AI Gold Rush or Bubble? Investing in AI in 2026
- Stephanie Roulic
- Oct 6
- 41 min read
Every founder is feeling the pull of AI right now—whether it’s excitement about new possibilities or pressure to figure out how to integrate it before competitors do. But is this a true gold rush moment for startups, or just another hype cycle waiting to deflate?
At Startup Boston Week 2026, we brought together investors and operators who have lived through multiple waves of tech booms and busts to dig into that exact question:
Maia Heymann - Cofounder & General Partner, Converge VC
Chip Hazard - General Partner, Flybridge
Jonathan Corbin - CEO and Co-Founder, Maven AGI
Heena Purohit - Director, AI Startups, Microsoft
Habib Hadad - Managing Partner, E14 Fund
They unpacked where AI really stands today, what makes an AI startup investable, and how founders can build companies that endure long after the hype fades.
From the importance of speed and execution, to redefining what a moat looks like in an AI-first world, the panel shared hard-earned insights and practical advice to help founders cut through the noise and focus on building lasting value.
5 Takeaways
1. AI Is Everywhere—But That Doesn’t Mean Every Startup Will Win
Chip Hazard put it plainly: “There will be no such thing as an AI company—every application will become an AI application.” The panel agreed that while AI is being infused across industries, we’re also seeing the classic hype cycle at work. The long-term potential is transformative, but in the short term, adoption challenges and overpromises are creating friction
2. Founders and Teams Matter More Than Metrics at the Early Stage
For pre-seed and seed companies, the investors emphasized that it isn’t about spreadsheets—it’s about people. Maia Heymann highlighted the importance of technical depth, vision, and the ability to adapt as markets shift. Hazard added that grit, resilience, and customer obsession are the real signals of a fundable founder
3. Moats Are Temporary—Speed and Execution Are the Real Edge
From proprietary data to unique integrations, panelists broke down the idea of “defensible moats.” The consensus? In AI, most technical advantages erode quickly. Jonathan Corbin noted that moats shift over time: “The only moat is speed—your ability to ship rapidly, learn, and evolve.” Habib Haddad added that the ability to attract talent early is itself a form of defensibility
4. AI Adoption Is Fast, But Trust and Ethics Can’t Be Ignored
Enterprises are adopting AI faster than expected, often experimenting aggressively to avoid being left behind. But speed must be balanced with responsibility. Haddad stressed the role of simulation, careful rollouts, and uncertainty detection tools to ensure that moving fast doesn’t mean breaking trust—or harming users
5. The Next Wave: From Physical AI to Human-Computer Interaction
Looking ahead to 2026 and beyond, the panelists pointed to several under-hyped frontiers:
Human-computer interaction (beyond voice, potentially even “thought-to-text” interfaces).
Robotics—still awaiting its “ChatGPT moment,” but with enormous potential in manufacturing and agriculture.
Applied AI at the enterprise layer—where companies that deliver working solutions will outpace enterprises trying to build from scratch
Final Word
So, is AI a bubble or a gold rush? The panel’s verdict: it’s both. Valuations may be frothy, but the underlying technological breakthroughs are real—and the winners will be those who combine speed, thoughtful execution, and resilience in the face of rapid change.
Full Transcript Below
Want to revisit a particular quote or share with a teammate? We’ve got you covered. Read the transcript here:
Heena Purohit
[ 00:04:32 ] Good morning.
Heena Purohit
[ 00:04:34 ] Quick show of hands, are we in an AI bubble?
Heena Purohit
[ 00:04:41 ] Are we in a hype cycle?
Heena Purohit
[ 00:04:45 ] Or is this all reality and this is real value?
Heena Purohit
[ 00:04:54 ] Awesome. I think I covered all hands. This was your morning stretch, because you are here on a Friday morning at 9 . 30 AM. Thank you all for joining. Super excited to have this panel today to talk about where we are at.
Heena Purohit
[ 00:05:10 ] Sorry.
Heena Purohit
[ 00:05:12 ] where we're at for AI in 2025, September 2025 and beyond. I'm Heena Purohit. I'm the Director of AI Startups at Microsoft. We help founders across from founders writing their first line of code to founders that are building scale-ups. like what we have on the panel today. And I couldn't be more excited because we've got a really exciting group of veteran investors who are investing, who've seen multiple rounds of hype cycles and real value both. And then we've got a founder who's raised a successful company and is executing in this. in this AI era. So I want to have the panel just start with introducing themselves. Chip, I'll start from you. If you could just introduce yourself. And since the question we're exploring today is, are we in an AI gold rush or bubble? What are your thoughts?
Chip Hazard
[ 00:06:08 ] That sounds great. Hey, everyone. Chip Hazard. I'm the founding partner at Flybridge. We are a preceding seed stage venture capital firm here in Boston and New York. About a billion of assets under management. Just raised our seventh seed fund. We're 100% focused on AI. at the enablement layer, the agentic application layer, and new experiences that couldn't be contemplated before AI. I've been investing in the Boston area since 1994. I was a partner at Greylock before starting Flybridge. My first investment in the industry closed the day that Netscape was launched. So I've seen a few cycles, good and bad.
Chip Hazard
[ 00:06:49 ] And the high level is obviously we're all in on AI. Couldn't be more excited.
Chip Haz
[ 00:06:55 ] It's a field we've been investing in. We called it different things. We called it modern data. We called it machine learning. We called it applied AI. And when ChatGPT came out in November 22, I said to my team, 'This is all we're going to need to be doing.'
Chip Hazard
[ 00:07:11 ] At the same time, I said that there will be no such thing as an AI company because it's going to infuse every aspect of business and society and culture, and every application will become an AI application. And so the distinction of saying I'm building an AI company will soon be. less relevant. But today it is relevant because it's still hard. And the reason you have this bubble hype question is the promise in 10 years is amazing. The classic line is in the short term we tend to overestimate the impact of new technology. In the long term, we tend to underestimate the impact of technology. And I think that's 100% right here. And you're starting to see a little bit of the trough of disillusionment with the MIT study. And people aren't getting applications into production.
Chip Hazard
[ 00:07:56 ] But if you power through that, I think the promise on the other side of a system that can work hand in hand and in concert with humans to really fundamentally unlock human potential is an amazing opportunity. And humans are the only species that uses tools to make us smarter, and AI is 100% a tool that will make us smarter. So we couldn't be more excited about the opportunity set.
Heena Purohit
[ 00:08:19 ] Thank you. Maya?
Maia Heymann
[ 00:08:21 ] Great.
Maia Heymann
[ 00:08:23 ]Maia Heymann. I am, like Chip, founding GP of an early stage firm here. In Cambridge and in Silicon Valley called Converge VC. I started the firm with my fellow partner, Nilanjan Abamek, and two years ago we added a third partner in Silicon Valley and she She's a five-time entrepreneur and her fifth company we backed in our first fund. So we brought her over to the investing side. I mentioned we focused at the early stage. So we are pre-seed, seed, and selectively we'll do some A. uh rounds um the definition of an a has been uh changing rapidly as as we all know um we invest uh like like chip at flybridge we invest at the application layer we go down the stack to enabling infra and then we have a third area that we invest in which we've been doing since our very first fund, which
Maia Heymann
[ 00:09:23 ] we call Digital Meets Physical. The physical world is becoming more automated, and there are some very interesting opportunities that we have and will continue to invest in there.
Maia Heymann
[ 00:09:39 ] I think the question bubble, right? So yeah, it's both. Yes, there are aspects of bubble in some regards, but we've never seen growth like we've seen.
Maia Heymann
[ 00:09:57 ] You know, bubbles are not all bad. Because companies get funded that maybe otherwise wouldn't have been funded, that deserve to be funded. So I think separating maybe difficult valuations from funding, those are two different things.
Maia Heymann
[ 00:10:21 ] I started my venture career in 98 in Silicon Valley. and was there for 10 years and saw the run up and then saw the deflation, which was actually a very good way to start a venture career because you see the dark side of. too much capital, the dark side of inefficient spending, et cetera, but you also see the incredible cohort of companies that get funded that in a downturn wouldn't have otherwise gotten funding.
Maia Heymann
[ 00:11:00 ] I used to say, you know, 10-year cycles, five-year memories. I think now it's like five-year, or maybe it's like, you know, five-year cycles, two-year memories. So, you know, there are patterns, bad patterns, that are getting repeated because...
Maia Heymann
[ 00:11:20 ] Amnesia, but even in that, there is still goodness and still tremendous, tremendous opportunity that we're very excited about.
Chip Hazard
[ 00:11:30 ] As a more senior venture capitalist, amnesia could be a good thing.
Habib Hadad
[ 00:11:36 ] So it feels like I've been waiting for this moment for maybe 20 years or so, for this bubble. My name is Habib Haddad, and I'm currently a founding partner, founding GP at E14 Fund. So we're a fund that has a deep partnership with MIT. I'm based here. My partner, Calvin, is on the West Coast. We invest in AI and deep tech. Starting from the infrastructure layer, whether it's hardware and accelerators or AI systems, all the way to the middle layer with companies like Mosaic ML and moving to the top with companies like Replit and Maven AGI and others. But I started my career actually as a founder. About 25 years ago, I moved here to the US as an immigrant student from Lebanon. And right after graduation, which was in computer vision, I moved to Boston to start a company with MIT folks. And at that time, computer vision was this very clunky thing that you used to use these fixed pipelines. So it wasn't GPU access. It was this graphics hardware. I worked on that company with a few founders. We got fired by the VC[s], so all of us left.
Habib Hadad
[ 00:12:37 ] Then I went and joined ATI. Some of you might know ATI. You might recognize the logo. ATI was the the first GPU of AMD. So I had the luck to work on the first GPU of AMD, which was called R600, and the first unified architecture. And at that time, at ATI, we used to always have rivalry with NVIDIA. And at some point, someone said, 'Oh, NVIDIA has this CUDA thing, but these suckers are using it for genomics. Like, you know, we're the real graphics people.' And so, of course, sure enough, you know, that kind of became history. And music sometimes is good. And then after that, I started a company in AI and linguistics. And again, it was early in the days of AI and linguistics. It was a transliteration tool for right-to-left languages. I ended up selling that to Yahoo in 2011. But that also got me into the early world of language models and thinking about that. And so, coming to today, it feels like, to me at least, and to many folks I talk with at MIT and outside, it's been a moment that we've been waiting for— a moment that new theories could happen but hasn't yet happened.
Habib Hadad
[ 00:13:42 ] There's actually a nice paper by the head of AI at Cohere called 'The Hardware Lottery'. And The Hardware Lottery talks about how, basically, you have all these inventions and all these scientific breakthroughs, but you have to wait for the right hardware for this scientific breakthrough to break out. You know, in a way, like we are in a bubble when you think about capital allocation and maybe company creation, but we totally are not in a bubble when it comes to scientific breakthroughs that deliver real ROI to customers and markets. I think we're just in the beginning of that. So I'm super excited at the same time I'm also worried— as a father of two young kids at home, looking at them— I was thinking the other day, there was one of the founding fathers of AI, Seymour Papert. Seymour Papert used to be at MIT. He's well-known for his work on education. And in one of his books, he said, 'You can't think about thinking without thinking about something.
Habib Hadad
[ 00:14:45 ] And it's like an interesting, deep thing. But then I was like, well, actually now with Chat GPT, you can think about thinking without thinking at all. And it's one of these things. And there was actually a study from the Media Lab around a student who showed they were doing brain measurement, and they showed that cognition would go down as students would rely more on child GPT. So these tools are amazing. We have to be careful how to use them. Again, as a father of two and young kids, I'm also thinking always like...' What's responsible in that? What's ethical AI? As investors, I think it's a big role for us to not just be so enamored with the bubble and think about what's our role in society.
Habib Hadad
[ 00:15:27 ] There's going to be disruptions, but we're going to also think carefully about that. So we invest in the physical layer that will disrupt labor and the knowledge layer that will disrupt knowledge. But at the same time, we're always thinking about how we make sure that this is done in a way that's ethical, but also at the same time not getting into these corner cases where you're really kind of causing mental health. and all these issues that are going to be very hard to reverse from.
Jonathan Corbin
[ 00:15:50 ] Great. And I'm fortunate enough to be able to work with Habib as one of our investors in our seed round. And then, of course, we get a chance to work with Hina as well at Microsoft. We have a good partnership with them. I'm Jonathan Corbin, co-founder and CEO of Maven AGI.
Jonathan Corbin
[ 00:16:07 ] I started off my career working for a company called Sybase many years ago, writing code. And so Habib and I are dating ourselves here.
Jonathan Corbin
[ 00:16:16 ] And I ended up going to go work for a local Boston company called Uniqlo. It was very early on in the marketing automation space.
Jonathan Corbin
[ 00:16:22 ] And I like to joke with my co-founders that was the beginning of my love affair with go-to-market software.
Jonathan Corbin
[ 00:16:28 ] Where I really became enamored with the idea of using technology in order to improve user experience and to help them to provide value for the customers they're interacting with. I worked for Unica, spent a few years at a big ad agency called WPP, one of the largest ad conglomerates in the world. We were working with brands like Coca-Cola on creating their 10-year digital strategy and working with a number of technology brands. Ended up working with one of the market leaders called Omniture. There was a leader in digital analytics. I went to work for them just as they got acquired by Adobe. Ended up spending five years there.
Jonathan Corbin
[ 00:17:03 ] amazing ride. We went through a transition period of going from box software to cloud-based technology. And so it was really great to see that evolution in terms of compute, what it unlocked for people and the capabilities that it unlocked for the companies we were working with. Left to go start a company called Veer. It was a parking app. And from that, we understood destination intent of where people were going, and we could engage with them around that. Ended up selling the technology for that, and I joined this company. Some people called it Big Purple. The name of it is Marketo. And it was a market leader in the marketing automation space. We ended up getting acquired by Vista Equity Partners, who then sold it to Adobe. And I was like, 'Guys, I left.' I wasn't trying to come back again.
Jonathan Corbin
[ 00:17:49 ] I joined a company called Sprinkler.
Jonathan Corbin
[ 00:17:53 ] Startup in New York that was focused on using social data along with first-party data in order to personalize experiences and so had a really great ride there, going from 100 to 300 million. Then, in 2019, I heard from this company in Boston called HubSpot.
Jonathan Corbin
[ 00:18:09 ] Back then, I was chatting with Brian Halligan, and he was like, 'We're the only market leader you haven't worked for. You've got to come work for us.' And so I joined in 2019, and it was a really fun place to be. At that point, HubSpot was really a single product company, very focused on inbound marketing.
Jonathan Corbin
[ 00:18:27 ] Over the next four years, we became one of the five fastest growing B2B SaaS companies ever after a billion dollars in revenue.
Jonathan Corbin
[ 00:18:34 ] Definitely a cornerstone of the Boston tech community. As we continued to grow from a customer standpoint, we went from 700 million to 2. 2 billion in revenue. We went from under 2,000 people to about 7,700 people and 220,000 customers. And so my team went from 238 people to 1,000 people. And Kate Buecher, who's the CFO, comes into my budget meeting and says, you guys are fucking killing it. And we're like, yeah, okay, we made it.
Jonathan Corbin
[ 00:19:03 ] And she's like, but because you're doing so well, I need you guys to figure out how you can decrease your costs. Figure out how to grow in a nonlinear fashion.
Jonathan Corbin
[ 00:19:13 ] And so my team's looking at each other around the room, how are we going to do that, right? We have this headcount model that's very predicated on the number of customers we're adding. We work with them, the certain number of customers we help, certain amount of revenue we bring in. And so she really pushed us to a place where we had to rethink the way that we're engaging with our customers.
Jonathan Corbin
[ 00:19:32 ] And so we did the things that you would normally do in a company when you're trying to conserve costs. Great, we're going to build near-shore teams. We're going to change the way we engage with our customers. But you didn't have the same impact that you were looking for with your customers. You couldn't drive the same amount of consumption and usage and adoption that you needed. And so we said, 'Amazing,' we need a technology platform that's going to enable us to do this. We're going to build it. It turned out we were always just about 18 months away from having something that was going to be useful. And so there was a lot of very heated conversations. And at some point there was a running joke. When I'd go into the room, Yamini, who's the CEO now, she'd be like, 'Okay, great. Jonathan's going to talk about. the thing that we need to build in order to enable our customers to better get value out of our product.'
Jonathan Corbin
[ 00:20:17 ] And so I said, 'Okay, I can keep having this running joke with Yamini or I can leave and go start a company.' So I started chatting with a couple of friends of mine. One was leading the Applied ML team at Stripe. He was trying to solve the exact same problem there. Another one was at Google. He'd built the Google News team from zero to a couple billion users. And the way they did that was through personalization. They said, 'How do we personalize every single user's experience in order for us to feel like this is unique to them?' And we said, 'How do we apply that same component to the way that companies engage with their customers?'
Jonathan Corbin
[ 00:20:53 ] And so the question at hand here is: Is this a bubble? And the answer is no. The opportunity for value creation is unique to this time and place.
Jonathan Corbin
[ 00:21:04 ] What we were trying to do before hadn't been possible with other technology in the past. For the past 20 years, I've been working at market leaders who are trying to create personalized experiences, and we couldn't do it.
Jonathan Corbin
[ 00:21:15 ] We came out of stealth in the middle of last year. We raised a couple of rounds of funding and we're working with over 50 different brands, creating these unique personalized experiences, starting customer support for every single customer.
Jonathan Corbin
[ 00:21:28 ] The companies that we're working with, we're answering over 93% of their customers' inquiries with no people involved. That allows the people who have been dealing with these frontline problems, even some cases very complex problems, to be able to focus on the customers who need their attention. So I believe this is not a bubble. The opportunity for value creation is larger than we've ever seen in the past. I'm very excited about it.
Heena Purohit
[ 00:21:52 ] Love it. A consensus on the opportunity that lies ahead of us. And all of you are tapping into this opportunity in different ways. Maya, I'd love to start with you this time because, as an investor, you're always trying to sniff out what's real versus what's hype. And you're doing it in a data-driven process, and you're looking at hard metrics. What are some of the metrics that you look for when you're evaluating AI startups? Because we have a lot of founders in the room, and they'd like to hear what's driving those lofty valuations they see.
Maia Heymann
[ 00:22:25 ] I'm smiling because we're in the thick of about to make an investment in a company that my partner refers to as it's three minutes old. The founder has just left a very large prominent company. And there are no metrics, zero, right? He has an idea, a very good idea. And so what we look at are several things. First, the founder, the technical depth and vision.
Maia Heymann
[ 00:23:00 ] of the founder and his or her ability to build. And that's...
Maia Heymann
[ 00:23:07 ] a multitude of things, prior experience, domain expertise, understanding of where the technology is likely to evolve at competition.
Maia Heymann
[ 00:23:23 ] That's one, you know, two things matter, team and market. How good is the team and how big is the market potential in our business, right? As we think about driving returns for investors.
Maia Heymann
[ 00:23:36 ] It sounds simple, but if you boil it down, because team equals product.
Maia Heymann
[ 00:23:43 ] and you know But you need the two because you can have a phenomenal founder But the mark they chose the wrong market or the markets not you know it's it's five years out Not three years out or what have you?
Maia Heymann
[ 00:23:59 ] Then when you, so that's one.
Maia Heymann
[ 00:24:04 ] answer when you when we are evaluating companies that have product and are beginning to generate revenue that has become very interesting because we think about the the risk of quote false positive that there's impressive revenue, but is it sustainable? Is it sticky? And that's particularly true in this moment in time where we see enterprises so voraciously looking to buy technology that's going to help them not be leapfrogged in their own businesses that I don't know, we don't know, but there are instances where we've thought, huh.
Maia Heymann
[ 00:24:54 ] that revenue may not be here two years from now, one year from now. So the quality of early revenue can be very compelling, but there's a but that sometimes, as we call it, it's a false positive.
Maia Heymann
[ 00:25:16 ] So stepping back, I sometimes say beauty is in the eye of the beholder. What we think makes a compelling investment will be different from what Chip and Habib— or it'll be similar. Because every investor is coming to opportunities with their own investment thesis, their own understanding of the buy side, the buyer, adoption rapidity. And then also their own investment thesis around— I'm going to say sustainability. I don't mean that in an environment. I mean, you know, the longevity of that technology. And it is.
Maia Heymann
[ 00:26:03 ] I know my colleagues will agree with this. The pace of change is mind-boggling. And it actually makes investing in this climate challenging because that we don't know how rapidly things, I mean we do, we are seeing how rapidly things are evolving and changing. And so to really get our arms around what has longevity. And that's why I go back to the team because a deeply astute team will know when to change, how to continually augment product and morph the product roadmap.
Maia Heymann
[ 00:26:49 ] in response to the competitive landscape and in response to customer demand. So those are my thoughts.
Jonathan Corbin
[ 00:26:59 ] I'm very excited to ask a question to the rest of this panel. I get asked all the time, how do I think about moats? How do I think about sustainable, durable moats? I'm very curious for the investors on the panel, how do you think about those? What you said is it sounds like it's about the founders.
Maia Heymann
[ 00:27:20 ] It is about the founders, and I think, so it depends where, you know, what the kind of technology is, but, right? I mean, there are different modes in different industries, different verticals, horizontal, right? But I think...
Heena Purohit
[ 00:27:36 ] With when we think of at the application layer Unique data, I would say four things a truly unique data set proprietary models deep integrations across across the you know deep integrations across multiple constituents right that can be functions can be it can be many things but the deep integration and then the fourth is um the feedback loop where there's constant improvement and and i think those four things create defensibility absolutely and i'd like to keep it i just expanded back where chip i'd love to hear since each of you have your own flavor so could you talk about the metrics you're looking for and what's driving your thesis and then we can talk about moat as well within that Yeah, I think for, it's always funny when you talk to early stage venture capitalists at the pre
Chip Hazard
[ 00:28:37 ]-seed and seed stage. you'll basically hear a very similar answer. I haven't opened a spreadsheet in, I don't know, 25 years. And so metrics, that's somewhat facetious, but the metrics at the very, very early stage are non-existent. And so you do tend to look for founder market fit, founders that you want to be in business with. You tend to look for very big markets and then you're trying to play out the chessboard in a way that you think you can build an enduring company. And so, for a founder that's raising a pre-seed or a seed round, how do you communicate that? So when you think about what are you looking for in a founder, you're looking for passion for the opportunity. So a lot of the questions we'll ask is like, of all the things in the world, you're incredibly smart, you're very talented. Of all the things in the world, why are you doing this one?
Chip Hazard
[ 00:29:24 ] What's the earned secret you have, the unique insight you have? We tend to love founders that have done a tremendous amount of customer discovery and they have an intimate understanding of their early customers. You tend to look for founders who, because of that passion and that drive, are going to put in the work. It's brutally hard starting and building a company.
Chip Hazard
[ 00:29:42 ] Do you have an evidence and ability to work very hard? It's a wild roller coaster. Do you have evidence in your background of like serious perseverance and grit? And so part of the reason why you see so many founders who are successful who come from the immigrant community is because they've already demonstrated massive resilience and perseverance to come to an entirely different world and try to make their way.
Chip Hazard
[ 00:30:06 ] And then if you put all those together, you tend to get a founder who has this unique ability to have that infectious enthusiasm that gets other people to want them to win. And if other people want you to win, they're going to bring you customers, employees, partners, and capital. And so all those things together is what you're really looking for when we talk about founders. Those are some of the characteristics that you're looking for founders. In terms of the market, I personally tend to like businesses that sort of fall into one of two categories from a market perspective. They're either going after just a massively large market and they're going to do so in a very disruptive new way. And so for 16 years, I've been on the board of a company called MongoDB, which is a big database software company. It was five people and an idea when we invested, but it was a $35 billion market growing. You know, double-digit growth rate on a sustainable basis. And that felt like there was going to be a lot of running room. And today we're $2 billion in revenue and 5,000-plus employees, and the market's a $100 billion market. So a market that's very large and growing is obviously a huge opportunity.
Chip Hazard
[ 00:31:10 ] Then markets that don't exist that you can completely create in the the classic Peter Thiel zero-to-one kind of market— and those those are super exciting. They're much harder to pitch as a founder, and so we look for founders who can articulate a vision for that market. How that market took it, could develop, and what are the chess moves they will make to go down the board? That unlocking one segment will unlock another, which will unlock another. And so, having a really clearly well thought out strategy for how you see a market developing, the role you can play in that market, and that communicates. Not only large aspirations, which I think are very important to venture capitalists, but also communicates a deep, deep understanding and level of strategic thinking about how to go after it. So those are the things when... when venture capitalists talk about founders and markets and what we're looking for, as you move down that journey and you're maybe raising subsequent capital where they are thinking about metrics.
Chip Hazard
[ 00:32:07 ] This issue of durability of business model, of is it real revenue, is critically important. And so anything you can be doing to show massive engagement across your customer base is what matters more than anything at this stage.
Chip Hazard
[ 00:32:23 ] And so we have a company that just raised a... very successful Series A, you know, 20 customers, $500 ,000 of revenue, and they raised close to $100 million valuation. And so you wouldn't think that would be the kind of metrics that would get you to raise a Series A. But if you look at those 20 customers, they're using the product like every day, all the time, mission critical. real production use cases. And so the more you can communicate that and show that and then back that up with a pipeline of customers that look like those initial customers, obviously that makes your fundraising job much, much easier.
Chip Hazard
[ 00:33:00 ] And then I got off track. Oh, moat. I was thinking about your list, Maya. The proprietary models I don't think is a moat anymore.
Chip Hazard
[ 00:33:08 ] Integrations I don't think is a moat anymore. It's just super easy. If you're building integrations and other software, you can... You can use MCP, you can use platforms, you can use cursor or cloud code, and write integrations yourself in a way that you never could. I do think the feedback loop is incredibly important. And we broaden that to think a lot about the user experience. And so, what is it about your user experience that unlocks? this magical change and so the the first generation of ai applications felt a lot like putting chatbots onto existing sas applications and that that by and large and this is a b2b setting that by and large strikes me as being kind of brain dead and and so you have this powerful new tool that can actually do work for you and so you need to think about it in an entirely different way and that's an entirely different user experience and the people who are able to capture that user experience in a new and innovative way and in one that brings feedback back into the system so the system recursively gets better. because I do think data will play a role in that.
Chip Hazard
[ 00:34:08 ] would be one source of competitive moat. But we're tending to look at moats as not any one big thing in the application space, but rather the concatenation of like 10 little things that in the way you've implemented this is just incredibly thoughtful and smart. And that's going to build a delightful consumer and user experience, and give you the running room to develop the momentum to build a more sustainable business.
Heena Purohit
[ 00:34:35 ] And it's always great to hear when there's differences in the moat. So Habib, I want to hear from you, especially since you've invested in a lot of AI apps, but the deepest corners of deep tech. And we're in Boston, so there's a lot of technical founders. What are you looking for, and how are you thinking about moat?
Habib Hadad
[ 00:34:52 ] Yeah, I think, I mean, the question of mode is very open-ended because it's very idiosyncratic to each company and the flavor of each company. So when you look at things down in the stack, it's the science and the work that they've done. And often we end up investing in folks who have spent 10 years working on things. I'll give you a couple of examples. One of the latest investments we've done is in a hardware company trying to decouple memory from GPUs, so really increasing the GPU utilization by 10x very quickly. And that's a company where the moat is obvious within the description. So the moat is the science. It is a team who has the best team at building this. You look at the company like Maven and Jonathan and Sammy and Eugene— kind of some of the world's best leaders in their own spaces. Some of the early days were actually integration mode. I don't know if you disagree, but some of your early mode was integration mode. But over time, that flywheel effect Once you integrate and once you execute, flawless execution, deep integration, you get into customers and guess what? You start having that usage and just that kind of flows back.
Habib Hadad
[ 00:35:56 ] In some cases, the mode is something you have to create. In some of our physical AI companies, there's not enough data for robotics, for example, to train robots.
Habib Hadad
[ 00:36:08 ] One of our very exciting company in manipulation and dexterity is basically building its own mode through simulation. So they're the world's best at doing simulation, and they're creating this chunk of data that they're basically using. So I think, of course, there's also the founder mode, the relationship. You have founders who are in an industry who know it very well. We have a founder in a company called Catrix. He works on software for FDA-regulated devices.
Habib Hadad
[ 00:36:38 ] He loves talking ISO. This is the key regulation. He's the only one in the world who can make regulation sound sexy. He loves it. And that's a mode in itself. But I agree with Chip. It's a concatenation of different modes, and you have to put them all together for you to see, to believe that that's going to be it. again the early stages you're doing a lot of these early stage investing it's hard and i think it's just getting harder and harder and in a good way it's exciting i mean when you see companies like cursor or delve in one of our portfolio companies these very fast zero to 20 to 50 million to 100 million and they're undergrads or the people who are in the 20 year old founders so the kind of the whole the whole game has changed and so that's kind of also very exciting in its own Absolutely.
Heena Purohit
[ 00:37:25 ] And I think, Jonathan, since you asked the question, I kind of want to go off script with you and hear the flip side. Because you announced your $15 million Series B round. So this is the second time in this hyper-competitive environment. that you convinced investors. So what metrics do you think set you apart? And what was your moat, if you feel comfortable sharing?
Jonathan Corbin
[ 00:37:49 ] Third time. Actually, we did a seed round, too. There you are. The way that we think about it is that you have short-term moats and long-term moats. And so short-term moat, I think Chip mentioned, Aviv mentioned, hey, you look for founders who have experience in this space. It gives you an unnatural advantage, right? So for us, when we went out and we started raising money, we were solving a problem that I witnessed firsthand. I lived it every day for many years.
Jonathan Corbin
[ 00:38:16 ] We had a team that had solved this problem in different forms. We had an expert in personalization. We had people who had been using transformer models since their early days back in 2017 and even before that. We had people who were trying to solve this problem in other places. And so, I think, the team was really important for us as we started looking at it. That was our short-term moat. We had an understanding of the problem. We had a team that could build the solution for it. Now, as you look at the long-term and you look at how do you create these durable, sustainable long-term moats, there are transitory steps that you take along the way. Habib mentioned, hey, great.
Jonathan Corbin
[ 00:38:53 ] Integrations was key for us, but that's transitory, right? You have to be able to access the information in order for you to be able to deploy AI agents at scale on top of it.
Jonathan Corbin
[ 00:39:03 ] The next step after that is, okay, great. Now that you have the data, what do you do with it? Well, you build a platform. And so that moat changes over time. And if you look at Google early days, right, they had a proprietary algorithm.
Jonathan Corbin
[ 00:39:17 ] Like, what was their sustainable moat? Well, it was transitory. First, it was actually the ability to be able to get in front of users and get people to use your product, right? So it was distribution. And then it became data. And so I don't think there's like one single moat that's sustainable over the long term. I think it's very hard to have a technical moat today. If you looked at OpenAI when they first came out, you'd say, 'Hey, wow, you have an amazing moat. You have the best technology in the market.' Then one day, out of nowhere, someone in China came out with a model that completely blew everything up. So I think it's very hard to have a technical moat right now. I met with the founders of XAI the other day, and they're in an incredibly competitive space. You're building frontier models. It's becoming more competitive by the day. It's like, guys, how do you think about moats? And they said, the only moat is speed. Your ability to be able to ship rapidly, evolve, understand the challenges, move past that. So I think going back to my original statement, there's short-term and long-term moats, and I think it's important for you to think about it that way.
Habib Hadad
[ 00:40:25 ] I think we also didn't talk about the hiring mode. I think that's part of execution. Just being able to attract a massive amount of talent is what's going to help you with the speed and execution. So we look at a lot of founders early on, even if they haven't hired. Who's going to be your first hire? Are you friends with them? Have you identified your first five, even if you can't afford them? So having that innate ability to attract talent and convince people to join you is a very big deal.
Heena Purohit
[ 00:40:53 ] Absolutely. And it all ties back to this is why early on the founder, do they have a vision that can gravitate people is incredibly important. I want to ask one last question and open it up. So folks, think about what you want to ask. But our session today is about 2026. So I'll just toss this out. What are some areas that y'all are excited about? Some areas y'all are investing in for 2026?
Chip Hazard
[ 00:41:23 ] Who wants to go first?
Habib Hadad
[ 00:41:24 ] I think we're probably all looking at the same areas. But maybe an area that I think is under-hyped or not as much talked about is human-computer interactions. And so I think the next wave of AI agents is going to require us different modalities of talking and communicating with these agents. You think about today—keyboard— probably going to die. Voice is becoming a very interesting modality. But what's after voice? So one of our companies just actually announced something on Monday: it's basically communicating at the speed of thought. And it's basically this device that you put as a bandana on your head, and it detects what you're thinking about. Well, maybe thinking about is an exaggeration, but you're silently speaking. So you probably can say now, he's full of shit. Maybe try it in your head. Well, if you had that on your head, you actually would go on your phone. So this is a device that will be available. And it's actually, funny enough, that technology was invented by the same inventor arnav at the media lab six seven years ago but at that time it was pre-llm it wasn't as interesting because you had to train so many different words but now you don't have to so now you're able to actually do that so
Habib Hadad
[ 00:42:40 ] i think that's an interesting one there's like all these things about what's true AGI and in that there's more than one modality to think about not just text so we're focused a lot on the physical AI as well when manipulation dexterity we're thinking out of the smell so all these areas to truly unlock
Habib Hadad
[ 00:43:00 ] human potential in the digital world is an area that I think maybe 2026, some of it will be there. Maybe it's a bit later, some of it. But I think that's going to have an interesting underhyped area for us.
Habib Hadad
[ 00:43:11 ] Exciting.
Maia Heymann
[ 00:43:13 ] You want to go? You're full of shit.
Chip Hazard
[ 00:43:20 ] I always struggle with this question. Not that we don't have a point of view on the thesis, but over the years, it always felt like you could have proprietary insights as a venture capitalist. And today, like all the ideas that you would say, oh, agentic applications, and everyone would be like, oh, Jan, you're full of shit.
Chip Hazard
[ 00:43:38 ] So it does go back to a lot of this, you know, where we light up is when a founder walks in and they're telling a story that is consistent with our point of view of the world.
Chip Hazard
[ 00:43:50 ] When we first sort of put together our AI thesis, I would have told you that I thought the enablement layer, so all the tools that you need to build high-quality production AI applications was going to be an amazing investment area.
Chip Hazard
[ 00:44:04 ] I still think it is. The problem is, everyone else decided it was an amazing investment area, and every little niche of observability, evaluation, simulation environments, or runtime environments for models is wildly competitive. And so we've moved up the stack. The other part of our thesis that was wrong: when we first put it forward, we thought every enterprise was going to build their own AI stack and build their own custom applications. And it turns out it's really, really hard to do that. And the average enterprise doesn't have skills to do that. And so we've moved more towards the application layer, more companies like Jonathan's, because... If you show up at a customer with a solution that actually works, that's 10 times better than what they can build themselves, even if they have all the modern tools and even if they have all the modern development environments. And so I think the application layer is going to be a source of significant growth and opportunity in the coming few years.
Maia Heymann
[ 00:45:02 ] An area that we are continuing to watch very closely and we're not, so, you know, put simply, when will we have the chat GPT moment in robotics? And we're not, you know, it's incredibly hard to.
Maia Heymann
[ 00:45:24 ] To combine the intelligence with the actuator, right, the actual physical. It will get solved, we're not, so it will. It will happen. But I think we're a ways away from that. But there are some just brilliant people working on it. And that, to us, is just— if you think of where the application of AI has the most impact. It's in the physical world, which has the most inefficiencies. And so automating certain, you know, manufacturing, ag.
Maia Heymann
[ 00:46:09 ] So there's huge opportunity. And from an investment point of view, if you take...
Maia Heymann
[ 00:46:21 ] let's say agriculture, which is massive and very hard to invest in. But through automation, you can disrupt the incumbents and own that end market. And that has incredible investment dynamics. There are some instances where that is happening now.
Maia Heymann
[ 00:46:53 ] We pick our spots very carefully.
Maia Heymann
[ 00:47:00 ] I'm a little bit in the weeds, but stepping back, I think continued automation of the physical world is something we're not there yet. We will get there, and there are very interesting spots to invest in.
Chip Hazard
[ 00:47:14 ] We're going to have a humanoid robotic that digs literal moats.
Heena Purohit
[ 00:47:20 ] I'll flip the question again for you, Jonathan, especially because as we talk about AI apps, one question that comes up when you're trying to sell to enterprises that have large armies of AI teams is: they want to build it themselves. Have you come across that build versus buy question? And if so, how do you overcome it?
Jonathan Corbin
[ 00:47:42 ] It's funny because, you know, when we first launched Maven, everyone was starting to hear about ChatGPT, just starting to use it. And I think you had these data science teams that were really excited. You had product and engineering that were like... this is the thing that i'm going to do it's going to have a massive impact on our business this is my path to promotion so you would go in they're like i've used chat cbt i've seen this before i can do it don't worry about it And what we saw is three months, six months later, people coming back to us and they're like, oh, wow, I can't just, you know, take a dump truck full of data, drop into a context window and expect real results.
Jonathan Corbin
[ 00:48:24 ] And the companies that we're working with, one of the companies we're working with is TripAdvisor. They have 5% of the world's population visiting their website on a monthly basis. You can't afford to give wrong answers. You can't afford for it to be an experiment on someone's path to a promotion, and people are realizing that.
Jonathan Corbin
[ 00:48:42 ] Folks that have come to us, they said, hey, we think we're going to go off and build this. We say, here's the lessons we've learned along the way. Let us know how we can help. And we continue to see folks coming back to us, large publicly traded companies. They have the ability to be able to go out and build this, and I think it becomes very similar to the Klarna case study that they put out. I don't know if any of you have read that.
Jonathan Corbin
[ 00:49:06 ] Klarna came out about a year ago, and they said, 'We've been working with open AI engineers for a year and a half. We don't need people and we don't need technology anymore. Bye-bye Salesforce. Bye-bye Workday. You're unneeded here.' See ya. And then seven months later, they came back and they were like, 'Oh my God, this is a lot harder than I thought it was.'
Jonathan Corbin
[ 00:49:28 ] By the way, everyone that we laid off, we'd like to invite you to come back and Salesforce, please let us renew our subscription.
Jonathan Corbin
[ 00:49:35 ] And the challenge is, is it's not just about building, it's about the maintenance of it. It's about the quality of your data. It's about maintaining that quality of data. And so, you know, when we went out there, we understood the challenges that enterprises are facing and we're able to mitigate those. We have the integrations. We have the data management platform. We have all the various components that are required in order to deploy AI agents at scale. To build and maintain those over the long term is not sustainable for most businesses.
Jonathan Corbin
[ 00:50:04 ] We're happy when people do that. That means they've educated themselves in the market. They understand the questions they should be asking when they go and purchase products. And so they're some of our paper buyers.
Heena Purohit
[ 00:50:15 ] I think it's awesome to hear the story about how they're coming back. You're thinking long term. These are long term relationships you're building with customers. Thank you, Jonathan. Any questions from the audience?
Heena Purohit
[ 00:50:28 ] I saw you raise your hand first, so I'll ask you and then you.
SPEAKER_0
[ 00:50:32 ] Thank you. First, introduce myself. I'm running an AI company that focuses on the biotech to try to accelerate some of the processes. So I deeply resonate when you talk about to build the mode where with the continuous feedback in the loop, because in the specific case, the data access is kind of limited. So we were thinking about using synthetic data.
SPEAKER_0
[ 00:51:00 ] decide to go the way along like using the human in the flywheel. I guess my question is more about how do you see the adoption cycle? Because we work in the AI biotech regulatory space, it's I think I see a lot of pinpoints there. It's a big issue. In today's process, we can solve it. But the adoption can be on the conservative side, because people in this space are naturally conservative. and have a lot of doubt about AI. So when you assess this type of opportunity, especially right now, the biotech is in a downturn. How would you say that would make me position myself better in fundraising? Thank you.
Chip Hazard
[ 00:51:59 ] I'm not super deep in biotech, but I think the general sense is you want to get people comfortable. It's just moving up the degree of complexity. And so what's a...
Chip Hazard
[ 00:52:13 ] somewhat lower risk use case that they can get comfort with. And then you take on a more demanding use case. And so you sort of work up the complexity ladder.
Chip Hazard
[ 00:52:23 ] You know, the challenge in biotech obviously is a wildly cyclical industry as it relates to funding and interest and and there's a bunch of things going on externally that are having an impact on that and so so as you think about qualifying customers you know you're gonna have to spend more time probably qualifying like how are they capitalized and what's their runway and what's their take capabilities the we see this with a lot of early stage companies like really narrowing your ICP not on on sort of obvious characteristics but sort of some with the non-obvious characteristics because you have limited resources you want to move very fast and so qualifying Customers who share your vision have the resources, can put some time and energy behind working with you, and that will take a little bit longer, but when you find those, those will build your momentum, get your product better, allow you to move up that complexity scale.
Heena Purohit
[ 00:53:13 ] And then, Habib, you can expand that, because I think it's a problem across other areas of deep tech.
Habib Hadad
[ 00:53:19 ] Yeah, I think, well, maybe biotech first and then I'll expand. So I think the one other thing is...
Habib Hadad
[ 00:53:25 ] Biotech is going to be very hard to find many customers. So I'd say we're not that deep, but we have a couple of investments. And it's just patience and keep knocking on as many doors. You just need one champion, one company to believe in you, to have a design partner.
Habib Hadad
[ 00:53:42 ] I don't have any advice except just knock on as many doors as you can. But the other advice would be back to Chip's point, and we're seeing a lot of Gen AI budgets in different industries where the board says, 'This Gen AI thing, we're going to reduce costs.' And it's not going to be part of the core processes in biotech. It might be something else that's around different workflows. So could you think about expanding your product line, whatever that is, into something that's not as core to drug development or drug discovery or things like that? And if so, then what that could look like, just maybe the things around it. And then maybe that becomes a bit easier for you to do that. And I think general advice is general AI budgets are a real thing. Maybe we're kind of getting to the tail end of them. But those budgets are being thrown. And you're finding a way to get some of that money early on. allows you to get into the door and potentially scale faster than typically going through long sales cycles with procurement and things like that.
Heena Purohit
[ 00:54:38 ] And they also belong in different groups. Some of them are in research groups. Some of them are in business groups directly, like those operational lines. So absolutely. I want to move to the next question. She raised her hand next.
SPEAKER_1
[ 00:54:50 ] Great. Um, it's nice to meet everyone. My name is Shreya. I'm the founder of Memoa. We're building the world's collection of collections at the intersection of AI, emotion, and e-commerce. My question is: Is the AI gold rush evenly distributed geographically, akin to the gold rush that actually happened in the 1800s? There's pockets like Silicon Valley in Boston where there is a lot of investment, valuations going up like crazy. But you also, even in the United States, have people who've never heard of AI or are so like afraid of what it's capable of. So my question is, both in an investment thesis process and in a founding process, how do you think about, you know, where the gold rush is actually happening in the demographics of the people, either customers, users, or other founders, when they're kind of navigating the gold rush?
Maia Heymann
[ 00:55:45 ] I think of two different answers, from the consumer and the individual, and then from the enterprise. I think, right, so I think in two buckets. And then you had a geographic layer on top of that. So sure, I guess there's also a third component of capital. Where is the capital flowing? So the enterprise is adopting at the very, you know, we think of this as sort of like. AI 1. 0 was machine learning, data science. Right now, at the very beginning, we had some question marks on how quickly the enterprise would adopt an O experiment. Completely wrong. I mean, it is mind-boggling the speed at which we're seeing adoption. And because companies themselves are so incentivized to not be left behind, leapfrogged, so the adoption curve is, and what we don't know is, you know, is everyone buying it once and experimenting and.
Maia Heymann
[ 00:56:50 ] you know, not getting the, we don't know. It depends. And then, so that's the enterprise. The consumer, that's just.
Maia Heymann
[ 00:57:01 ] that's just demographics right that so i and i don't we i can't comment on on consumer adoption you see it massively in areas and um and then i think capital the the It's mind boggling the amount of capital being invested in the Valley.
Maia Heymann
[ 00:57:24 ] And of course, you know, and in other and, you know, New York here.
Maia Heymann
[ 00:57:31 ] just because of our partner who's on the ground in the valley and the speed at which with zero, like very little, I'm exaggerating, but with...
Maia Heymann
[ 00:57:50 ] Very little, I'll call it work being done, just money flowing. And that has some longer-term negative effects in the industry. But you didn't ask that. That's a personal commentary.
Heena Purohit
[ 00:58:06 ] However, I'll also add that there are many organizations that are trying to invest in those what you consider underserved segments. So I can talk more offline about some of those. I'd love to get some more questions just for time.
Heena Purohit
[ 00:58:21 ] Maybe we could have one person answer each of them, since there's a lot of interest.
SPEAKER_9
[ 00:58:25 ] Yeah. One quick question is about the unit economics that they're really going to catch up with AI. Because until now, everything is funded by VC dollars, right? And it's very different financially than traditional.
Chip Hazard
[ 00:58:42 ] Yes, there actually are costs to goods sold. So I think we spend a lot of time thinking about gross margins because it's very hard to grow a negative gross margin business very, very quickly.
Chip Hazard
[ 00:58:54 ] So paying attention to gross margins is not that they need to be amazing out of the gates when MongoDB launched their cloud service called Atlas, which today is over a billion dollar business. The gross margins were like 10%. And I sat in the boardroom, I'm like, how are we gonna build a business on a 10% gross margin? And they're like, trust me, trust me, it's gonna come down. And so if you have a point of view on how your margin profile will improve over time, such that you can build an enduring business, that will be incredibly important, even if today, maybe it's not the best unit economics.
Chip Hazard
[ 00:59:24 ] And so just being ready for that question, being ready for the roadmap that you're answering. But if you look at the cost of tokens, they're continuing to drop just massively. And so I think the tailwinds are generally in your favor on that approach.
Heena Purohit
[ 00:59:38 ] I want to get that last one.
SPEAKER_6
[ 00:59:40 ] Hi, I'm Jessie. I'm a recovering founder and just AI super curious.
SPEAKER_6
[ 00:59:45 ] You made a comment about the speed being a super mode. for companies and i had a visceral reaction to this because um as jonathan mentioned you know with tripadvisor serving so many people like any acts accidents or issues that occur can have a real impact for people and can do some harm. How do you think about advising your portfolio companies or companies that you look at on moving very quickly?
SPEAKER_6
[ 01:00:16 ] and taking into consideration the customer experience, the ethics, the go-to-market strategy of all of the things that they do, knowing that we want to move super fast and carefully. We want to run fast but not break things.
Habib Hadad
[ 01:00:29 ] Yeah, that's a great question. And I think moving fast doesn't mean you have to move fast at the expense of consumers or at the expense of ethics or at the expense of uncertainty. There's a lot of ways to do that. There's simulation. So that's one thing you could do. Simulate your environment before and stress test that. Their sandbox, so when Maven AGI went quickly to deliver to customers, that wasn't from day one, switch on, and every users were having it. It was a rollout. There's also tools where you can have uncertainty estimation in real time in the AI models. So we have actually a company, a portfolio of spin-off MMIT. that basically detects uncertainty and hallucinations in real time. So there's lots of different tools to do that. I mean, at the end of the day, there's lots of times where we pass on founders because we think that there's that misalignment in in culture, in thinking, in what the world should look like, and uneven access to AI. It's also something we think a lot about. But I think, by and large, all of those are manageable.
Habib Hadad
[ 01:01:30 ] Speed doesn't mean breaking things. Speed means executing, building, and getting to that result, that answer faster than you initially would have.
SPEAKER_6
[ 01:01:37 ] Can we package that and tell that to people so there's less fear and uncertainty in the market?
Habib Hadad
[ 01:01:43 ] Yeah. And to be honest, maybe we're unique. I mean, on this panel, I know all the GPs are like, I'm sure we share our same views. But in general, it's not all VCs would think like that and would prioritize that.
Habib Hadad
[ 01:01:56 ] you know mileage may vary i think what i have in boston is unique and one thing we didn't talk about so i will need to conclude on that is the role of academics and in the world where where the research is happening outside of academia because of compute and money being spent academia is becoming much more important so keep an eye on what it means the the research being happening on the humans and ai and and really how to focus on human flourishing in the world of ai that's a whole new field that only academia can excel at all right we're at time for today but thank you so much
