This is a weekly feature that runs down the week鈥檚 top 10 announced funding rounds in the U.S. Check out last week鈥檚 biggest funding deal roundup here.

Startups raised funding rounds with a lot of zeroes at the end this week. Three companies 鈥� , and 鈥� secured financings of $1 billion or more. Additionally, a robust lineup of companies in sectors including AI, e-commerce, cybersecurity, biotech and even mining also announced sizable new rounds.

1. , $1.37B, manufacturing: Hadrian, a developer of highly automated factories, raised $1.37 billion in Series D funding led by , , , , and . The financing sets a $7.87 billion valuation for the 6-year-old, Torrance, California-based company.

2. (tied) , $1B, energy storage: Austin-based Base Power, a developer of residential battery energy storage systems, secured $1 billion in Series D financing at a $13 billion post-money valuation. , , and led the financing, which coincided with the launch of the company鈥檚 Base Core home battery.

2. (tied) , $1B, nuclear power: Valar Atomics, a developer of technology and infrastructure to deliver nuclear energy, closed on $1 billion in Series B funding led by . The El Segundo, California-based company also secured a $200 million credit facility led by and .

4. , $700M, AI connectivity: Lumilens, developer of a connectivity platform for AI infrastructure, emerged from stealth and announced more than $700 million in new funding. , , , and led the financing for the San Jose, California-based startup.

5. , $545M, live shopping: Live shopping marketplace Whatnot bagged $545 million in Series G funding. The round reportedly a $20 billion valuation for the Los Angeles-based company, with , and as lead investors.

6. , $310M, critical minerals: Mariana Minerals, a software-focused developer of projects for supplying critical minerals, picked up $310 million in Series B financing led by . The 4-year-old company engineers, builds and operates mines and refineries using its software platform.

7. , $300M, AI infrastructure: Volta, a developer of AI cloud infrastructure, from stealth and said it raised a Series A at a $2.4 billion valuation, led by , , and .

8. , $250M, cybersecurity: San Francisco-based cybersecurity provider Horizon3, announced a $250 million Series E. and led the round, which set a valuation of more than $2 billion, triple the value set for its Series D last year.

9. , $188M, biotech: Watertown, Massachusetts-based drug discovery startup LifeMine Therapeutics secured $188 million in Series E funding led by . The funding will go toward clinical development of its lead program and advance its pipeline of transplantation and immunology therapies.

10. , $150M, agentic AI: HappyRobot, developer of an agentic AI platform geared for enterprises in sectors including logistics, financial services, utilities and manufacturing, raised $150 million in Series C funding led by and .

Methodology

We tracked the largest announced rounds in the 兔子先生传媒 database that were raised by U.S.-based companies for the period of Aug. 1-7. Although most announced rounds are represented in the database, there could be a small time lag as some rounds are reported late in the week.

Illustration:

Among lead investors, topped the ranks last month, while was by far the busiest backer by deal count. The highest-spending investors for the period, meanwhile, appear to be and .

For more detail, below we ranked active investors for July by several metrics. These include most prolific venture dealmakers, most active lead backers, biggest spenders and highest-volume seed investors.

Active lead investors

We鈥檒l start with active lead investors for the month, which, as usual these days, featured a heavily AI-centric lineup of deals.

Khosla Ventures ranked as the most active lead investor in rounds of $5 million or more, with eight deals in July. The largest were a $300 million Series A for quantum computing startup and a $120 million Series C for AI-enabled legal tech provider .

took the No. 2 slot, with six lead deals, followed by , with five. Below, we charted the top eight lead investors for the month by deal count.

Busiest venture investors

The ranks looked quite different when we widened the category to include both lead and non-lead investments in rounds of $5 million or more.

By this metric, repeat frontrunner Y Combinator once again took first place, participating in at least 19 such rounds. The storied accelerator typically takes a non-lead stake in follow-on rounds for startups it incubated.

Insight Partners and Andreessen Horowitz were next on the list, with 10 deals each, followed by Khosla and , with nine each. For a bigger-picture view, below we ranked the top 18 busiest venture investors for July.

Highest spending investors

When we focus on investors who led the most expensive assortment of startup financings last month, the lineup shifts once again.

For July, Coatue ranked as the apparent highest-spending听1 lead investor, backing a $10 billion financing for 鈥� rocket company, . (It should be noted though, that Blue Origin, founded in 2000, is probably too old to be considered a startup, although it is still a private company.)

Nvidia also stepped up, backing a $5 billion financing for foundational AI startup . Index Ventures and Andreessen Horowitz ranked high as well, each leading or co-leading rounds collectively valued above $2 billion.

Below, we rank 18 of the highest-spending lead investors for the month.

Seed dealmakers

Seed dealmakers were a bit more challenging to rank for July, in part because there鈥檚 often a time delay before smaller deals enter the dataset. One thing that is apparent is that Y Combinator was the most prolific investor at this stage, while other 鈥渦sual suspects,鈥� like and , also ranked high.

Related reading:

• A Record 14 Billion-Dollar Rounds In July Pushed Venture’s Historic Run Higher

Illustration:

To dig into this trend, 兔子先生传媒 News analyzed a dataset of repeat founders who have gone through YC鈥檚 cohorts. That analysis revealed some very interesting insights. The dataset, shared with us directly from YC, consisted of 454 repeat founders through the program as well as 935 founder-company records spanning 2005 through 2026.

What we found is that repeat participation to date has mostly been a two-chapter journey: 428 founders (94%) went through YC exactly twice, while only 25 appeared three times. and co-founder was the sole four-time founder.

Other highlights: Founders typically returned to YC five years after their previous appearance, with an average gap of 5.1 years. However, the data reveals two distinct themes. Nearly 30% of return participations occurred within two years 鈥� including 38 in the same calendar year 鈥� while 61 returns happened after a decade or more. Some founders jump straight into their next venture, while others take years off to build experience before coming back around.

Repeat founder numbers peak in the most recent data, hitting 65 in 2025. But that doesn’t automatically mean people are returning at higher rates. In recent years, YC cohorts have grown significantly, and the 2025-26 numbers include newer batch formats alongside potentially incomplete data.

It’s also clear that returning to YC isn’t always a solo journey.

Several complete founding teams returned together for subsequent companies, including those behind , and , as well as and .

A trend YC partners are watching closely

, a general partner at the San Francisco-based accelerator who worked the spring 2026 batch, has enjoyed a front-row seat to the shift. In that cohort, he said he had 鈥渁 bunch of repeat, second-time founders鈥� he’d worked with before 鈥� several during their previous YC company.

鈥淚t definitely feels like more of a trend now,鈥� Epstein said. Still, he’s careful not to overstate the novelty.

鈥淚t’s not a new thing. But the alumni base of past YC founders continues to grow,鈥� he said in an interview with 兔子先生传媒 News, and that naturally translates into more people eligible to come back.

Epstein has worked with more than 1,000 startups at YC. Before that, he was a startup entrepreneur himself, co-founding (YC W10), a marketplace for graphic design assets that he sold to in 2014 before spinning it back out as an independent company in 2017.

Ask him what separates second-time founders from first-timers, and he points to experience using the program itself.

鈥淭hey know exactly how to get the most out of the advice, network and resources available to them,鈥� he said. 鈥淗aving been through the startup grind, they get really good at focusing on the signal that matters and cutting out the noise.鈥�

That experience also helps them avoid a specific, costly mistake.

鈥淭he biggest mistake I see second-time founders avoid is overhiring or overspending pre-product-market fit,鈥� Epstein said. 鈥淭he biggest regret of all the successful first-time founders I know is that they hired too many people, moved way slower and didn’t like working at their own companies anymore.鈥�

Leaner teams, powered by AI

That instinct toward leanness shows up in another pattern: Many repeat founders are choosing to start solo the second time around.

鈥淪ome of them (repeat participants) are solo founders, but they’re not building alone,鈥� Epstein said. 鈥淭hey already have networks of people they can bring in as founding employees. This helps them move faster, and feels more fun and less lonely.鈥�

He compares this shift to how cloud computing eliminated the need for startups to raise large sums just to pay for servers.

鈥淚t wouldn’t surprise me if 10-15 years from now you look back at all the money startups had to raise to hire people and realize that’s not a requirement,鈥� he said.

AI is accelerating that shift, and Epstein sees it pulling former company builders, including himself and YC CEO , back into hands-on product work.

鈥淚t’s so easy to get back into it and start building again. And it’s incredibly exciting,鈥� he said. That mix of hard-won product sense and new tooling, he believes, is changing what one person can build alone.

鈥淭hey actually become the people that can produce at 10x or 100x what a traditional engineer would be able to build,鈥� he said.

As an example, Epstein pointed to , a founder he first worked with on in 2020 who’s now building an AI tool that helps founders manage their projects and automate tasks.

Even so, Epstein believes founders keep coming back for the same core reasons: personalized advice from partners, a community of ambitious peers, access to top investors and alumni, and the urgency of the batch environment.

鈥淭he pressure cooker environment of the batch, which pushes them to move even faster, and distribution to thousands of companies within the network,鈥� he said. 鈥淚t’s extremely hard to replicate those things on your own.鈥�

From Opkit to Sazabi

One of the repeat founders Epstein has worked with is , whom YC has now backed twice.

Callaway’s path to Silicon Valley began almost by accident. As a college sophomore, he skipped a lined-up investment banking internship after reading about a software bootcamp in San Francisco 鈥� a decision he calls 鈥減robably the single most important鈥� of his life.

From then on, his goal was clear: 鈥淚 wanted to do my own venture-backed tech startup, and I wanted to do a YC venture-backed tech startup.鈥�

After gaining experience at and fintech , he founded his first company, , in YC’s fully-remote summer 2021 batch. Opkit was a healthcare-fintech startup building insurance verification and revenue-cycle-management software.

鈥淚t was, in retrospect, not the right thing for me to be working on, but a really fun and interesting and rewarding first venture,鈥� he said in an interview. Opkit was later acquired by .

That experience shaped his second company, , a name chosen deliberately in contrast to Opkit.

鈥淥pkit wasn’t very personal to me. It was more of an MBA case study approach to starting a business,鈥� he said. 鈥淲ith Sazabi, it needs to really be in alignment with who I am and my passions and interests.鈥�

Sazabi, an AI-native observability platform competing with incumbents like , draws directly on work Callaway has done throughout his career 鈥� a return, in his words, to 鈥渨hat I know best.鈥� He sees it as part of a common pattern: First-time founders often avoid building in the field they know best, then return to it with their second company.

Callaway hadn’t originally planned to go through YC again, and the reconnection happened almost by chance through an email that looped in his former partner on Opkit, Epstein. Once Callaway decided to return, he was more strategic about timing, even deferring his batch to build out more of the product first.

鈥淚 wanted to use YC as a go-to-market acceleration event,鈥� he said, something he likely wouldn’t have known to do without having gone through the program before.

The founder was back at YC in person for the first time this spring. He described the second-time experience as something entirely new: 鈥淚t was really something special.鈥�

This time around, Callaway also noticed a more experienced cohort than his own first batch, along with new concerns specific to the AI era. 鈥淭here’s a lot of anxiety around what the durable moat is in an AI world when lines of code are effectively free,鈥� he said.

On fundraising, he drew a pointed comparison to 2021. 鈥淪pring 2026 felt similar to fall 2021,鈥� he said, 鈥渂ut unlike 2021, where interest rates and ZIRP drove a lot of that energy, in 2026 it’s driven by AI and by real material gains.鈥�

The company鈥檚 thesis is resonating with investors. In late June, Sazabi announced an $8 million seed round led by , and Y Combinator, with participation from and more than 60 angels from companies including , and .

鈥淎I has changed how software gets written. Now it is changing how software gets operated,鈥� Callaway said. 鈥淪azabi is rebuilding observability from first principles for a world where agents are part of every engineering team.鈥�

Overall, as AI continues to lower technical barriers and YC’s alumni pool keeps growing, second-time founders like Callaway are becoming an increasingly visible part of the accelerator’s lineup.

Illustration:

In November 2022, was doing something rare for a longtime educator: taking time off. Having launched his career as a teacher in Atlanta Public Schools, Khan became an assistant principal before founding his own public high school in Denver. After a year spent coaching principals at the district office, he decided to take a 鈥減ersonal sabbatical.鈥�

Then, ChatGPT came out.

Khan began tinkering with the new technology, fascinated by its potential.

鈥淚 actually went out to my old school building, the one that I founded, and started using it with teachers,鈥� Khan recalls. He ran workshops and asked the educators to use the tool in as many scenarios as possible.

The responses were varied. Most teachers barely touched it. A few tried, but felt doing the work manually was faster. However, some had a lightbulb moment.

鈥淭here were one or two teachers who told me, 鈥楾his has completely revolutionized the way I teach,鈥� 鈥� Khan said in an interview with 兔子先生传媒 News.

Seeing that divide sparked something in him.

鈥淚 thought this technology could impact every teacher, not just teachers who are really enthusiastic about using new technologies,鈥� Khan said. 鈥淪o the task then was, 鈥楬ow can we take all the power of this new technology and make it really accessible to teachers?鈥� 鈥�

Building the 鈥榲ertical AI鈥� for K-12

That experiment set the groundwork for , a platform designed as an all-in-one AI operating system for K-12 educators and students. For teachers, the tool acts as a daily assistant. It performs tasks like building rubrics, differentiating assignments for varied learning levels, and generating practice worksheets and reading materials.

The platform also helps educators offer monitored AI experiences directly to students. Those experiences range from algebra tutors to writing assistants customized with state exam rubrics that deliver tailored feedback to help students revise their essays.

鈥淵ou can think of MagicSchool as the vertical AI solution for K-12 schools,鈥� Khan said. “Enterprises are adopting generative AI in other fields 鈥� Like in law, there鈥檚 and that are vertical AI for legal firms. We鈥檙e kind of that, but for K-12 schools.鈥�

Today, the company鈥檚 primary customers are school districts that want to provide a safe, governed environment for generative AI that aligns with data privacy rules and local curriculum priorities. MagicSchool now partners with large school systems, including Denver Public Schools, and Florida鈥檚 Broward County Schools and Hillsborough County Schools, as well as private institutions.

“One in five children in America go to a school that is in partnership with MagicSchool,” Khan noted. Additionally, roughly 8 million educators worldwide have signed up for the platform, he said.

The uphill battle to raise capital

Despite the platform鈥檚 rapid adoption, Khan’s path to raising capital for MagicSchool was a challenge. In the beginning, he worked with hourly contractors and lacked a formal business model.

鈥淚 had no real business plan,鈥� Khan said. “The most successful tech companies from my perspective as a consumer were the ones that just got a lot of users, and that was my goal 鈥� I was like 鈥榣et’s just get a lot of people using this, and we’ll figure it out from there.鈥� 鈥�

Once MagicSchool鈥檚 user base neared 1 million, Khan began pitching venture capitalists. However, when compared to standard Silicon Valley profiles, his background as an educator initially proved to be a hurdle rather than a selling point.

鈥淣obody wanted to give a former principal money,” Khan said, recalling 鈥渜uite literally hundreds of meetings鈥� before securing an institutional investor.

鈥淚 think that investors are taught to pattern match,鈥� he noted. 鈥淭hey’re saying, ‘Hey, well, did you go to ? Are you a tech person? Did you work at ? ‘… I have none of those things on my resume.鈥�

Even edtech-focused investors were hesitant, leaving Khan frustrated as he watched other founders secure millions based purely on tech-heavy resumes.

鈥淚 remember seeing other edtech companies right around our size raise seed rounds 鈥� with no product, no sales, no nothing,鈥澨� he recalled. 鈥� I would think, 鈥楶eople know what our product is. Millions of teachers know what our product is, and nobody’s heard of that one.鈥� 鈥�

To overcome the skepticism, Khan relied strictly on impressive growth metrics, convincing investors during every fundraising stage that the business鈥� 鈥渢raction was undeniable.鈥�

The strategy paid off. Following early angel investor checks, MagicSchool went on to raise a $2.4 million seed round led by Colorado-based . To date, the company has raised nearly $63 million in total funding, driven by strong financial growth, including 3x year-over-year revenue growth at the end of last year, according to Khan. The startup鈥檚 other backers include , , and.

Expertise as the next wave of innovation

Although Khan no longer manages the high school he founded, he stays connected to the classroom through district visits. The school remains a top-performing public school in Denver under a former founding team member, he noted.

鈥淥f course, I miss that,鈥� Khan admits. 鈥淭here’s nothing that can replace the relationship you build over a long period of time with students.鈥�

Yet, his deep-rooted experience in education ultimately became MagicSchool’s greatest asset 鈥� a trend Khan sees taking hold across the broader AI landscape as domain experts step up to build industry-specific tools.

鈥淚 think that what we’ve learned over time is that the model is no longer the differentiator,鈥� Khan says. 鈥淗ow you contextualize the model with the real problems that people have in the work that they’re doing, and specific expertise, is the thing that’s going to unlock the next wave of innovation and impact for generative AI.鈥�

Related 兔子先生传媒 query:

Illustration:

July ranked as the third-largest funding month of the year, up 10% over June, following on the heels of a record-breaking first half of 2026, when startups raised $515 billion globally.

Fourteen startups raised billion-dollar rounds in July, the highest count in a single month, though not the largest amount raised in such deals, an analysis of 兔子先生传媒 data shows. The tally includes nine U.S.-based companies, two each from Germany and China, and one company headquartered in Singapore.

The largest startup funding deal last month was a $10 billion investment in , the first external financing for the -founded space exploration company.

, a frontier lab founded by former Chief Scientist , reportedly raised $5 billion from . The next two largest deals were Beijing-based frontier lab 鈥檚 $3.5 billion raise after releasing its latest Kimi K3 model, and raising $2.8 billion for short-video generation.

Two Germany-based companies in defense tech also raised billion-dollar rounds: and . In the U.S., companies that raised billion-dollar-plus rounds spanned the energy, industrial robotics, AI training, security and semiconductor industries.

Funding to AI

A total of $35 billion, or around 53% of global venture funding, went to AI-focused companies in听 July. Other leading sectors were aerospace, defense and energy.

U.S.-based companies raised a total of $39 billion, or around 59% of global venture capital, last month with roughly half of the capital invested in its AI-focused companies.

Exits

July was also a robust month for startup exits, including via acquisition and public-market debuts.

Venture-backed M&A totaled more than $9 billion in July, with five companies exiting at prices听 over $1 billion, 兔子先生传媒 data shows. Notable acquisitions included London-based data center provider 鈥檚 roughly $1.65 billion acquisition of software layer , which was built to manage AI workflows, and in the security sector, AI-native security company 鈥檚 $1 billion acquisition of , a service to manage non-human identities.

Twelve venture-backed companies went public above $1 billion in value in July, including five from China, six U.S.-based companies, and one from Italy. The largest was Chinese chipmaker , which went public at around $85 billion and . Italy-based , an acquirer of software companies including and , went public at a value of $18.5 billion. And last-mile transportation company , founded in 2017, went public at $1.6 billion in value, raising $167 million in the process.

In closing

If the first half of 2026 established that venture has entered a new era of mega-financings, July reinforced that the trend is broadening rather than fading. Record numbers of billion-dollar rounds in both hardware and software, alongside a healthy IPO and M&A market, point to an ecosystem where capital is not only concentrating in category leaders but is also beginning to recycle through exits.

Related 兔子先生传媒 queries:

Methodology

The data contained in this report comes directly from 兔子先生传媒, and is based on reported data. Data is as of Aug. 3, 2026.

Note that data lags are most pronounced at the earliest stages of venture activity, with seed funding amounts increasing significantly after the end of a quarter/year.

Please note that all funding values are given in U.S. dollars unless otherwise noted. 兔子先生传媒 converts foreign currencies to U.S. dollars at the prevailing spot rate from the date funding rounds, acquisitions, IPOs and other financial events are reported. Even if those events were added to 兔子先生传媒 long after the event was announced, foreign currency transactions are converted at the historic spot price.

Glossary of funding terms

Seed and angel consists of seed, pre-seed and angel rounds. 兔子先生传媒 also includes venture rounds of unknown series, equity crowdfunding and convertible notes at $3 million (USD or as-converted USD equivalent) or less.

Early-stage consists of Series A and Series B rounds, as well as other round types. 兔子先生传媒 includes venture rounds of unknown series, corporate venture and other rounds above $3 million, and those less than or equal to $15 million.

Late-stage consists of Series C, Series D, Series E and later-lettered venture rounds following the 鈥淪eries [Letter]鈥� naming convention. Also included are venture rounds of unknown series, corporate venture and other rounds above $15 million. Corporate rounds are only included if a company has raised an equity funding at seed through a venture series funding round.

Technology growth is a private-equity round raised by a company that has previously raised a 鈥渧enture鈥� round. (So basically, any round from the previously defined stages.)

Illustration:

Menlo Ventures XVII will invest primarily in seed and Series A companies, while Menlo Inflection IV will provide growth capital to startups at Series B and beyond. The new funds will target companies throughout the AI market, from foundational models and infrastructure to enterprise, healthcare and consumer applications.

The new capital gives the Silicon Valley firm more flexibility to back companies from their earliest days through later funding rounds that can require hundreds of millions of dollars. It also shows how important AI has become to a firm previously known for investments in companies including , and .

In recent years, has become the most prominent company in Menlo鈥檚 AI portfolio. The firm first invested in the AI model developer in 2023 and has added to its investment in later rounds. Menlo鈥檚 other AI investments include app-building platform , music-generation startup , AI model marketplace , voice productivity company , AI infrastructure companies and , robotics startup , and AI research company .

, a partner at Menlo since 2015, has played a central role in developing that strategy. He invests across AI infrastructure, developer tools and AI-native software and has led Menlo鈥檚 investments in companies including Anthropic, Lovable, OpenRouter, AI-powered software delivery platform , code security startup and legaltech startup .听

Before joining Menlo, Murphy spent 15 years as a general partner at Kleiner Perkins, where he was an observer at Google from the firm鈥檚 initial investment through its IPO, helped launch the $200 million iFund with Apple and worked on investments including DocuSign, AppDynamics, Upstart and Shazam. Earlier in his career, he held operating roles at Netboost and Sun Microsystems.

兔子先生传媒 News spoke with Murphy about why AI is pushing Menlo toward larger and more concentrated investments, what the firm has learned from its relationship with Anthropic, and where he sees the next opportunities 鈥� as well as potential bottlenecks 鈥� across the AI market.

The interview has been edited for brevity and clarity.

兔子先生传媒 News: Inflection IV puts Menlo in competition with some of the biggest late-stage investors in the world. How do you keep the firm鈥檚 close, founder-focused approach when you鈥檙e writing much larger checks?

Murphy: AI companies need more capital than previous generations of software companies. They鈥檙e staying private for longer, and the winners are quicker to break from the pack.
For us, a larger fund gives us the ability to partner with founders from company formation through hypergrowth. Through our venture fund, we invest in seed and Series A companies, but the inflection fund gives us the scale and flexibility to back the clear winners as they emerge.听

This was our strategy with Anthropic, Suno, Wispr, OpenRouter and Lovable.听

You鈥檝e recently invested $100 million in companies including Lovable and Suno. Is that level of concentration becoming a bigger part of Menlo鈥檚 strategy, or is it reserved for a small number of standout AI companies?

The Anthropic investment is an example of us doubling down when we had incredible conviction. Remember, we first invested in the [Series] C round, which gave us a chance to get close to the team, see how well they were executing, and understand where they were going.

When we led the [Series] D round, it was still the largest investment the firm had ever made. We learned from that experience and success, and it’s become a standard part of our approach now. Also and importantly, the market has changed.听

There’s a gold rush around later-stage AI, and the companies that break out are growing at rates we鈥檝e never seen before, at scale. These companies need capital to sustain that growth and, frankly, have earned higher private valuations given the growth rate.听

We’re changing how we invest, but overall we鈥檙e pursuing more of a barbell right now. On the later end, we’re much more aggressive, stage- and capital-wise, for the right companies.听

That said, the bar is still very high. Many AI categories are overfunded, and there is a huge amount of speculation. The winners of this era separate quickly, and we believe they will compound at unprecedented rates.听

Your relationship with and Anthropic gave Menlo an early view into where the AI market was heading. What are you seeing now that you think other investors may still be missing?

I don鈥檛 know that it’s counterintuitive, but I’d say we are moving from Phase 1 to Phase 2 of the market and are seeing an entirely different set of opportunities and challenges.听

In Phase 1, developers just picked a model to start building AI. In Phase 2, we are seeing companies get to scale using AI and looking to optimize their spend and infra choices. A whole host of companies are seeing tailwinds alongside Claude and Claude Code, such as OpenRouter, Fireworks, Modal and .听

It will be a multi-model world. One size won鈥檛 fit all, and we鈥檝e been active in that area as well, including more vertical models such as for life sciences and for robotics.听

The Anthology Fund has helped you spot promising AI companies early. As the application layer matures, what specific bottlenecks are you seeing founders run into when building enterprise-grade defensibility on top of frontier models?

The Anthology Fund has been an incredible source of deal flow and has given us a broad aperture around what areas of AI are disproportionately taking off. It’s been a great program for getting closer to a broad set of application and infrastructure companies and building relationships before deciding where to lean in.听

I wouldn鈥檛 say it’s been the key factor in identifying bottlenecks across the AI ecosystem. For sure it is part of it, but from the broad set of portfolio companies and new companies we meet, the No. 1 bottleneck has been how to take all the new code that has been written and get it into production faster, safely, and securely.听

This has created a big tailwind for companies helping with software delivery, like Harness with application and code security, like Semgrep; and code review and testing like .听

Additionally, the rise of custom models based on open-source/open-weight models has created a number of bottlenecks as companies look for compute, training, sandboxes, and more. Both development and runtime resources have become essential to accommodate this next wave, and companies like Modal and Fireworks are addressing that with their offerings and the compute capacity they鈥檝e been able to aggregate across various compute providers, including Nebius and CoreWeave.

Valuations across the AI market have risen dramatically. Which parts of the market do you think are most likely to produce strong, sustainable businesses: infrastructure,听model tools or industry-specific applications?

We鈥檝e been active across models, infrastructure, and applications. All are showing tremendous potential and tailwinds right now. At the moment, infrastructure is seeing a disproportionate spike in opportunities as enterprises and AI-native companies embrace a multi-model approach and scramble to keep up with the compute and infrastructure management needs that it requires. Coding tools are now mainstream and putting tremendous pressure on organizational processes to release software faster and more efficiently, which is leading to tailwinds for companies like Harness and Gimlet.

It’s fair to say the majority of companies are optimizing for market share right now rather than gross margin, but there are many opportunities for margin improvement over time, and this is a rare land-grab moment.

You鈥檝e backed new AI research labs before they even have a product, including . At that stage, what convinces you that a team has something truly different, and that it can compete with much larger technology companies?

As I mentioned, we believe in a multi-model world, where one size won鈥檛 fit all needs and use cases. We have an explicit strategy to gain early exposure to some of the most compelling AI research teams with distinctive techniques or capabilities, even at a very early stage. Many of these companies are raising $100 million-plus rounds, and while we occasionally lead in this category, we prefer to write smaller checks initially. This helps us build broader exposure across the category and talent pool, and then double down once we see one really taking off.

Frankly, there are too many right now, and all claim some differentiated technique or team. Of the roughly 60 model companies, we believe we鈥檝e invested in more than five of the best and expect to lean into one or two of them as they ramp.

Related 兔子先生传媒 queries:

Related reading:

• Anthropic Backer Menlo Ventures Raises $3B In New Funds To Back AI Startups Across Stages听

Illustration:

This is a weekly feature that runs down the week鈥檚 top 10 announced funding rounds in the U.S. Check out last week鈥檚 biggest funding deal roundup here.

Another week, another bevy of big rounds. For this past week, the largest round was a reported $5 billion -backed financing for foundational AI unicorn , followed by a $1 billion investment in . Other sizable rounds went to companies in sectors including energy storage, health testing, nuclear power, fintech, and, yes, AI.

1.听, $5B, foundational AI: AI lab Safe Superintelligence announced a long-term with Nvidia to rapidly accelerate its growth. The deal included a $5 billion investment from the chip giant with an eye toward boosting compute resources for Safe Superintelligence, a Silicon Valley startup founded by co-founder .听

2.听, $1 billion, fusion energy: Commonwealth Fusion Systems, a startup working on fusion energy technology and developing a grid-scale fusion power plant, raised $1 billion in fresh funding from unspecified investors. The round brings the total invested to date for the Massachusetts-based company to $4 billion.

3.听, $550M, thermal energy storage: Antora Energy, a company that provides energy through thermal batteries to data centers, announced that it has closed on $550 million in a Series C funding round. and co-led the financing for the nine-year-old, San Jose-based company.

4.听, $450M, health testing: Austin-based Function, a provider of lab testing, imaging, and personal health information that markets to consumers, secured $450 million in growth financing from .听

5.听, $370M, nuclear energy: Antares, a nuclear fission energy company that develops compact microreactors for defense and space applications, raised $370 million in Series C equity funding. and led the financing for the three-year-old company, which raised $100 million in debt funding alongside the equity investment.听

6.听, $200M, AI simulation: Simile, a developer of AI tools for running simulations, said it picked up over $200 million in fresh funding at a $2 billion post-money valuation. led the round, which comes just five months after the Palo Alto-based company launched its product.

7.听, $190M, cybersecurity: Cybersecurity provider ThreatLocker closed on $190 million in Series F funding to hone its platform and expand internationally. led the round for the Orlando-based company.

8.听, $170M, fintech: New York-based CAIS, an alternative investment platform for independent financial advisors, secured $170 million in Series D financing. led the round, which set a valuation for the company of more than $2 billion.

9.听, $160M, fintech: PEX, an AI-enabled provider of prepaid and charge cards for businesses, along with tools to track finances, raised $160 in equity and debt funding, with as lead investor.

10.听, $145M, AI infrastructure: Eliyan, a developer of connectivity technology for AI infrastructure, completed its Series C with a total of $145 million at a $1 billion valuation. led the financing for the five-year-old Santa Clara, California-based company.

Illustration:

That was the intent of our latest 兔子先生传媒 News data dive into seed-stage trends. For this installment, we focused on mid-sized rounds of between $5 million and $10 million, analyzing around 800 global seed financings that closed this year.听

Why this range? In a startup investment climate characterized by the ascendance of megarounds, the idea was to focus on rounds more representative of the classic seed deal: a risky bet on unproven founders, technologies or business models.

Using this methodology we identified multiple popular investment themes and zeroed in on five. The first 鈥� cybersecurity 鈥� we tackled in a separate piece. Here we delve into the other four: proptech, cancer therapeutics, space tech and robotics.

No. 1: Proptech

Real estate is the world鈥檚 most valuable asset class, providing startups a huge and varied addressable market. By one estimate a few years ago, real estate accounted for a staggering two-thirds of global net worth.

Given the size of the space, actual venture investment tied to real estate and construction looks comparatively meager. Last year, per 兔子先生传媒 analysis, proptech startup investment totaled just over $10 billion, far below peaks hit several years ago.

Seed investors seem to believe there鈥檚 a good case for startup driven growth ahead. In particular, they鈥檙e funding a lot of rounds in the $5 million to $10 million range for companies looking to add efficiencies to the planning and building process, streamline rental operations, reduce building power consumption, and more.

To illustrate, below we put together a sample set of 15 companies that closed seed rounds in our target range this year:

A few standouts include , an AI-powered home management system, , a developer of software to support real estate decarbonization, and , an AI-enabled construction supply chain platform.听

No. 2: Cancer treatments

Startup founders don鈥檛 need persuasive superpowers to convince investors that cancer is a sufficiently serious area to address. Today, it鈥檚 that 39% of Americans will be diagnosed with cancer at some point in their lives. Cancer also ranks as the second leading , behind heart disease.听

Seed-stage companies aren鈥檛 expected to bring down numbers in the near term, but as they progress, it鈥檚 increasingly plausible. That鈥檚 the apparent mindset for investors at this stage, who鈥檝e backed a good-sized number of rounds in the $5 million to $10 million range this year for developers of cancer therapeutics and diagnostics, charted below:

Three California startups secured $10 million, the largest financing in our sample set. They include: , which is working on AI-driven discovery of undetected cancer targets, , a developer of targeted therapies for solid tumors, and , which is focused on cancer diagnostics.

No. 3: Space and satellite tech

This year鈥檚 most attention-getting event in space tech finance was obviously the IPO of sector pioneer . But while that debut may have dominated headlines, quite a few smaller, earlier, lower-profile deals were also getting done.

Per 兔子先生传媒 data, space tech was a popular area for seed financings in the $5 million to $10 million range. To illustrate, below we put together a sample set of nine such companies that raised rounds this year:

The largest fundraiser in our target range was , which is focused on developing reusable satellites. Next was , focused, as its name implies, on in-space propulsion systems, followed by , developer of an ML-native operations platform for satellite fleets.

No. 4: Robotics

Robotics is a perennial favorite in our seed-funding data dives, including the last one, focused on AI. This time, the sector made the ranking again, thanks to a bevy of intriguing seed-stage companies that met our parameters.

Turns out, you can jumpstart some highly ambitious ventures on a $5 million to $10 million seed round. To illustrate, below we aggregated a sample of 18 funded this year:

Robotics was also the most geographically dispersed sector in our lineup, with startups hailing from Asia, North America, Europe and Australia. A few that stood out include , a developer of what it calls 鈥渋ntimacy robots,鈥� , a maker of autonomous underwater robots, and , focused on robots for greenhouse harvesting.

Big picture: Midsized seed rounds for outsized ambitions

Overall, seed funding trends reviewed above may tell us more about the kinds of companies investors are willing to bet on than about the sectors attracting interest, which are already well-established.

Clearly, startup investors still believe that small, modestly funded teams with grand missions remain a worthwhile and viable wager. That鈥檚 particularly encouraging these days, when the venture and seed financings we most commonly hear about tend to be the largest ones.

That鈥檚 not to diss large rounds. Startups that are led by prominent serial entrepreneurs or have established traction hold obvious appeal, even at pricier terms. But for those of us who enjoy rooting for the underdog, it鈥檚 encouraging to see lower-profile companies with outsized ambitions are still in the game.

Related 兔子先生传媒 lists:听

Related reading:

• AI Seed Investors Flock To Cybersecurity听
• In Charts: Seed Deals Keep Getting Bigger As Odds Of Reaching Series A Fall Dramatically听
• AI Seed Trends: More Multimedia, Backend Automation, Agentic Security, And Yes, Robots

Illustration:

Before that, Larco had spent nearly eight years as a partner at (NEA), one of the world’s largest venture capital firms.听

There, she served on the firm’s investment committee and led investments across enterprise software, developer tools, and consumer technology, including , , , , and . She also served as a board observer at leading up to its 2021 IPO.听

Known for her sharp product intuition and hands-on operational experience, Larco focuses heavily on helping founders evaluate market dynamics, navigate product-market fit and scale resilient teams.

Before transitioning to venture capital, she built a career as a product leader and founder. After earning a degree in computer science with honors from the , she began her career at working on and , before leading core product teams at companies like and . She also founded an app development startup that she successfully ran and sold before joining NEA.

兔子先生传媒 News recently sat down with Larco to discuss how changing founder preferences and the (SVB) collapse drove her to launch a specialized pre-seed and seed fund designed to make early-stage founders a top priority.听

Among other topics, we also discussed how she evaluates startups based on founder potential rather than initial ideas, looking for teams that leverage AI to make products dramatically faster, cheaper, or easier to use while avoiding rigid, single-model wrappers.

This interview has been edited for clarity and brevity.

兔子先生传媒 News: You were at New Enterprise Associates for nearly a decade before branching out on your own. What led you to start your own firm? Was there a specific gap in the market, or was there a premise you felt couldn’t necessarily be fulfilled at a fund that size?听

Larco: There were a lot of things. At a multi-billion-dollar fund, writing $2 million checks is never going to be a top priority. They invest across all stages, but when you have to deploy between $3 billion and $6 billion depending on how you look at it, it鈥檚 impossible to do that $2 million at a time with standard team sizes.

Even if you still write those checks, founders have gotten wiser to what it feels like when they are a top priority versus when they aren’t. One founder put it to me this way: 鈥淚 want my investor at every round to feel like the check size hurt 鈥� that it’s a big percentage of their fund 鈥� because that鈥檚 how I know I鈥檓 going to be a top priority when push comes to shove.鈥�

So, for a pre-seed round, they want a pre-seed fund where the check size hurts. For a seed round, they want a seed fund where the check size hurts. For a Series A, they want a mid-sized fund where the check size hurts.

That frank conversation put a lot into perspective. Founder preferences have shifted over the past few years. Emerging funds over the last three to four years are winning very competitive deals, securing lead slots against more established, bigger firms. This was virtually unheard of before.

How has that happened?

A side, unintended consequence of the SVB collapse was this change in founder preference. When SVB was going under, every single founder called everyone on their cap table saying, 鈥淚 can’t make payroll on Wednesday. Can you help me?鈥�

Every VC was getting dozens to hundreds of calls. Depending on portfolio size, you can’t help everybody or be on the phone with every single company. Everyone had to prioritize. If firms scraped together money to help cover payroll, they couldn’t cover everyone across the entire portfolio. Very quickly, founders got to see where they sat on the priority list.

That’s interesting. As I cover rounds lately, I鈥檝e noticed the lead investors aren’t as often the big mega-funds.

Not at pre-seed or seed.

Even Series A. You’re seeing less of it happening.

Part of it is that fund sizes got really big, so they are writing bigger checks, which inevitably leads to more calculated ROI risk and moving to later stages. Part of it is that founders want to be a top priority, and they saw what happens in a crisis.

Founders are on WhatsApp channels, hacker houses, and communities, so one bad story spreads faster than ever. It used to be just repeat founders who wanted specialized, focused firms at the earliest stage for signaling risk and other reasons. Now, even first-time founders hear those stories and want a specialized investor.

When customer preferences change in any market, you realize there’s an opportunity. We asked ourselves: 鈥淐an we capitalize on this shift? If you were to build something from the ground up targeting this specific ICP, what would you build?鈥�

We did what we tell our founders to do: a listening tour. We interviewed people in our ICP and asked: What do you wish you had? What works, what doesn’t, what taglines are you skeptical of, and what is tangibly helpful? We doubled down on what we could provide well and cut out things people assume are best practices that founders don’t actually value.

We think of Premise as a startup, and our product happens to be a fund, so it still has to be something people want.

Do you invest strictly at those very early stages, or across other stages?

Strictly pre-seed and seed. Check sizes range from $500,000 to $3 million.

It鈥檚 noisy out there. How are you able to cut through that noise to identify real potential versus people riding the AI bandwagon? As a journalist, I struggle with that, so I imagine investors do, too.

We spend a lot of time with founders before backing them. During diligence, we talk one to three times a day for three to five days, alongside extensive reference and back-channel checks. Because of that, most of our investments are in cities where we have strong networks, like SF, New York, and Atlanta.

We try to get a deep sense of who the person is, what motivates them, and what key attributes they possess. Mercedes and I looked across all the best founders we saw at our previous firms and identified seven core attributes. There isn’t one single persona; founders have different strengths and weaknesses. We look for founders who are world-class in at least two of those seven attributes. In our investment memos, we justify those choices with anecdotes and reference feedback. Nobody is the best at all seven 鈥� some attributes even contradict each other.

At the pre-seed and seed stages, whatever idea you pitch 鈥� while we want it to be a good idea because it shows your ability to plan and generate ideas 鈥� the likelihood that it’s what the company looks like in five to ten years is very slim. A lot of it is gauging the potential of the person to find the right market and product fit to build an iconic company.

It is tough, but it’s not that different from the crypto, Web3, or early AI waves. Tailwinds always attract fair-weather founders. The core tactics to figure out who really wants to build something interesting, who has unique insight, and who is tenacious enough to endure the ups and downs haven’t changed in the last decade.

I’ve seen you discuss AI as a concierge service, shifting from “do-it-yourself” tools to “do-it-for-me” agents. You’ve also mentioned that an AI agent shouldn’t just be a wrapper; it needs to significantly re-architect the cost structure. When looking at a seed-stage deck today, what stands out as evidence that a team actually knows how to fundamentally change that cost structure?

Those can actually be two separate things. If a traditional wedding planning concierge service costs $20,000, and you offer it for $1,000, you’ve blown the cost structure out of the water 鈥� even if you’re just a wrapper using $100 in API credits. You can be a wrapper, pay for APIs, and still charge a fraction of traditional costs because the legacy price anchor is so high.

What I look for in any company to be competitive is whether it is faster, cheaper, or easier than existing options. A 10% discount isn’t enough, but at 50% off, people will switch. If a tool reduces a weekly five-hour administrative task to five minutes, sign me up. The bar now is enabling people to do things they couldn’t do before or lacked the confidence to do. For instance, I can build a cap table in Excel, but it takes me forever. If a tool makes that effortless, I’m in.

So ideally, a startup delivers on at least two of those three pillars: faster, cheaper, or easier.

I’m not against wrappers, but founders must understand the underlying mechanics. If you scale and the wrapper gets too expensive, or the model degrades, you need to know how to split tasks across open-source, closed, Google, or other models to deliver the best product at the best price.

Technical founders obsessively optimize models for specific features across their product. Less technical founders often use a single model for everything, which doesn’t guarantee the best price or performance. My hesitation with wrappers isn’t that a team launched quickly; it’s when they don’t know how to continue innovating because they’re wedded to a single model.

The counter-argument to my own point is (AWS). When AWS came out, critics said, 鈥淎nyone can start a company over a weekend on AWS; it’s not defensible, there’s no moat, you don’t own servers.鈥澨�

Yet many great companies were built on it. It鈥檚 the same argument. People said the same things about the cloud and mobile waves 鈥� that mobile was a toy and no one would buy a $1,000 phone or pay for subscriptions. Looking back, those criticisms sound funny.

You mentioned you look for seven distinct founder attributes, and that a founder needs to be world-class in at least two or three. Without giving away the whole secret sauce, what is one attribute on that list that would surprise people?

The one that catches people off guard is what we call 鈥渦rgently dissatisfied.鈥� These founders can come across as disagreeable: they’re more focused on the goal than on making people feel good, and their standards can be genuinely difficult to work around.听

But the people who’ve worked with them tend to say the same thing: that the founder pushed me to accomplish things I didn’t think were possible. This shouldn鈥檛 be confused with ego. It’s about managing hustler, relentless energy and pointing it at the right problems. The best founders I’ve backed have this quality. They have a high bar for themselves and their teams 鈥� as in everything should have been done yesterday, and they should have acted accordingly.

On the flip side, given how fast the tech landscape is shifting right now, is there an attribute that used to be a ‘must-have’ for a Series A founder five years ago that you now consider a nice-to-have at the seed stage?

The attributes themselves are pretty universal truths about what makes a great founder. What’s changed is the intensity and pace at which they have to show up. Five years ago, shipping an exceptional product, not just features, every six to twelve months was the bar. Now it’s every three to four months.听

So being a decisive execution machine still matters enormously, but what we’re evaluating is whether a founder can operate at this new compressed pace without sacrificing quality. That’s a harder thing to assess early, but it’s become one of the most important signals.

Right now, a huge portion of the VC ecosystem has completely retreated from consumer tech to chase B2B enterprise AI. Are you still actively looking at consumer behavior change as an investor? Do you think the rest of the market is miscalculating the size of the consumer AI market, and if so, why?

I think the retreat is short-sighted. Consumer software has historically produced some of the most important companies ever built, and it doesn’t make sense to vacate that entirely because the sector has been in a lull the past few years.听

The first principles of what makes a disruptive consumer company are exciting again because consumer behavior is rapidly changing with AI. We price in that risk. Fintech is another space where I’ve seen a meaningful pullback, and we’re still active there for the same reason. If everyone is running from a category, that’s usually worth paying attention to in case new tailwinds emerge.

You spent years as a product leader at places like and . We鈥檙e hearing a lot of talk about how AI will automate the tedious parts of product management 鈥� writing tickets, reviewing specs, tracking bugs. If AI absorbs the execution workload of a PM, what does a top product leader actually do day-to-day in 2026?

The job of a PM has always been consumer empathy: understanding what someone is trying to accomplish and why, and then making sure the product actually gets them there.听

AI only changes the artifacts you produce. A few years ago, you were writing specs. Now the best PMs I talk to are writing evals to define what 鈥済reat鈥� looks like for the agents they’re building and testing whether the agents actually deliver it.听

Someone somewhere still has to care deeply about the end user, ask the hard questions about what success means, and hold the bar. That’s still a human job.听

I love the analogy that AI wrappers are just the new AWS. But with AWS, the 鈥渕oat鈥� eventually became workflow stickiness and data accumulation. In a world where technical founders are constantly swapping models to optimize cost and performance, what does a 鈥渕oat鈥� actually look like for an early-stage company? If it’s not the underlying model, then what is it?听

I think it鈥檚 still workflows and data accumulation. I don鈥檛 think the moats changed much. The real question is how you retain your customers when competitors can clone you in three days. There are small non-durable moats you can lean on before you build out the data/workflows/network effects/integrations/etc moats.听

You made an interesting distinction between how technical and non-technical founders approach model selection. Given that, are you leaning heavily toward funding purely technical, AI-native architectures right now, or can a world-class product-and-distribution founder still win you over if they hire the right engineering talent?听

Never say never, but I am heavily biased towards a founder or founding team that has exceptional AI talent. I find that these folks enjoy being at the cutting edge, staying up to speed on the latest breakthroughs, and don鈥檛 mind blowing up their roadmap to move fast on a new functionality that enables them to build better products for their customers.听

You talked about AI shifting from ‘Do It Yourself’ to ‘Do It For Me,’ like giving everyone a concierge wedding planner or a financial analyst. When an agent moves from just giving advice to actually executing transactions and making decisions on behalf of a user, what is the biggest hurdle you see startups face? Is it a trust problem with the user, or is it an execution infrastructure problem?

Few people want 鈥淒o it entirely for me, and I have no idea what you did or how you did it鈥� right now. Most concierge services do the research, ask you questions to personalize the recommendations, and then filter down the options they present. If you have questions, you can dig into their reasoning, what they ruled out, etc. If you don鈥檛 like the options, they can go and find a new set. Rarely do wedding planners, travel agents, etc just go off and book everything for you without your input. I think that鈥檚 where we are with agents. It鈥檚 not just a trust problem, but more that people still want to make the decisions themselves 鈥� just not do all the research.听

Related 兔子先生传媒 query:

Illustration:

Think of the most famous boxers you know, likely the heavyweights: Muhammad Ali, Joe Louis, Mike Tyson. In a clash between titans, the advantages seem easy to understand, since the bigger fighter looks like the stronger one.

But size alone is not a strategy. Sugar Ray Robinson was a middleweight, not a heavyweight, and in the 1950s, A.J. Liebling said he looked 鈥渕ore like a loose-limbed dancer than a boxer.鈥�

Robinson鈥檚 advantage was completeness: speed, footwork, intelligence and stamina. In a famous 1951 match against Jake LaMotta, Robinson schooled the reigning, heavier middleweight champion with a 13th-round TKO.

Completeness also applies to companies. The market tends to assume that big companies will capture the biggest gains from AI. But AI is tough to get right at any size.

Look at , valued at $6 billion in 2024. It made headlines claiming its -powered chatbot could handle millions of conversations and do the work of 700 customer service employees. Customers hated the rollout, and by 2025, Klarna was . Or , which watched its value after ChatGPT commoditized its main offerings.

If everyone can get AI wrong, who wins?

Enter the scrappy middleweight

Each year we speak with thousands of operators and founders, and one pattern is clear: The biggest long-term gains from AI will not flow to heavyweight incumbents or many AI-native startups but to scrappy middle-market technology companies, the middleweights.

The next phase of AI disruption will be challenging, but middleweights can gain serious ground.

One objection: Won鈥檛 hyperscaler companies go after certain verticals? If Copilot inside 365 or 1听agents can run a workflow, how does a middleweight company survive? The answer depends on what constitutes durable advantage. Horizontal platforms are built for generalized work, not the messy, regulation-heavy, category-specific workflows of the real world. Middleweights can win by making their software the system of record that AI calls into instead of software that AI replaces.

The odds for making big, impactful gains with AI right now favor the middle market, where proven growth companies can use customer trust, domain expertise, capital structure and speed to transform their businesses, taking market share from slower incumbents. With three-quarters of AI鈥檚 economic gains now being captured by just per , entrepreneurs who stand still may already be losing the round.

What makes for a winning middleweight company?

The best middleweight technology companies share the five traits below, all working together as a system.

Disciplined self-assessment. Middleweights are designed to act quickly on honest feedback, and their boards help them test where AI generates value versus where it merely consumes engineering capacity and budget.

Seat-based pricing is one area for brutal assessment. When autonomous agents do the work, the revenue model should reflect outcomes, not users. In 2023, customer service platform made a bold switch, pricing its AI agent Fin at 99 cents per resolved conversation. That agent became the company鈥檚 core offering, and it recently . Outcome-based pricing might seem painful at first (and reorganize your GTM team and their incentives), but it anticipates an agentic future.

Agility. Enterprise companies are weighed down by technical debt and legacy infrastructure. Middleweights have enough scale and proprietary data but not so much organizational mass that every experiment needs 10 layers of approval. Their agility is as much cultural as structural.

These are ambitious, scaling companies growing 20% or more with strong unit economics, and a tech-first mindset runs through the entire business, not just the engineering org. Take the restaurant software , where early AI gains came from product leads ; those product teams then built a flywheel connecting new product features to external communications, with LLMs continuously editing and improving instructions for AI agents.

Workflow ownership. In the AI era, the strongest moat is owning a complex workflow. Middleweights have spent years gaining this position 鈥� integrating into customer systems, accumulating exception-level data, learning operational nuances that take a claims process from 95% accurate to 99.5%. (The last 4.5 points are the moat.)

An company, , doesn鈥檛 just apply AI to contracts; its moat is absorbing the decision workflow around each contract. As a contract moves through approvals, negotiations and redlines, the important part is learning from the history of why internal teams decided the way they did. Well-positioned companies will hold the institutional memory that AI agents need to query to do their jobs.

Technical capacity. Most large companies are stuck in AI pilot purgatory, and the market still underestimates how operationally demanding AI deployment is. Middleweights have something most AI-native startups lack: years of working with real customers. , another FTV company, started in 2007 as a service-heavy cybersecurity business that has learned deep detection logic from operating in more than 1,000 customer environments, including some of the largest global enterprises. As the company saw rapid automation from machine learning, then more sophisticated AI, it moved in-house SOC analysts into higher-value product development roles, allowing engineers with deep cyber expertise to drive key R&D.

A well-capitalized balance sheet. Companies with cleaner balance sheets can move faster, absorb experimentation costs, pursue selective M&A, and keep investing through periods of disruption. Large legacy software companies carrying heavy leverage, optimized for cost-cutting and growing at 5%-10%, can鈥檛 be light on their feet and will struggle to reallocate capital aggressively enough into AI R&D.

The imperative

Plenty of boxers can be complete for one season. Sugar Ray Robinson executed consistently in 200 professional fights, mastering the sweet science with a reliable system. That same high standard now applies to companies in the AI era.

The window for transformational gains with AI is not open indefinitely. Speed is a middleweight leader鈥檚 advantage. Do not wait to perfect your AI strategy; start executing.

If you don鈥檛 know where to start, pick key workflows, map them and ask whether AI makes them more defensible or more exposed. The answer may determine whether you give up a round or win the match.

is managing partner at , where he oversees the firm鈥檚 global strategy and investment decisions. He has been a growth equity investor at FTV for more than 20 years, leading investments in enterprise technology and services and financial technology and services. Bernstein has over 25 years of private equity experience. Prior to FTV, he was a partner at and its predecessors where he managed the business and financial services group. He began his private equity career with and started his professional career in the investment banking division of in New York.

Illustration: