Nobody cares about money anymore.

[Robotics has] gone from very theoretical potential and rudimentary progress to being on what feels like the cusp of that GPT-3 moment.

I hate this era, or this stage, right now.

I think there's room for a bio model company still, because there's no open data set. That could be Boltz or Chai. A bio-data or bio-modeling company.

If I had to do a bet, I would bet on MatX. These are the guys that built the TPU at Google. They're one of the three new chip companies, and they're my favorite of the three.

[Aaru is] either a $50 billion business or a zero. And based on the day, I oscillate back and forth.

VC Sentiment Report

What percentage of startups backed by top VCs do you think are meaningfully misrepresenting their financials to investors?

VC SENTIMENT REPORT: A confidential survey of insider VCsGet access to the report at newcomer.co

Sept. 2026

8. What percentage of startups backed by top VCs do you think are meaningfully misrepresenting their financials to investors?

Financial misrepresentation

Our question about founders misrepresenting their financials, which we’re hearing more about these days, ended up revolving around what it means to “misrepresent.” Most of our confidants said that outright fraud was pretty rare, less than 5%, though a few people had seen more than that.

The more interesting issue is when “fake it ‘til you make it” crosses the line into unethical misrepresentation. Investors said financial projections especially were increasingly rife with exaggeration and misleading optics. Some of our respondents' comments suggest ethical backsliding by founders: spinning numbers has become standard practice for many, with young entrepreneurs especially seeming unaware of the lines.

“Intentionally it's only like 10%,” said one confidant, commenting on how many pitches he sees with dubious numbers. “But I think unintentionally it's like 75%.” Another investor added: “We have accepted that playing fast and loose with that, in certain ways, is part of hustle.”

In a frothy moment like the present, though, more entrepreneurs are going to cross the line. Numerous respondents said they’d walked away from deals because they were uncomfortable with the data, though only a couple cited outright fraud.

Right now people are being very fast and loose with their numbers. CARR versus ARR, live versus not.

I think it's about a third of companies. I think it's pretty high. I think people are very optimistic or labeling things very differently.

If you say you have a contract with Anthropic and it turns out you have an email that speculates what a contract could look like, that's a lie.

You want founders who are great fundraisers and storytellers and walk this line of reality and optimism. And my bet is a lot of them have some number in there that is not totally accurate to today's reality.

I would call it mislabeling, which is still a misrepresentation, because I've caught people on this. And I'm like, 'This is not ARR, man. It's not recurring.'

[Growth investors] literally are throwing money at [OpenEvidence] with no diligence and committing $50 million on a single Zoom call with no data, with nothing.

If you're a founder and you raised $20 million and you utterly blow it, you can raise another $20 million a month later. If you're a founder that raises $20 million and you misrepresent your numbers or steal it or do something like that, of course you're never going to raise money again.

A lot of venture is a lot on trust and handshakes and reputation. And there are definitely people who've burnt reputation by overplaying their hand.

I remember the first time [redacted startup] told me they're targeting a billion in ARR at the end of this year. I literally laughed at them. I was like, 'You cannot put that in a pitch deck.' And so, I don't know, I'm just wrong.

We have accepted that playing fast and loose with that, in certain ways, is part of hustle, versus [dishonesty].

Quite often we pass because they misrepresented the nature of how the product is being used and its core utility.

If I think you're a fraud, I walk immediately. I don't care what company you're building. I don't care how good it is. I don't care about the round. If I think you're a liar, I'm gone.

Founders, their job is to tell the most optimistic form of the truth. And sometimes that can cross over into things that are less optimism and more hope, and less truth and more truthy.

We have definitely backed out of a number of deals because we could not get the data that we wanted to get comfortable.

What percentage of the projections that a company gives you is full of shit? 99% of the time it's bullshit.

I think intentionally it's only like 10%, but I think unintentionally it's like 75%.

I don't know that we've seen a deck without something like that in it.

The VC needs to do their homework and dive into what their definition is. That's always been the case. The VC has to do the homework. The VC can't not do their job. Some VCs don't do their job, and I doubt they're going to be top tier for very long if they're not doing their job.

VC Sentiment Report

If you could be an investor in any multi-stage private markets firm besides your own based on their investments in the last 12 months, who would it be?

VC SENTIMENT REPORT: A confidential survey of insider VCsGet access to the report at newcomer.co

Sept. 2026

9a. If you could be an investor in any multi-stage private markets firm besides your own based on their investments in the last 12 months, who would it be?

Best multi-stage firm

When it came to judging their peers, there was a lot of agreement among our confidants. Founders Fund was the clear pick among the large, multi-stage firms, with people citing their contrarian strategy as a clear winner. Fund boss Peter Thiel is, ironically, among the few venture investors who’s made it big by going against the grain; he led the way in defense tech investing, for example, long before it was in fashion. SpaceX too was considered uninvestable by standard VC analysis — it was far too dependent on long-term government contracts — and we know how that one came out. Sequoia also got a nod from six of our investors, though the tone was a bit grudging; it’s no fun betting on favorites! Thrive also picked up some votes, which are certainly merited based on recent performance.

“Thrive, maybe... but I wanted to give something other than Sequoia,” said one confidant. “Like everybody, it feels like the easy answer, the cop-out answer.”

I do respect that [Sequoia is] like, 'Hey, we're not doing SaaS-is-dead. We're going to do this thing, and we're going to go hard into the paint for it.' I'm like, great. That's discipline.

For return or for information? Information, investing—all of them. [For return,] Sutter Hill.

Thrive is like, we have scale, we see everything, and we make really clear judgment calls and have a lot of opinions on what we want to invest in. That, to me, is truly venture capital.

[a16z has] talented people there, and they're good, and they're run well.

[Sequoia.] Probably not that original. But good investments.

Thrive, maybe... But I wanted to give something other than Sequoia. Like everybody, it feels like the easy answer, the cop-out answer.

We have very little transparency into how many of them are actually doing.

I think Benchmark continues to clean up.

I think [Founders Fund is] just investing in categories that have the highest likelihood of breaking out over the next decade. They're not just pure software investors. They're actually in a bunch of different categories. They're bold.

I really like some of the stuff [Founders Fund is] doing with New Limit and Fractal. They're kind of back to taking these huge swings on moonshot-oriented companies. They're pretty fucking good at nailing those, at getting one to really work.

I think Spark has done a good job. You have to give [them] credit. [They] had a big win and then parlayed that into the next and the next, and I don't think [they've] taken [their] foot off the gas.

[Menlo] did Anthropic early. So I wouldn't mind being an LP in that fund.

I would definitely not say Sequoia. I love them, but they're literally Andreessen now. They just do everything.

Weirdly, Menlo is number two. And it's so weird, because their brand does not have the same allure as any of the other names I mentioned. But I really credit them. They've just done really well in this moment in time.

VC Sentiment Report

If you could invest in any venture fund started in the last 5 years, what would it be?

VC SENTIMENT REPORT: A confidential survey of insider VCsGet access to the report at newcomer.co

Sept. 2026

9b. If you could invest in any venture fund started in the last 5 years, what would it be?

Best emerging fund

Among emerging managers, the picking was even easier: Sarah Guo’s Conviction was the big winner with twelve votes. Yet there was also an undercurrent of skepticism towards the emerging manager category. “They’re kind of all bad,” snarked one confidant. Few have yet delivered DPI.

There were some pretty interesting names among the pack though. Alexis Ohanian’s 776 got a strong endorsement — “he’s a great founder, founders love him and therefore he gets incredible access,” said one person. Nebular, Neo, and Nova got two mentions apiece. Dimension, Compound, Abstract, and Hyperion rounded out the pack.

One respondent said they’d most of all like to be invested in the venture arms of OpenAI and Anthropic. That strikes us as a prescient thought, given how big a role corporate venture is playing in the AI revolution.

Abstract is the most interesting new venture fund in the game.

[Alexis Ohanian is] a great founder. Founders love him, and he therefore gets incredible access.

There's Conviction. There's Chemistry. There's a few others. I love those guys. I've invested in a few of them, personally. But I'm not looking to invest in them. Most of them have paper returns, not actual returns.

That's a tough question, because they're kind of all bad... Probably Theory. I think Tomas is quite good.

I do think there's some question mark of how much do they own of those companies? How big are the checks that they've participated in? What are the companies that are pure-play Conviction companies? It's one thing to be the investor of record. It's another to be a participant.

[I'm assuming] people talk about Conviction because it's the hype, the most hyped-up name. But the fund is so hyped up, it feels like valuation is going to come down for that reason. It's guaranteed overpay.

Sarah [Guo of Conviction is] a close friend and she's doing terrific. I think on a dollars-in, dollars-out basis, Dimension's probably stronger.

My answer would be the OpenAI venture fund or Anthropic's venture fund, because they have this pretty unique insight into what's working and what isn't.

I just don't think there's been a legit [one].

VC Sentiment Report

What is the most overpriced startup that has raised money in the last 12 months?

VC SENTIMENT REPORT: A confidential survey of insider VCsGet access to the report at newcomer.co

Sept. 2026

10a. What is the most overpriced startup that has raised money in the last 12 months?

Most overpriced startup

In asking about overpriced and underpriced startups, a number of our confidants had to first get over their gut feeling that everything is overpriced. Once they worked through that, we got some interesting opinions: chipmaker Etched, which raised $700 million in an August round that valued it at $21 billion, took the top spot, with 5 votes. Perplexity has long drawn skeptics, but only two people thought it merited most-overvalued. Twenty one companies got the nod as overpriced, which probably proves the first point above!

[Figure AI is] raising from not nefarious, but questionable individuals. That cap table reads of Theranos. It's not like institutional VCs are giving him these rounds. It's a whole bunch of sovereigns and random groups.

[Perplexity is] just between a rock and a hard place. And also can't access compute.

There's a lot of circular shit happening around [Cognition].

Bret Taylor could probably find someone to buy [Sierra] for $25 billion. But does the revenue and growth rate justify that valuation? I don't know. In a world where the company is still fundamentally selling Anthropic's models.

Probably Instinct right now, at two and a half billion, if that's real. But two as of last night.

[Baseten is] burning a ton of cash. You have to raise all the time because of the high burn. And people are seeing their technical team is actually not that impressive. Lacking a little bit, but really, really good at PR.

I think [Lovable is] overpriced because it's a self-serve product. And so their retention's terrible.

I actually am both the biggest believer in Town and the most baffled at the price.

Hadrian's the one where I'm like, I don't know what's going on here. The best working hypothesis that I've heard is that they basically Barnum-and-Bailey'd themselves into being.

I struggle with a lot of neolabs that are working on the same research problems that everybody else seems to be working on too.

[Reflection] just raised $2 billion from Nvidia or something. They haven't put out a model yet.

[The neolabs] have basically all been pre-product, pre-model. And to me, all the multi-stage funds are investing in them as a call option on the possibility that one of them becomes a future OpenAI. But it's totally disconnected from the substance of what's actually been produced so far.

Jeff Bezos's stupid-ass thing. Project Prometheus.

[Etched has] built a commodity chip, which their customers will tell you. In a market that's becoming increasingly competitive [and] where OpenAI now has a chip that's more performant.

I think [Wispr Flow is] a commoditized technology.

Cursor was ridiculous. But kudos to them. Very lucky that xAI needs help on the front end. So it worked out.

All the neolabs feel overpriced to me.

If Perplexity is actually raising at $30 billion, that still feels nuts to me.

You're kind of burning the candle on both ends with these [chip] businesses.

[For Colossal,] I ran the comps on SeaWorld, and it doesn't feel like there's a venture-backed outcome there.

VC Sentiment Report

What is the most underpriced startup that has raised money in the last 12 months?

VC SENTIMENT REPORT: A confidential survey of insider VCsGet access to the report at newcomer.co

Sept. 2026

10b. What is the most underpriced startup that has raised money in the last 12 months?

Most underpriced startup

Underpriced companies were tough for many of our confidants to single out — the corollary of everything being overpriced. But a couple still feel that Anthropic was a good buy at its current valuation, and Modal also got two votes. Beyond that it was a grab-bag. Only one company, Cognition, had the distinction of making both the overpriced and the underpriced lists.

Someone's going to pay a big price for [Modal] in the next couple months, probably, I would guess. And so maybe it won't be underpriced at that point in time.

There's a wave of companies that have seen huge acceleration from the rise of agents. Temporal is a perfect example of this, where they were priced at two to $4 billion, and it was mispriced because they're seeing rapid acceleration.

The fastest-growing demand for U.S. Treasuries right now is stablecoins, and they're all short-duration, three-month T-bills. It's such a clear and huge market opportunity, and it's wild to me how the market is so fixated on AI.

[Zipline is] the only company that I know of that has FAA clearance to do what they do. You can now order a burrito in Fort Worth and have it delivered within 12 minutes to you by a drone, and it costs them 10, 20 cents to deliver it.

Probably Anthropic at every price.

I don't remember seeing a deal underpriced.

Vercel, in terms of what they've built, looks a lot more like Cloudflare, for people that look under the hood, than it does just a cloud business.

The last round [for Railway] was done at $425, and they haven't raised again yet. Their revenue is going to be at more than a quarter of their valuation.

[Anything] scaling really quickly outside of AI [that] no one's paying attention to.... We have [non-AI] companies in our portfolio that are profitable, growing 200, 300% a year.

11. What are you seeing and how do you feel about dual-valuation / tranched rounds?

Dual-valuation & tranched rounds

Our question about dual-valuation, or tranched rounds, where prestige investors pay a lower price than others in what is essentially the same round, probably drew more heat than any other topic. Nearly everyone agreed the practice had become more prevalent, though there was some difference between those who see it as all but ubiquitous these days and those who see it as an occasional deal term.

Most respondents took a dim view of the practice, though many were resigned to it being a new feature of the landscape. It’s not good for employees, who end up with a higher option strike price, and it can create unachievable expectations, the critics said. Some were also bothered by the misleading nature of it: a company announces funding led by a prestige firm at a high valuation, while concealing the fact that most of the prestige firm’s money came in at a much lower price.

A number of our confidants specifically cited Sequoia as the main driver of the practice. “It's everything that people hate about VC. We're not helping ourselves as a brand when people do shit like that,“ said one confidant. “I think it's fraud. You shouldn't be selling securities at different prices to different people,” said another.

But founders like high valuations and many confidants said it was just the market at work. Having a Sequoia or Benchmark in your corner as an early-stage startup does in fact make it more valuable. “We're doing it right now with two companies, where it's like, this is our price and everybody else pays a higher price. I've been doing that for years,” said one investor. “I love it. It's great. The founders love it too.”

We're doing it right now with two companies, where it's like, this is our price and everybody else pays a higher price. I've been doing that for years. I love it. It's great. The founders love it too.

It's some kind of Jedi mind trick that Sequoia has with certain founders.

I think people are getting fed up.

It is worse for employees, because the 409A gets struck at a higher valuation given the higher tranche.

It's not nearly as in vogue as people are making it sound.

I don't actually view it as bad as other people think it is, because it's supply and demand. The founder gets demand for a higher price. That's just how efficient markets work.

It's a structure that is downstream of how many new foundation labs, incredibly capital-intensive companies, get started.

It's everything that people hate about VC. We're not helping ourselves as a brand when people do shit like that.

It is accurate to frame this as the valuation. But yes, it's absolutely incorrect, all things considered. But is it wrong? It's not wrong. It is the valuation the last investor came in at.

[If I were the CFO], I [wouldn't] love it, because it sends a bunch of different mixed signals to the folks coming onto your cap table in terms of the value that you're ascribing to them, and creates a headache from an accounting standpoint.

I feel like I've seen it for years. It doesn't feel that new to me. And then it just became this thing, this Twitter thing. What would often happen is you'd have a priced round, and then any overfill would fall into a capped note at a slightly higher cap. Now maybe we're having subsequent equity rounds. It's a slightly different structure, but it's the same intent.

I think what they say is, 'Hey, don't worry about it. We got your round. We'll price it at 500. Does that work with you? Cool. You know what? I bet you can go now immediately mark it up 30.' That's how they do it. It's not, 'It's priced at 600, give it to us at 500.'

[Hone] was a triple tranche. Three separate prices, all roughly 10% ownership, from three mega funds, all in the span of a week.

The number one place it happens is in neolabs, because they're just so capital-consumptive.

That's different than a fast follow. If the round is done and some other investor shows up and says, 'I really want to invest, I'll invest at twice the price,' okay, I think that's okay.

On a one-to-ten deception scale, [Starcloud is] a six. Serval's a 10.

The practice feels icky to me.

I think it's fraud. You shouldn't be selling securities at different prices to different people.

It's the company bringing in a lot of money but then lowering their dilution, because they can do that. So I understand why it's evolved this way. But I think overall it's not great, because companies with too much money, not a great history of that being successful as a strategy.

VC Sentiment Report

If you could tell a new college grad to go into any profession today, what would it be?

VC SENTIMENT REPORT: A confidential survey of insider VCsGet access to the report at newcomer.co

Sept. 2026

12. If you could tell a new college grad to go into any profession today, what would it be?

Advice for a new college grad

One narrative that our confidants definitely aren’t buying is the end of the software engineer. Nine recommended software in some shape or form, potentially as a researcher at a big foundation lab. “It's actually an incredible time to be a software engineer, because you basically have an army of infinite junior engineers,” said one.

Another suggested rotating through a bunch of San Francisco startups for a few years as a way to gain skills and insight. “Go learn how to have emotions” was perhaps the most challenging career direction, with another offering a more nuanced version of that, citing the deeply human “skills of coercion, of persuasion, of collaboration, of consensus-building, of moving things forward.”

Only one person suggested professions most obviously immune to AI, like sports and live entertainment. Nobody was very keen on pointing young grads towards venture capital.

It's actually an incredible time to be a software engineer, because you basically have an army of infinite junior engineers.

If you can get a job in venture, I would say it's a great place to be. But don't stay too long, or you're going to be a venture lifer.

Definitely not be a VC. Our market is shrinking, not growing.

I think [live sports and entertainment is] the most protected career path. If they're an athlete, great. But the better option would be working within the team, or being part of the infrastructure of the league.

Look at Anthropic's careers page, open roles, and look at AI research and engineering. It's like, Life Sciences Operations Lead. Research Engineer, Life Sciences. Research Operations Lead, Biology. Research Scientist, Life Sciences.

Go learn how to have emotions.

Having a grounding in the sciences, combined with being AI-native, is a really exciting combo for the next 50 years.

What will forever be a trait that remains native to humans is the ability to organize, galvanize, inspire, coordinate. That will always stand with humans. And so no matter what profession you take, if you can develop the skills of coercion, of persuasion, of collaboration, of consensus-building, of moving things forward.

[Start a company.] The ability to raise capital at the earliest stages is only increasing. It's a benefit of being a somewhat frothy market.

Mathematics, philosophy, critical thinking broadly, entrepreneurship. Those would be the things that I would tell my kids to study right now.

Dude, I tell them, do drug discovery. There's a whole revolution happening in AI right now, and you're not going to get automated away by some other AI.

[Software engineering] was a way of thinking and problem-solving. Not a syntax that we were learning.

I kind of want to be contrarian. I would still do computer science.

Pick a startup in San Francisco to join for a year, and then go to another one, learn for another year, go to another one, learn for another year, in whatever role you can, that they will hire you for.

13. What's a story we should be chasing right now?

The story we should be chasing

We got a lot of great story ideas!

I would be interested in the anecdotes of the people that have been at Anthropic for four years as a sales rep, or five years as a sales rep. And now you're worth 40 million bucks, or soon to be worth 40 million bucks. How do you even think about it? Do you just keep going because you still believe in the mission? Do you think of cashing out and going to live on the beach? Do you become an artist?

Companies doing a few million dollars in ARR and growing 100 to 300% are having a hell of a time fundraising.

SpaceX is buying companies so that they'll use their data centers. It's nuts.

The top tick of this AI bubble is going to be when OpenAI's revenue starts significantly declining.

All of these companies that have raised in the billions in valuation that have still not delivered even a product. Forget about revenue. There's not even a product.

Why is Nvidia doing all these deals, like Perplexity and Hugging Face?

[The Instinct round] got priced at two and a half billion. For them to make venture returns on that, that has to be a 25, 30 billion dollar company, given additional dilution it'll take from employee options, additional financing, all this stuff.

In China you have this mirror ecosystem emerging that's almost more technologically progressive.

It's almost like you can perform better as a person who has jumped into venture today and just invested in your smart friends, if you have good access, than somebody [with a long track record].

It's hard to look at Grok Bot, Instinct, Town. These are commoditized products now, with really the same foundation to them.

How much secondary matters in companies for talent now. You have to orchestrate, every 12 to 24 months, tender offers for top talent. Otherwise you can't compete.

The hype-versus-substance gap is much smaller for Fireworks compared to Baseten.

All of the Gulf region capital right now, because of the Iran war, I heard a stat that they're spending $100 million a day on their self-defense. That comes totally out of the dry powder that they hold to invest in funds.

Linear. $100 million of ARR, 177% net revenue retention. It's valued at two and a half billion, as of a tender offer today. And so just the discrepancy between that and the stuff that is raising on nothing at billions in valuation.

Distillation of frontier labs from the U.S., coming back as fine-tuned models here, enabled by China, is just changing the entirety of the venture landscape. I think that's just a provocative reality that no one's really said cogently, clearly yet.

How much the average application investor has turned deep tech in the last three months. Kind of crazy.

How far is physical AI away from the ChatGPT-3 moment?

Probably Palantir's the biggest beneficiary of the Trump administration policies on defense spending.

The really spicy one is, how many VCs are going to survive in the next fund cycle?

All of these venture firms you're talking to, they've made money primarily in consumer internet and in enterprise software. There are some anomalies, SpaceX is a massive anomaly, but by and large that's where venture profit has been made. And now there's a lot of stuff happening outside of that.

I think a very interesting story is the rise of Gavin Baker. He's been somebody that many of us have respected and appreciated for years. And in this AI economy platform shift, he's become one of the most trusted voices. His star is definitely rising.

If we see five more deals like [the Poolside acquisition] in the next six months, then maybe the neolab asset class is underpriced.

We used to try and get 20% ownership, and they got thrown out the window. But what is the right ownership? [The big firms] don't own anything in these companies at $20 billion.

Just the Apple story and AI, what they can do on device and where that goes, I think feels underreported. I just wouldn't count them out yet, in a way that it feels like a lot of people have. A lot of what Apple is well set up to do could be in personal agents.

How are lenders going to lend [OpenAI] more money, and how will they honor these contracts?

Bending Spoons basically setting 2x as the Airtable revenue multiple is not great.

Even the paper returns for a lot of funds aren't that good, even if they have five or 10 unicorns in the portfolio. The entry prices are just so bad for so many of these things.

Just understanding what's happening right now vis-à-vis space superiority, and the race that is silently underway there between China and the U.S., could be a pretty interesting story.

I think the old-school LPs are still the ones who are allocating. And maybe they're grumbling about there not being DPI, but they're still cutting new checks into Lightspeed's latest fund.

All of these researchers getting funded as neolabs, that used to apply for grants for funding, and now they can just get VC capital. So they're kind of taking funding intending to stay researchers, but it's a much easier life.

The VC Sentiment Report

Introduction

Compared with the previous quarter, how aggressively are you currently deploying capital?

Deployment pace

Compared with the previous quarter, how is your interest in defense tech?

Defense tech

Compared with the previous quarter, how is your interest in AI applications?

AI applications

Compared with the previous quarter, how is your interest in AI infrastructure?

AI infrastructure

Compared with the previous quarter, how is your interest in AI labs & foundation models?

AI labs

Compared with the previous quarter, how is your interest in fintech?

Fintech

Compared with the previous quarter, how is your interest in healthtech?

Healthtech

Are there any verticals not asked about that you're increasingly interested in?

Other verticals

Which of the following best characterizes your view of the current state of the venture economy?

State of the venture economy

What do you think are the chances that AI-related valuations (public & private) will be down by 25% or more 12 months from now?

Bubble odds

What was the most competitive early-stage round over the past 3 months?

Most competitive early-stage rounds

What was the most competitive late-stage round over the past 3 months?

Most competitive late-stage rounds

What would you do with shares in OpenAI?

Buy/Sell/Hold OpenAI

What would you do with shares in Anthropic?

Buy/Sell/Hold Anthropic

What would you do with shares in Stripe?

Buy/Sell/Hold Stripe

What seed to Series B company are you most bullish about being the next generational startup?

Next generational startup

What percentage of startups backed by top VCs do you think are meaningfully misrepresenting their financials to investors?

Financial misrepresentation

If you could be an investor in any multi-stage private markets firm besides your own based on their investments in the last 12 months, who would it be?

Best multi-stage firm

If you could invest in any venture fund started in the last 5 years, what would it be?

Best emerging fund

What is the most overpriced startup that has raised money in the last 12 months?

Most overpriced startup

What is the most underpriced startup that has raised money in the last 12 months?

Most underpriced startup

Dual-valuation & tranched rounds

If you could tell a new college grad to go into any profession today, what would it be?

Advice for a new college grad

The story we should be chasing