What AI Hype Cycles Actually Look Like From the Inside With Mercedes Bent

• Mercedes Bent grew up surrounded by tech entrepreneurs, cut her teeth at Lightspeed through four distinct market cycles in six years, and came out the other side with a sharp eye for what's real and what's noise.

• She sees AI creating massive value but warns that labor displacement, unit economics, and an inevitable security reckoning are the unresolved tensions most people are ignoring.

• Now building a new venture firm and running technical communities like Surreal and SciFi Tech Club, Mercedes draws her biggest inspiration from science fiction—and the futures it dares her to build toward.

Let's dive in. No floaties needed…

Revenge of the Nerds

Mercedes Bent, Investor & Co-Founder

She didn't discover tech—she grew up in it. Her parents worked at companies like IBM and Apple, then started their own ventures in video streaming and RFID in the '90s and 2000s. She thought everyone's family built things. Banking and consulting were the exotic career paths she discovered in college.

After years in finance and operations, Mercedes joined Lightspeed Venture Partners in 2019, where she invested from pre-seed to Series B, co-led the firm's LATAM region and Scout Fund, and backed companies like Stori, Honeylove, Magic Eden, and Outschool. Now she's co-founding a new venture firm focused on AI-native tools built by technical founders—and running communities that bring engineers, researchers, and builders together around sci-fi and emerging tech.

Looking Back at Your Path From Finance to Operating to Venture, What Experiences Most Shaped How You Think About Technology & People Today?

The earliest influences were just the formation stages of life. I was very lucky to grow up in a household where my parents were big into tech. They worked at large companies like IBM, RCA, and Apple, and they were also entrepreneurs—they started a video streaming company and an RFID company in the '90s and 2000s.

I was always around tech. Frankly, I thought everybody did it when I was younger. I didn't know people had normal jobs like lawyers and doctors. That was new to me. I learned about banking and consulting when I went to college.

During Your Time at Lightspeed, What Patterns Did You Start to Notice About How Technology Cycles Actually Play Out?

I joined Lightspeed in 2019, and what followed was just whiplash after whiplash. In March 2020, the pandemic hit, and deal flow completely dried up. Nobody knew if they were going to have a business anymore. I remember sitting there in May, genuinely worried that the industry had just died. One of my partners told me to take it as a breather—that it would come back in a few months. I wasn't sure he was right, but he was.

Then 2021 roared back. Interest rates had dropped, crypto was reaching new heights, and the market went completely crazy. Founders would show up with no pitch deck, no metrics, and four competing term sheets due by tomorrow. Almost no diligence was being done. It was a gambling casino. To put it in perspective, in 2019, a Series A company with $2M ARR was often raising $6M or $7M at a $35M post. A year later, people were asking for $20M on $150M valuations with no deck.

Then whiplash again in 2022. Interest rates started rising, and by April, nobody wanted to invest. That lasted through 2023 until AI picked things back up. By late 2023 and into 2024, things were getting frothy again. So in six years, I lived through at least four distinct cycles. And if you zoom out to the broader context—the dot-com bubble, the financial crisis—I think 2021 was the mobile bubble peak, and now we're in an AI bubble. I just can't tell you yet how far into it we are.

When You Look at Consumer AI Products Today, What Tends to Catch Your Attention Early & What Quietly Makes You Skeptical?

What captures my attention is anytime someone can show amazing retention—especially cohort retention that's increasing over time, cohort over cohort. There's a great chart that was published showing OpenAI's consumer subscription cohorts, and it's the most beautiful chart. You see an initial drop-off of people churning, but then it flatlines and even starts to tick up into a smile at the end.

With cohort charts, what you're always looking for is the shape and the trend. And with that ChatGPT paid subscription chart, you could clearly see month over month that more and more people were retaining—or said another way, fewer and fewer people were churning. That's what gets me excited: finding a product where people care enough that they just don't leave.

As AI Starts to Take on Roles Traditionally Filled by Humans, What Feels Genuinely Exciting & What Still Feels Unresolved?

One of the biggest unresolved elements is labor displacement and how we're going to account for this massive dislocation. I believe AI is going to make some people work more and more, but it's increasingly going to be the haves and the have-nots—fewer people taking more of the profits. Right now, AI can replace pretty much any white-collar, knowledge-worker, entry-level job. So the question is: what happens to all those new grads coming out who need to get trained for the first time?

Hopefully, they're adopting AI skills and learning to manage it. But companies are also freezing their workforces at current sizes. Over time, more people won't be able to fill those roles. I think we're going to see a lot more joblessness, including in creative fields where AI is starting to do more. That feels like one of the biggest unresolved issues.

The other big unresolved issue is unit economics. Right now, a lot of companies are taking on debt to fund the capital expenditures going into data centers and energy infrastructure to support computing. But is the revenue actually big enough to make up for all of that investment? If you analogize this to the railroads in the 1800s, there was massive investment, and not all of those investors made their money back. I think this cycle is going to have similar issues with people recouping their costs.

In the Current AI Landscape, Where Do You Still See Startups Having Real Leverage Over Large Incumbents?

The classic advantage hasn't changed—startups can move faster, experiment, build, and ship quicker. Most big companies just have too many ongoing priorities to compete. I think Google is one of the only exceptions, because AI was so existential to their core product—search—that they had to respond quickly. Everyone else? Apple—what are they doing? There's no competing hardware computing platform yet.

Incumbents will have an advantage at improving the profits of their existing cost structure by leveraging AI, but they're not going to be good at generating new revenue from AI-native products. That's just not their forte. If customers are looking for a 10x better experience using the latest technology, they're going to new companies for that.

At This Stage of Your Career, What's Been Capturing Your Curiosity & Attention the Most?

Right now, AI has really sucked the oxygen out of the capital environment. Anything that isn't an AI company is having a hard time raising money. The game changed very quickly. A few years ago, the playbook was the 'triple triple triple double double'—you 3x revenue three years in a row, then double twice, and you could get to over $100M ARR in five years. That's now no longer considered fast enough, which is difficult for a lot of startups to digest.

When you see GenSpark go from zero to $100M in revenue in nine months, or Higgsfield go from zero to $200M in a little over a year, it's really hard to justify investing in anything else. But the question remains—the very first thing I mentioned—great retention. And those companies aren't publishing their retention numbers widely. That's going to be what helps everyone figure out whether this level of prioritization toward AI is actually justified.

I also think there's a reckoning coming. There's going to be a huge security incident. AI-generated code isn't always the best at creating secure permissions. There have already been instances where public API keys were released in documentation. I think we'll see a big incident that causes real backlash—after people get their identities and money stolen due to poorly coded AI solutions.

What Books, Essays, or Podcasts Have Been Shaping How You Think Lately?

I run a sci-fi tech club on the side where we discuss the technology and science behind our favorite science fiction books and movies. Most of my reading lately has been related to that. We just read Kim Lue's All That We See or Seem, which features a group dreaming technology but also has AI woven throughout—everyone has a personal pocket AI, and everyone's data is easily trained on.

I love reading books like that because it helps me imagine a future state of the world. I started the group during maternity leave about a year ago. I was bored, wasn't working, and I figured sci-fi isn't technically work—so I could read books and call it a hobby. I've always felt it's my favorite pastime. Anytime I'm on vacation, I'm reading sci-fi. And one day it just flipped for me: why don't I spend more time doing the stuff I love and build a community of other people who love it too? That's where I'm drawing most of my inspiration today.

Quick Poll

Additional Reads

• Platform Shifts, AI, and the Future of Consumer Investing — Venture Unlocked

• Global Investment Opportunities in Fintech and B2B Marketplaces — Fintech Leaders

• Closing the Gender Investing Gap — CNBC Television

Rate This Edition