鈥淚 learned that you can course-correct if you focus on the early life of the baby. If they’re missing certain beneficial bacteria, it can really cause a lot of issues for the infant,鈥� Sew Hoy recalls. 鈥淚 was reading all this literature and asked my doctors about it, and none of my doctors knew what I was talking about.鈥�

A week after her son鈥檚 birth in 2020, she incorporated Austin-based . Today, the company announced it has raised $33 million in a Series B funding round led by , it told 兔子先生传媒 News exclusively.

The financing brings Tiny Health鈥檚 total funding to $46 million, following a $4.5 million seed round in 2021 and an $8.5 million Series A in late 2023. Existing investors , and participated in the latest round, along with new investors including and . B Capital鈥檚 will join the company鈥檚 board.

Demand for Tiny Health鈥檚 at-home microbiome tests is growing. The startup鈥檚 overall revenue has more than doubled each year, according to Sew Hoy. Its B2B arm, Powered by Tiny, which sells tests through healthcare providers and other companies, grew fourfold last year. Sew Hoy did not disclose revenue figures or the company鈥檚 valuation, though she said the valuation was 鈥渕uch higher鈥� than in the Series A.

From baby gut tests to adult testing

Tiny Health spent two years developing its platform before launching an at-home baby gut test in 2022. Healthcare was new territory for Sew Hoy, whose previous startup, digital coupon company , was sold to in 2013. Before raising a seed round, she put $50,000 of her own money into an initial study and spent about a year collecting stool samples and building the platform.

The baby test gave Tiny Health a focused entry point. In Sew Hoy鈥檚 view, parents of children with conditions such as eczema wanted more information as to the cause, while clinicians had few practical tools to apply emerging microbiome research. The company also offered pregnancy, vaginal microbiome and adult tests, but initially focused on infant testing. Adults have since become its larger testing segment, according to Sew Hoy.

So, how does it work? Customers collect a small stool sample with a swab. After sending in the swab, they receive a report on the microbes detected, along with explanations and suggested next steps. Tiny Health also sells a subscription called Tiny Plus that includes a baseline test and a retest.

Sew Hoy believes one of Tiny Health鈥檚 biggest differentiators is that its tests use shotgun metagenomic sequencing to analyze microbial DNA. The method provides a broader view of the organisms in a sample than tests that look for a limited set of microbes, she said. Chief science officer , who joined Tiny Health in 2021, leads its scientific work. The company鈥檚 other leaders include co-founder , who joined Sew Hoy two years after she started Tiny Health and leads its technology work, and chief medical officer , an integrative pediatrician who previously advised the company.

The reports examine what the detected microbes may be doing and suggest steps such as dietary changes, with links to supporting research.

鈥淲e are a wellness test,鈥� Sew Hoy said. 鈥淲e鈥檙e not a diagnostic 鈥� It鈥檚 very actionable, very evidence-backed.鈥�

Insurance does not cover the tests, which can identify microbes in a sample, but cannot establish what caused a person鈥檚 symptoms.

Other startups are taking different approaches to digestive health. Austin-based makes a device that attaches to a toilet and tracks stool patterns and hydration. Tiny Health analyzes microbial DNA from a collected sample. Throne recently raised $10 million in a Series A round led by .

Beyond direct-to-consumer testing

Tiny Health鈥檚 fastest-growing business is Powered by Tiny, which sells testing through healthcare providers and companies that incorporate it into their own services. Partners include , and 鈥檚 Executive Health and Longevity Programs, according to Tiny Health. It also works with supplement makers and academic partners that want to measure the microbiome in clinical studies.

Customers began taking their direct-to-consumer results to doctors, according to Sew Hoy. Some clinicians then approached Tiny Health for help interpreting the reports. Today, the startup serves more than 6,000 practitioners and offers training on using its tests in patient care.

Its enterprise business supplies testing to other health and wellness companies, including through technology integrations. As mentioned earlier, that B2B arm is growing fourfold. Expanding both, along with clinician training, is a priority for the new funding.

, senior principal at B Capital, told 兔子先生传媒 News that his firm sees an opportunity to make microbiome research more useful to consumers as interest in preventive health grows.

鈥淲e’re in the middle of a real shift in how people engage with their own health. Consumers are paying out of pocket for wearables, blood panels, and full-body scans because they want to understand what’s happening inside their bodies before something goes wrong,鈥� he wrote via email. 鈥淭he gut microbiome is one of the most important and least understood pieces of that picture 鈥� Tiny Health turns that science into something people can actually act on, and we believe the company that does this well will define the category.鈥�

Whitehead also cited Tiny Health鈥檚 standing with clinicians the firm consulted during its diligence.

鈥淭iny Health was consistently the test they rated highest,鈥� he said.

Another attractive feature, Whitehead added, is the company鈥檚 data from families tested beginning in infancy.

Putting its data to work

Tiny Health has collected nearly 200,000 microbiome samples over time, according to Sew Hoy. To make use of those samples, the startup is developing a tool called TinyAI that will draw on its data and research reviewed by its scientific staff to help people understand their results. The company also uses health information customers provide, with identifying details removed, to inform that work.

Long-term, Tiny Health aims to identify patterns that could help spot health risks before a condition develops. has associated gut microbiome imbalances with more than 100 diseases.

鈥淲e want to be able to predict disease before it happens and intervene before it happens,鈥� Sew Hoy said in an interview with 兔子先生传媒 News.

Indeed, one pressing question is whether microbiome-guided interventions can improve health outcomes. Tiny Health says it has published four scientific papers, including a study involving infants, and plans to support further research involving adults.

Presently, the company has about 40 employees and plans to hire 30 more. It will use the new capital to expand its work with healthcare providers and other companies, and to fund research into whether changes in the microbiome can improve health outcomes. It plans to devote $5 million to a program that will give researchers and clinical partners access to its testing and data.

Related 兔子先生传媒 query:

Related reading:

• Pooptech Startups Are Passing The Smell Test
• Investments In Poop Startups Are Piling Up

Illustration:

She later became a general partner at , where she led early institutional investments in AI chipmaker , and later invested at . Now she鈥檚 founder and managing partner of , a $52 million fund backing startups that use AI to do work in industries such as construction, industrials and insurance.

In an interview with 兔子先生传媒 News, Venkatachalam discusses why she looks beyond familiar founder profiles, what makes an AI company durable, and how an early investment in Groq shaped her approach.

This interview has been edited for clarity and brevity.

You invested at Khosla Ventures before starting Axiom. What did you take from that experience, and what did you want to do differently?

Venkatachalam: One thing I took was 鈥檚 open view of where great founders can come from. Silicon Valley has gravitated toward a fairly narrow idea of who can build the next great AI company: Someone with a computer science or machine learning background, or experience at . We鈥檙e looking for more nonobvious founders, particularly in nonobvious industries.

The other thing was how we think about risk. Instead of asking a company every conceivable diligence question, we focus on the risks that matter for its next set of milestones. Can this team do what it says it will do? And if it can, could the result be massive?

That means accepting that many bets won鈥檛 work out while aiming for the outliers. I think that鈥檚 what my investors are backing me to do: identify categories of the future, rather than participate in the categories everyone already recognizes.

At Axiom, the biggest difference is that we鈥檝e built the firm around people who are actively working with AI. If you aren鈥檛 building, productizing, pricing or taking AI to market regularly, it鈥檚 very hard to keep up. Our team includes people doing exactly that in their other roles. They help keep our investment thinking current, and founders want to work with them because they鈥檝e encountered many of the same challenges.

We also use AI throughout the firm. We鈥檝e built what we call the Axiom Brain to help us make sense of market trends, identify interesting people and companies, and move faster on diligence and other work. For me, the value is the ability to act quickly.

You mentioned that some of those AI practitioners have other jobs. How does their role at Axiom work?

Venkatachalam: They have dedicated time to Axiom and work with us part time. They also receive carry in the fund. They鈥檙e partners in the work, rather than people whose names appear on an adviser list.

Their other jobs are central to the model. Some of the best angel investors are people who are still operating and building. I don鈥檛 need these people full time. In fact, they would be less valuable to Axiom if they left the work that keeps them close to the market.

Axiom says it invests in 鈥淎I for the real world.鈥� What does that mean when you鈥檙e evaluating a startup?

Venkatachalam: Our view is that AI should benefit a much broader population than the early adopters who are already using it. We look at industries underserved by technology, where AI can produce an outcome rather than simply provide another software tool.

That may lead us to construction, industrials or insurance. Some of those companies involve hardware, sensors or robotics; others are entirely software-based. What connects them is that they鈥檙e doing work that matters to customers in real-world industries.

We generally don鈥檛 invest in products that look like conventional enterprise software tools. We want to see AI delivering a result.

You鈥檝e described a shift from software people use to digital workers that perform jobs. Are customers actually paying for AI from labor budgets?

Venkatachalam: Yes, and that鈥檚 one of our investment criteria. Even when a portfolio company is at an alpha or design-partner stage, we do diligence to understand whether customers are willing to buy it that way. We鈥檙e often seeing contract values in the hundreds of thousands of dollars, rather than the much smaller contracts you might expect for a midmarket software tool.

We鈥檝e seen that buying behavior play out across the majority of the portfolio companies we鈥檝e invested in.

AI products are becoming faster to build and easier to imitate. What makes one durable enough to become a large company?

Venkatachalam: If you鈥檙e doing important work inside a customer鈥檚 business, and that work is worth a lot of money, you become difficult to replace. You鈥檙e handling what we call the last mile of the job.

In industrial settings, for example, delivering an outcome means integrating deeply with the systems customers use. You have to understand their data, train on it, learn the workflows that matter, and stand behind the result. That takes more than putting an interface on top of a model.

Those relationships and capabilities can become difficult for another startup to replicate. They also involve work that the large AI model companies may have little interest in doing themselves.

Before Axiom, you backed Groq when AI inference was far from an obvious investment category. What led you to it?

Venkatachalam: I came from both hardware and software. At one point, I became interested in why was building its own networking switches when it could buy them from existing suppliers. As I looked into that, I learned it was building its own chips, too.

That led me to , who had been involved in that work and had left to start Groq. I began learning why large technology companies were building chips to train models. Then Jonathan made the case that the much larger future market would be inference.

I鈥檒l be honest: In 2016, I barely understood inference. But if you believed these models would spread, it made sense that people would build on top of them and need the infrastructure to support that. That insight drove my investment.

How did that experience shape what you look for now?

Venkatachalam: It taught me the value of being a little early and having some patience. You don鈥檛 have to be wildly contrarian, but you do have to see the opportunity before it becomes obvious to everyone else.

In a way, our thesis hasn鈥檛 changed. We鈥檙e still asking what will be built on top of AI infrastructure and models. We want to invest while the answer is emerging, before there鈥檚 a consensus.

What happens when one of those early bets doesn鈥檛 work out?

Venkatachalam: We plan for it. We have a $52 million fund and expect to make 35 investments. We fully expect about half of them to fail, whether that means a company shuts down or simply never reaches the growth trajectory we鈥檙e looking for.

The model depends on finding an exceptional outcome. We need one outstanding investment to return the fund. If you invest early enough and the company becomes very large, that can offset many bets that didn鈥檛 work.

That willingness to accept losses is part of making investments before an opportunity is obvious. It鈥檚 built into how we approach the fund.

Related 兔子先生传媒 query:

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.

This week delivered a bountiful supply of big startup funding rounds, led by two $400 million financings for cybersecurity unicorns and . Rounding out the week were large financings for startups across hot sectors, including foundational AI, drug discovery, neurotech, and even rainmaking.

1. , $400M, cybersecurity: Island, a developer of tools for secure enterprise digital operations, raised $400 million in Series F financing. led the round, which set a $6.4 billion valuation for the Dallas-based company, more than double its 2024 valuation.

2. , $400M, cybersecurity: Cyera, a provider of enterprise data security tools that govern both humans and AI agents, picked up $400 million in Series G extension funding led by and backed by . The round brings total funding for New York-based Cyera to $2.7 billion, .

3. , $350M, foundational AI: San Francisco-based Snorkel AI, provider of tools for frontier labs and AI teams to develop specialized training data and environments, secured $350 million in Series E funding. and led the round, valuing the 7-year-old company at $3.2 billion, and it now has more than $375 million in annual recurring revenue.

4. , $311M, biotech: Enveda, an AI-enabled drug discovery startup, closed on $311 million in Series E financing led by and joined by a long list of new and existing investors. The round brings the Boulder, Colorado-based company’s total capital raised since inception to more than $845 million.

5. , $250M, neurotech: New York-based Precision Neuroscience, a startup focused on brain-computer interface technology, secured $250 million in Series D funding with and the as lead investors.

6. , $200M, wireless communications: Hubble Network, operator of a satellite network that extends connectivity to Bluetooth devices, pulled in $200 million in Series C funding. led the financing, which set a $1.6 billion valuation for the Seattle-based company.

7. , $155M, pharmacy benefits: New York-based Rightway, a provider of pharmacy benefits and care navigation services, raised $155 million in Series E funding led by . The startup will use the financing in part to expand its AI capabilities.

8. (tied) , $100M, cloud seeding: El Segundo, California-based Rainmaker, a startup focused on atmospheric research and producing new freshwater through cloud seeding, picked up $100 million in Series B funding. The round includes investments from , , , and .

8. (tied) , $100M, tax compliance: San Francisco-based Numeral, an AI-powered sales tax compliance platform, closed on a $100 million Series C round led by .

8. (tied) , $100M, AI data: San Francisco-based Micro1, a provider of AI training data to major AI labs, has reportedly raised a fresh funding round of more than $100 million at a valuation of $4 billion, according to a citing people familiar with the deal.

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 Sept. 19-25, 2026. Although most announced rounds are in the database, there may be a small time lag, as some rounds are reported late in the week.

Illustration:

So Shariff offered to build an AI agent for a couple of properties. The first, called Alfred, answered guest questions and helped with check-ins and check-outs. Within 90 days, he said, online reviews improved at both properties. He then began deploying it at more hotels, before deciding that guest messaging was only one part of the opportunity.

That work led him to team up with to start , the San Francisco startup they founded in July 2025. The company is now emerging from stealth with $6.7 million in seed funding to build agents that handle reservations, guest requests, staff coordination and other hotel tasks. led the raise, with participation from . The round marks Dextr鈥檚 first institutional financing.

From guest requests to bookings

After early guest management deployments, Dextr turned to incoming calls with a voice reservations agent called Daisy. If a caller has a question about a reservation or something they can’t find on a booking site, answering it could lead to a booking, Shariff found. As hotels began using Daisy, he realized that the phone calls were 鈥渓eading to revenue.鈥�

In fact, today, one large hotel that Dextr works with handles $100,000 to $300,000 a month in bookings through the voice agent, according to Shariff.

Dextr has since expanded into agents for group bookings, staff management and other work. Its goal is to connect those agents so information gathered in one part of a hotel can prompt action in another. For example, if a guest declines housekeeping, a guest management agent could pass that information to an operations agent so that staff schedules could be adjusted accordingly.

Another use case could involve a property operating without an overnight front desk employee. A guest arriving late could call the hotel, have Daisy verify their reservation and receive instructions for entering their room. Daisy and Alfred, Dextr鈥檚 guest management agent, would work together to handle the arrival.

For some properties, using agents to handle late check-ins could remove the need to staff an overnight front desk shift. Shariff estimated that such a shift could cost a property from $70,000 to $80,000 a year in a market such as California.

鈥淚t was very hard to hire for these shifts to begin with,鈥� he said.

Building ROI

Shariff says combining agents is just one of three things that set Dextr apart from other hotel AI tools. An agent that answers calls or messages can handle a particular task, he said, but connecting it with other agents opens up additional uses for a property.

鈥淚t’s important that we do multiple use cases to see the higher ROI,鈥� he told 兔子先生传媒 News in an interview.

The second distinction, according to Shariff, is Dextr鈥檚 approach to getting the agents running. Putting them to work takes more than installing software, he said. Every hotel differs from another in how it manages reservations, assigns staff, and handles guest requests. So Dextr sends engineers to work directly with customers, identify where agents could be useful and then connect them to existing systems. Those engineers stay involved as the property puts the agents to use, he said, and work to find applications that have the potential to deliver a return on investment.

鈥淓very property, every operation is different in this space,鈥� he said. Dextr integrates with property management systems including Oracle Hospitality, OPERA Cloud, , and .

The third is how employees use the agents. They can direct them by text or voice through a companion called Doss, without having to learn a series of steps in a new software system. Shariff said frequent turnover makes it difficult for hotels to train staff on complicated tools and retain that knowledge when employees leave. With Doss, 鈥測ou can call it, you can talk to it, you can text it,鈥� he said.

Hotels and more

Dextr says its agents now handle more than 1 million interactions a month across hundreds of hotels. The startup鈥檚 customers range from a 21-room property to one with 550 rooms. They include independent hotels and properties operating under the , , and brands, as well as reservation centers, vacation rentals and outdoor hospitality businesses.

The startup works directly with the owners and operators of those branded properties, Shariff said, and is beginning its first direct relationship with a hotel brand that he declined to name. It also works with hotel and property management companies, as well as reservation centers, vacation rentals and outdoor hospitality venues such as campgrounds, RV parks, lodges and golf courses.

The company charges a subscription based on the property and the agents it uses, with usage-based fees for some agents. Shariff declined to disclose revenue, profitability or the round鈥檚 valuation. After spending much of last year on pilot projects, he said, Dextr began growing more quickly this spring. He said annual recurring revenue has been growing about 20% to 30% month over month, though he did not disclose hard figures.

Dextr plans to use its new funding to hire more engineers who work directly with customers, expand its software integrations, and develop additional agents. Its team has grown from three people a year ago to 21.

, an AI partner at Elevation Capital, told 兔子先生传媒 News that Dextr鈥檚 early traction got the firm鈥檚 attention. He pointed to customers using the agents to generate bookings and upsells alongside reducing operating costs.

鈥淲hat stood out immediately was Sajid and Scott’s execution. Dextr scaled to [hundreds of] contracted properties and meaningful ARR without raising a single dollar,鈥� he wrote via email. 鈥淚n a space where most AI companies are still searching for product-market fit, Dextr had customers, zero churn, and a clear ROI story. That combination of founder quality and early traction is exactly what we look for at Elevation.鈥�

Related 兔子先生传媒 query:

Illustration:

So far this year, companies in the space have raised around $2 billion in seed- through growth-stage funding. That’s already ahead of the 2025 full-year total, driven in large part by big rounds for companies at the intersection of AI and gaming.

Lead fundraisers

A few particularly large rounds contributed an outsized share to this year鈥檚 funding tally.

Sunnyvale, California-based , a developer of foundation models for AI-powered 3D generation, was the largest fundraiser, pulling in $400 million in a July Series B at a $1.5 billion valuation. While not a pure-play gaming company, Meshy highlights gaming as a core use case for its 3D AI agent.

, developer of a platform for training AI models, was another investor favorite, pulling in $300 million. Like Meshy, Israel-based Decart is not an AI pure-play; it鈥檚 also known for a video simulation technology pitched as compelling for game development.

We also saw plenty of pure gaming startups in the largest-round list, including motion-based family game developer , which closed on $150 million last week, and Turkish mobile gaming company , which picked up $70 million in May. Below, we highlight nine standout funding recipients this year.

Investors are also scaling up

Gaming-focused venture investors are also attracting fresh capital. One of the larger raises closed last month, with San Francisco-based picking up $250 million for its fourth flagship fund. The closing follows a string of successful investments, including a lead seed stake in viral puzzle game maker , which later exited at a $5 billion valuation.

Makers Fund has continued to invest actively, taking part in at least 10 known rounds this year, per 兔子先生传媒 data. This includes financings for high-profile startups such as , the sports betting platform that recently drew attention for a controversial ad campaign.

A few months earlier, announced a fresh raise of its own, securing $100 million for a fund focused on indie games. The firm鈥檚 Special Opportunities Fund intends to provide financing in exchange for a share of a game鈥檚 revenue, in what it hopes will be a more appealing investment structure for indie studios.

The beginning of an up cycle?

Overall, while gaming startup investment remains far below the peaks seen a few years ago, the substantial year-over-year gains we鈥檙e seeing look encouraging. The direction of funding charts certainly points to an upturn still early in the making.

Related 兔子先生传媒 query:

Related reading:

• Gaming Startup Funding Levels Down Further In 2025

Illustration:

So far this year, global startups have secured at least 114 Series A rounds聽1 of $100 million or more, per 兔子先生传媒 data. That鈥檚 the highest annual total in years and on track to top the all-time peak.

Moreover, many of those jumbo early-stage rounds far exceed the $100 million threshold. Collectively, this group of Series A recipients has raised around $33 billion this year, with at least 12 rounds valued at $500 million or more.

An AI thing

The funding bump was mostly an AI-driven phenomenon. Per 兔子先生传媒 data, more than 70% of Series A rounds of $100 million or more went to AI-focused startups.

That figure encompasses some of the year鈥檚 largest early-stage financings. For instance, it includes a $1.2 billion round for Silicon Valley-based , a platform for developers to train and serve custom models, and a $900 million financing for China-based , a developer of AI-enabled humanoid robots.

Below, we put together a sample of 10 of the largest Series A rounds, including mostly AI but a few other areas as well.

The high preponderance of AI deals reflects what we鈥檝e been seeing across stages. In the first half of this year, venture and growth funding to artificial intelligence startups totaled an estimated $394 billion, roughly 77% of all investment capital. Granted, most of that was for later-stage financings. But our Series A data shows early-stage doesn鈥檛 look too different for AI鈥檚 share.

US leads for jumbo Series A deals.

Roughly half of this year鈥檚 $100 million-plus Series A rounds and funding went to U.S.-based startups, per 兔子先生传媒 data. That translates to about 62 deals with a collective value of around $15 billion so far in 2026, which puts it on track for a record tally.

Still, megaround funding at Series A is more globally dispersed than overall venture investment this year. In the first half of 2026, more than three-quarters of global seed- through growth-stage financing went to American companies, largely due to megarounds for Silicon Valley-based and .

When investors like the same things

One can point to several potential causes behind the rise in Series A megarounds beyond AI growth alone. For one, leading startup investors have exceptionally large capital reserves to deploy. Additionally, exit multiples historically, and to an even greater extent recently, reward those who are anything but modest in their ambitions.

At Series A, another factor may be that investors seem to agree more than usual on the sectors, business models and founding teams they want to back. And given that a pricey share of a winner still beats a discounted share of a laggard, they鈥檙e piling in to perceived early-stage leaders.

Related 兔子先生传媒 queries:

Related reading:

• 兔子先生传媒 Data: Global Startup Investment Hit Record $510B in H1 2026 As AI Boom Accelerates Funding And Exits

Illustration:

led the San Francisco-based company鈥檚 Series A, with participation from , , and of . The financing brings Baselayer鈥檚 total funding to about $40 million since its 2023 inception, according to co-founder and CEO . The company declined to disclose its valuation.

Baselayer combines business identity, credit and fraud data to help banks, fintech companies and other financial-services providers evaluate prospective customers. It sells its products directly and through software companies that resell or put their own branding on Baselayer鈥檚 technology. Its automated platform initially focused on Know Your Business, or KYB, identity verification, fraud detection and risk management.

More than 2,000 financial institutions 鈥� representing over 20% of such institutions in the U.S. 鈥� use its technology to onboard, underwrite and open accounts for merchants, according to Awad. Baselayer also works with Fortune 500 companies and has about 50 employees across offices in San Francisco and New York. Since its founding, the startup claims it has helped customers prevent more than $1 billion in fraud losses.

Awad declined to reveal hard revenue figures, saying only that Baselayer reached eight figures in revenue in less than two years.

Now, Baselayer is using its new capital to address a newer 鈥� and growing 鈥� identity problem: determining whether an AI agent is actually authorized to act on behalf of a particular person or business.

With people using AI agents left and right these days to book a restaurant reservation, for example, it鈥檚 becoming increasingly challenging to determine whether an AI agent鈥檚 automated activity is legitimate or if it鈥檚 a bot attempting to scrape data or commit fraud.

Alongside its raise, Baselayer today is also announcing the launch of its Agentic Identity Suite, extending its identity network from businesses to the AI agents transacting on their behalf.

From businesses to the agents acting for them

Awad and co-founder started Baselayer in February 2023, initially focusing on the lengthy and fragmented process financial institutions use to verify businesses and assess risk.

鈥淲hat we set out to do was essentially bring risk assessment to the 21st century,鈥� Awad recalls.

Awad describes Baselayer as both an identity network and a fraud consortium. Because its technology is used across thousands of financial institutions, Baselayer says it can recognize when the same person or business applies at multiple institutions and incorporate that activity into its risk scoring.

The company processes tens of millions of applications and says it sees many of the same businesses multiple times a year. That data becomes more useful as additional institutions and reseller partners join its network, according to Awad.

鈥淲e鈥檝e essentially streamlined 10 years鈥� worth of selling into two years,鈥� he said.

An AI agent presents a different problem, however. It may be created for a single task and disappear immediately afterward, leaving little or no history for a bank or risk provider to evaluate.

鈥淎gents spin up and they spin down,鈥� Awad said. 鈥淗ow can you trust this random one-task agent?鈥�

To address this dilemma, Baselayer is developing what it describes as 鈥淜now Your Agent,鈥� or KYA. The system is being designed to do things such as determine not only who deployed an agent, but also who that agent represents and whether it actually has permission to carry out a particular task.

It wants to do this by providing an authorized agent with a credential it can present when attempting to make a purchase or interact with another business. Then, when presented with a credential, a merchant, financial institution or online platform could use that information to decide whether to allow the transaction to proceed, Awad explained.

The startup is working with agent developers, payment processors, merchants and fraud-detection providers to issue and recognize its credential. They include , and Socure, among others. Unless agents can establish that they are acting on behalf of legitimate people or businesses, 鈥渁gents will just get blocked everywhere,鈥� Awad said.

AI can also make fraud easier to scale

Ironically, the same technology that allows legitimate agents to do more tasks can also help fraudsters operate faster.

In the past, identity fraud involved someone getting their hands on stolen personal and business information, creating a credible-looking identity, and then repeatedly applying for bank or credit card accounts until an institution approved one. At one point, the process took significant time and manual work. But today, AI agents can automate parts of it and run continuously.

鈥淚t鈥檚 fraud on steroids right now,鈥� Awad said. 鈥淚t鈥檚 so easy, it鈥檚 so cheap, it鈥檚 so fast, and it鈥檚 24/7.鈥�

Reports of AI agents bypassing restrictions have also raised questions about how to identify and control autonomous software. , for example, recently reported incidents in which its models took unauthorized or deceptive actions, including activity involving the e platform.

Baselayer鈥檚 technology would not keep a model from disregarding instructions or exploiting a vulnerability, Awad acknowledged. But its goal is to verify an agent’s credentials when it attempts to interact or transact with an outside party.

Without a way to identify themselves, he said, legitimate agents may resort to trying to get around websites鈥� restrictions just to be able to complete their assigned tasks. Or, they could simply become less useful because they are repeatedly blocked as suspected bots.

Competing to establish a standard

M13 managing partner told 兔子先生传媒 News in an interview that he met Awad about a year before his firm invested in Baselayer. At the time, he saw the startup primarily as a provider of Know Your Business technology.

鈥淭he business did not feel like a business of the future,鈥� he admits. 鈥淚t just felt like he was solving a KYB banking verification problem.鈥�

The investor鈥檚 view changed as more companies began exploring payments made by AI agents and Baselayer began applying its business-identity data to the field.

鈥淓very agent ultimately is going to have to be tied to something real, and they understand the real world,鈥� Alomar said.

He believes Baselayer鈥檚 existing data, identity network and relationships with financial institutions give it an advantage over a startup entering the market from scratch.

“AI agents are rapidly becoming economic actors, but the identity infrastructure underneath commerce was never designed for software that can open accounts, make purchases, move money or enter into transactions on someone else’s behalf,” Alomar added. 鈥淭hat creates an enormous new trust problem, and we believe identity will become one of the foundational infrastructure layers of the agentic economy.鈥�

So far, no dominant standard exists. But Baselayer still must work to persuade agent developers, merchants, financial institutions and payment companies to recognize its credential.

That could take time. Awad said relationships with financial institutions typically take 12 to 18 months to establish, while large merchant partnerships can take up to 24 months. Baselayer may be able to reach some institutions more quickly, however, through its existing reseller relationships.

The company also sees potential use cases beyond payments. For example, Alomar said the technology could eventually authorize agents involved in cryptocurrency transactions or smart contracts, among other things.

鈥淭his is not just a fintech business 鈥� it鈥檚 a security business,鈥� he said. 鈥淚t begins with payments, but ultimately that technology applies directly to anywhere that an agent is making a decision that you need to verify it is permitted to make.鈥�

Related 兔子先生传媒 queries:

Illustration:

A decade ago, Espahbodi co-founded aerospace and defense startup accelerator and moved back to the U.S. to expand it. On the advice of friends at , he opened an office in El Segundo, California, near , just as more alumni of that company were leaving to launch hard-tech startups of their own and next-generation defense startups including were emerging.

Espahbodi eventually sold his stake in Starburst and launched , which invests in companies spanning industrial infrastructure, manufacturing, energy and water desalination. The firm has backed 14 companies since making its first investment in January 2023.

He also advises federal agencies on working with nontraditional, venture-backed companies. In an interview with 兔子先生传媒 News, he discusses how AI is changing hardware economics, why software investors are rushing into industrial technology, and what he believes many of them misunderstand about the sector.

This interview has been edited for length and clarity.

兔子先生传媒 News: What led you to leave Starburst and launch Generational Partners?

Espahbodi: About four years ago, I noticed that my friends from SpaceX were leaving the space vertical and moving horizontally across physical industries. I reached an inflection point: I didn鈥檛 want to remain locked into the space sector. I wanted to follow my friends.

I sold my equity in the accelerator, and part of the investment team left with me to start Generational Partners. For the past four years, we鈥檝e invested in what you might call the SpaceX-mafia and hard-tech sectors 鈥� anything involving industrial infrastructure, manufacturing, energy or water desalination.

We made our first investment in January 2023, in a North Dakota-based drone company. It was a trial by fire and an opportunity to prove the thesis. We鈥檝e invested in 14 companies since then.

You were already investing in physical, safety-critical industries before the generative AI boom. Has AI materially changed where you invest, or has it mainly reinforced your existing thesis?

Espahbodi: I tend to arrive earlier than others. I embraced the idea that hardware does not have to be capital-intensive. People often confuse hard tech with deep tech, but nomenclature aside, you don鈥檛 need to invest in science to win in these categories.

AI has dramatically changed that narrative and encouraged more people to get on board. I鈥檓 not looking to invest in science. I don鈥檛 necessarily see opportunities in quantum computing, nuclear fusion or other technologies being spun out of laboratories.

People who worked at companies such as SpaceX, and laid their companies鈥� foundations digitally. AI has significantly improved that augmentation and performance, enabling these companies to tackle legacy industries more aggressively and, more importantly, with new business models.

Another major component of the AI question is that frontier labs have become more expensive and capital-intensive than traditional hardware companies. The success of frontier AI labs, combined with the SpaceX IPO becoming an enormous wealth-creation event, creates a new environment. It raises questions about what is truly capital-intensive, what makes a product or its intellectual property defensible, and where companies are reengineering products around different business models.

Hardware has historically been capital-intensive, slower to commercialize and difficult to scale. Under what conditions does its technical defensibility compensate for those challenges?

Espahbodi: Fundamentally, it comes down to the business model. I look for creative software talent combined with commoditized hardware, significant customer demand and a new business model.

One of our portfolio companies was founded by the team that built the factory for user terminals. When you buy a retail Starlink antenna, these people built and scaled the assembly line that produced it at high volume.

While deploying those terminals globally to provide internet access, they observed that poverty often stemmed from a lack of access to clean water. They asked whether they could replicate the proliferated satellite-and-user-terminal architecture for edge water desalination.

Rather than investing in multibillion-dollar, nation-state infrastructure like that used by Gulf countries, they wanted to mass-produce every component in a vertically integrated stack. Their goal was to produce a cooler-sized device that could clean water at the point of need.

used a digital, software-based approach to build the bill of materials needed for mass manufacturing. AI is part of its business and operations, but the company鈥檚 real innovation was inverting the infrastructure model and scaling it.

I helped Vital Lyfe win its first customers within the and . Those organizations can use its devices in the field rather than shipping pallets of bottled water by air freight. That created a signal for overseas partnerships and nonprofit humanitarian-aid applications. It showed that there could be a different way to provide clean water.

Those are the kinds of unique business models that excite me.

What other companies founded by SpaceX alumni demonstrate how hardware businesses can overcome the traditional challenges of the sector? What can these founders build today that would have been difficult five years ago?

Espahbodi: Another example is the team SpaceX recruited to build the autonomous drone ships that catch boosters in the middle of the ocean. The team included former Coast Guard personnel and oil-and-gas technicians.

At SpaceX, they had the freedom to use software and AI tools to automate station-keeping 鈥� the ability of those drone ships to position and navigate themselves and reach the right location.

That team spun out and brought in many former colleagues to change commercial maritime shipping. They retrofit legacy boats operating in harbors and waterways and move supply-chain goods.

They brought a digital-first foundation to automating the controls on tugboats and barges. That had never existed before because the communications link to those ships didn鈥檛 exist. Starlink changed the concept of operations. The company can use its software expertise to change how physical devices operate aboard these boats and allow their sensors to send signals anywhere in the world.

That makes it possible to retrofit and overhaul how legacy shipping vessels navigate harbors and waterways in the U.S. It鈥檚 another example of SpaceX alumni applying the playbook and technologies they learned at SpaceX to a much broader commercial industry.

You鈥檝e said AI is eroding traditional software moats. What evidence are you seeing that investors are responding by moving into hardware and industrial technology?

Espahbodi: I meet many software investors who feel they鈥檙e missing out on hardware but don鈥檛 necessarily understand it. I鈥檝e met beauty investors who now say they鈥檙e defense-tech investors.

Los Angeles is a hotbed of firms that historically invested in software, media or consumer packaged goods. But people forget that Southern California, particularly El Segundo, is the aerospace capital of the world and has the largest concentration of mechanical-engineering talent.

Across the region 鈥� from China Lake to San Diego 鈥� technicians, builders and vocational talent are intersecting with the democratization of software and access to AI tools. Many local VCs have never taken advantage of the hardware talent located around them, so they鈥檙e being thrown for a loop.

Ironically, Bay Area VCs have been among those leaning most heavily into this. But it鈥檚 happening everywhere. I鈥檓 in Washington, D.C., now, and one of the first investors in , the hypersonic missile company, was in Virginia 鈥� before and others became involved.

Los Angeles VCs in particular know there is a talent war underway and that many people are leaving established companies to launch new businesses in these categories. But they struggle to underwrite those deals. They don鈥檛 know how to distinguish a strong opportunity from fear of missing out or something merely cosmetic.

So investors鈥� lack of experience in the space isn鈥檛 deterring them from writing checks or competing for deals?

Espahbodi: You have to ask why. The answer is their limited partners.

Sophisticated allocators, such as endowments, foundations and pension funds, along with more FOMO-driven family offices and high-net-worth investors, are watching this wave of SpaceX, and Anduril alumni create new companies and raise extraordinary rounds.

Many of those companies are no longer raising solely to pursue intellectual property. They鈥檙e building war chests to acquire other companies. The lines between private equity and venture capital are blurring. VC-backed companies are doing private equity-style buyouts, while venture deals are bringing in private equity checks.

That leaves LPs pushing for more. The success of the frontier AI labs has also perpetuated a fear of a 鈥淪aaS apocalypse,鈥� which I don鈥檛 think is real 鈥� although I sometimes question 鈥檚 1聽stock price for fun.

It creates what venture does best: tourists and FOMO. LPs ask why their managers aren鈥檛 investing in the same companies and how they can participate, raise more money and show that they aren鈥檛 missing out. That鈥檚 how I鈥檝e seen investors unfamiliar with these sectors enter the market.

Some of the largest Silicon Valley firms … missed this dynamism wave. Now they鈥檙e leaning in hard, sometimes at ridiculous valuations for companies that have yet to produce anything.

If more venture funding continues to flow into defense, aerospace and industrial technology, what prevents hardware from developing the same problems software experienced, including too many competing companies?

Espahbodi: Bring it on 鈥� hard and fast, and as much as possible.

Venture as a category exists because it was always about hardware. I would argue that the SaaS era, from the dot-com boom until now, was a blip compared with what venture was originally intended to underwrite.

I would move away from the hardware-vs.-software distinction and ask who is reframing the business model. Is there a way to reengineer a combination of software and hardware to unlock customer value? That鈥檚 the more important question.

How important is geography for these startups? Does locating near a major government customer help a company win contracts, and how do startups navigate procurement if they aren鈥檛 based near Washington, D.C.?

Espahbodi: It鈥檚 a common misconception that Washington is where the money is. The Los Angeles Air Force Base houses , which is another way of saying it holds 鈥檚 wallet. El Segundo makes the purchasing decisions for the fastest-growing portion of the military budget.

Washington is a place of considerable activity that needs to be influenced. Venture has never had this degree of influence on an administration and its executive orders. We鈥檙e also seeing portfolio companies backed by influential investors win government contracts worth as much as $1 billion at a time. That鈥檚 extraordinary.

Geographically, companies need to be where the talent is as much as where the customers are. Government customers should signal what matters, but companies shouldn鈥檛 organize themselves entirely around the government.

My catchphrase is that I want everyone to be commercially focused but mission-aware. I don鈥檛 want them to be mission-focused on the government. I want government to signal what it cares about while companies remain commercially focused.

The talent war for this convergence of hardware and digital technology is centered in Southern California. If you aren鈥檛 building and recruiting there, you鈥檙e falling behind. I like that the Bay Area is trying to attract more hardware talent and capitalize on the automotive and humanoid-robotics markets.

But I think the talent base for the factory of the future starts in Southern California and can then be used as a model for expansion into other places, as companies such as Anduril have done in Ohio and Louisiana.

We invested in a company founded by people from SpaceX and . They immediately moved to Austin to build a smart factory for raw-material processing. They wanted to automate the process at its source.

The largest concentration of cotton farming is around Lubbock in the Texas Panhandle. The company is building automated factories from the ground up to mill cotton into yarn and then complete the digital, vertically integrated stack by producing textiles at prices that beat outsourcing to China, Vietnam and other countries.

It sounds crazy, but the founder is determined to do it. If you can prove the model in textiles, you can apply it to copper. If you can do it with copper, you can do it in pharmaceuticals. From there, it could go in any direction.

Do startups located near Space Systems Command have an advantage?

Espahbodi: Not for that reason alone. The advantage is that they鈥檙e part of the ecosystem and geography. They鈥檙e spending time in the same bars and restaurants, and their children attend the same schools. They鈥檙e witnessing the same velocity.

Space Force itself is facing greater demand than ever to protect assets in space. Whatever happens with funding for individual programs, it remains the fastest-growing portion of the Pentagon budget.

I don鈥檛 think startups should locate there solely to be close to the customer. They should be there for the talent they need to build.

Related 兔子先生传媒 queries:

Related reading:

• Sector Snapshot: Space Tech Startup Funding Orbits New Highs
• Sector Snapshot: Defense Startup Funding Hits An All-Time Record As VCs Begin To Eye Exits

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.

After a week of multiple billion-dollar-plus rounds, startup investors have reduced the number of zeroes on their funding checks. This past week, the largest U.S. startup funding rounds were in the hundreds of millions, topped by a $550 million financing for AI infrastructure company and a $308 million investment in space vehicle developer .

The remaining list of big rounds featured mostly AI-focused companies in sectors including investment management, networking, coding and marketing as well as some energy and biotech. Data center developer , also made official its previously reported $3 billion-plus raise, co-led by , and .

1. , $550M, AI infrastructure: Temporal Technologies, developer of an open source platform for building and operating long-running AI agents and other enterprise systems, secured $550 million in Series E funding at a $12.55 billion valuation. , , , and led the financing for the Bellevue, Washington-based company.

2. , $308M, space tech: Redondo Beach, California-based Impulse Space, a developer of space vehicles for moving payloads across and between orbits, secured $308 million in Series D extension funding. The financing brings the combined round total to $808 million.

3. , $250M, investment management: Ridgeline, an AI-enabled investment management platform, picked up $250 million in a Series E funding round. The financing, led by founder and chairman , set a $1.45 billion valuation for the Incline Village, Nevada-based company.

4. , $205M, networking: Wayne, Pennsylvania-based Cornelis Networks, a developer of networking technology for AI and high-performance computing workloads, closed on $205 million in new funding backed by .

5. , $200M, AI software development: San Francisco-based Factory, a provider of AI tools for enterprise software development, announced a $200 million funding round at a $5 billion valuation, backed by a long list of venture firms and individual investors.

6. , $180M, AI marketing: Profound, a startup offering marketing software to help users appear more prominently in AI results, raised $180 million in Series D funding at a $1.8 billion valuation. and led the financing for the New York-based company.

7. (tied) , $150M, foundational AI: Arcee AI, a developer of open-weight AI models, closed on $150 million in Series B funding at a valuation of more than $1 billion. , and led the round for the San Francisco-based company.

7. (tied) , $150M, gaming: Nex, a developer of family-oriented digital games that rely on body motion rather than controllers, secured $150 million in new equity and debt financing, including a Series E led by and . The San Francisco company did not break out how much of the round consisted of equity.

9. , $135M, geothermal energy: Mazama Energy, a Seattle-based geothermal energy developer specializing in superhot rock geothermal power, picked up $135 million in a Series B round led by and .

10. , $123M, biotech: Sling Therapeutics, developer of a small-molecule therapy for thyroid eye disease, closed on $123 million in Series C funding. led the financing for the Ann Arbor, Michigan-based company.

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 Sept. 12-18, 2026. Although most announced rounds are in the database, there may be a small time lag, as some rounds are reported late in the week.

Illustration:

Ten years ago, a seed-stage startup needed a product, a team and a pitch deck to raise capital. Today, that’s just the start. Technology and strategy have become inseparable, each fueling the other, and the rules that once defined success have quietly shifted under everyone’s feet.

Last month, my firm reviewed more than 2,500 inbound applications. Here are the key shifts we鈥檙e seeing in the startup ecosystem at the seed stage.

Broadening capital strategy

Equity is a powerful tool for building high-growth companies. But it鈥檚 no longer the only option. Non-dilutive growth capital is increasingly playing a strategic role for companies with revenue visibility and clear ROI channels.

For example, we recently financed a company with $1 million in growth capital it needed immediately to expand its team and infrastructure. Raising that through equity alone would have likely taken months, with significant time and execution cost along the way.

We鈥檙e now writing financing checks like this on a near-weekly basis.

Distribution focused

Leading with a 鈥渂etter鈥� product isn鈥檛 enough to propel growth. The breakout companies are investing in building stronger distribution systems 鈥� aka what founders refer to as 鈥渢raction.鈥� Distribution is a critical moat for early-stage startups. Rapid scaling is no longer achieved by launching new products; it鈥檚 through distribution loops.

Distribution is something startups can now architect intentionally with social platforms, marketplaces and other ecosystems. One of the most common founder mistakes we see is delaying the distribution strategy until after the product launch. At that stage, the architecture is harder to retrofit. Strong startups design distribution before they scale their product.

For example, some of the fastest-growing startups now design their products around existing ecosystems from day one 鈥� building apps that tap into merchant marketplaces, AI tools distributed through or Teams integrations, or fintech products embedded directly into banking and payroll workflows. In many cases, the distribution channel becomes more valuable than the underlying product itself.

One of the most common mistakes we see is founders postponing distribution strategy until after the product is built. By then, the architecture is far harder to retrofit. The strongest startups design distribution into the company before they scale the product itself.

Learning over speed

鈥淢ove fast鈥� is often dolled out as the best startup advice. Operating in a fast-paced environment remains a strategic asset, but it is not enough to maintain a competitive advantage.

Everyone is fast. It鈥檚 no longer a unique attribute. Instead, learning velocity is becoming the defining advantage in early-stage startups. How quickly can you reduce uncertainty? Competitive edge is achieved not by executing blindly, but by closing knowledge gaps faster than everyone else. Execution without learning equals wasted motion.

The founder focus advantage

Last year, my team reviewed close to 25,000 applications for our investment funds and bespoke accelerators. The ones that stand out are the companies doing the fewest things exceptionally well. The most-fundable companies can describe their business in one tight sentence. They can also defend exactly what they are not doing.

Disciplined constraint is one of the highest-leverage traits in venture-backed companies. When we review applications, this pattern consistently stands out.

When a company is focused, the residuals compound: stronger early retention, faster iteration cycles, cleaner capital deployment. In a capital-selective market, focus compounds faster than ambition.

Based on tens of thousands of applicants, close to 82% of the ones that stayed in business a year later had a strong go-to-market foundation in their deck. GTM is built on agility and learning fast.

AI as infrastructure, not experimentation

There鈥檚 no lack of interest in AI. But there is an implementation problem. We鈥檝e seen companies struggle when AI is approached as experimentation rather than architecture. Rather than bolting tools onto already fragmented stacks and workflows, designing intelligent systems should be mapped from the ground up.

More than 78% of the founders applying to today are leveraging AI in at least one way in their startup.

The most successful playbook combines execution with operational clarity and emphasizes infrastructure over experimentation. We鈥檝e seen successful implementations that center around two practical paths. The first is validation, with rapid prototypes and identifying market signal opportunities before investing in a full build. The second is system, designing and integrating custom AI agents directly into operating workflows for revenue-generating companies facing operational complexity. Both are required to move AI agents from concept to capability. A disciplined system design often matters more than flashy tooling.

Marketing is the moat

Marketing execution is one of the largest performance gaps we see across early-stage startups. Startups lose when they don鈥檛 distribute fast enough once there is something worth selling. Marketing is the propeller for the distribution engine.

Most startups fail at marketing because it is a business function that becomes a founder hustle with support from one junior hire. But breakout growth requires process, cadence and accountability. That鈥檚 not achievable without an experienced team and clear plan.

One of the biggest mistakes founders make is treating marketing as something that starts after launch. Founders must create unique strategies, test them and then analyze what works and what doesn鈥檛. From there, they must keep iterating and creating to unlock the most product-market fit and traction.

If we see classic strategies in a pitch deck, it is an auto-reject. And beyond being unique, your strategies must have been tested by your team.

The key is simple: Test ideas early, measure what actually works, refine aggressively and scale the strategies that compound over time.

, a serial technology entrepreneur since age 12, is now a value-driven venture capitalist with a track record of backing more than 1,000 startups across the globe. He is a co-founder and general partner at , one of the world鈥檚 most active venture capital firms. Before becoming involved in venture investing, he built and scaled into the world鈥檚 largest debate-focused social network and edtech platform, reaching millions of users and serving students across more than 500 school districts, colleges and universities.

Related 兔子先生传媒 query:

Illustration: