The Company Before the Company

FGV Capital

FGV Capital's best asset may be the conflict it has to keep under control. The growth platform helps the firm win cap tables; the investment team still has to prove those commercial relationships do not choose the investments.

That question sits inside an oversubscribed $35M Fund II, closed above a $25M target and including co-investment capital. The August 25 close brings FGV Capital's reported AUM above $60M while Fiat Ventures and Fiat Growth come together under one FGV brand.

Marcos Fernandez leads the capital arm as Managing Partner, alongside General Partners Drew Glover and Alex Harris and Partner Rohit Ramkumar. This fund is being sold as the institutional version of an operating model the team began building in 2018.

That proximity can make diligence sharper. It can also make the boundary between commercial work and investment judgment harder to see. Drew Glover told TechCrunch that the advisory and investment businesses remain separate entities with clear processes intended to keep business relationships from biasing capital decisions.

Reinsurance Group of America, MassMutual, Bank of America, and Stellar Development Foundation are among the named LPs backing Fund II. Those are more than allocator logos for a fintech fund. They are regulated institutions with distribution, operating knowledge, and potential partnership value for companies building financial infrastructure.

TechCrunch reports that FGV plans to invest $1M-$1.5M in at least 25 companies over 2 years, with 13 already backed. The mandate centers on fintech, financial access, and AI, with the broader portfolio extending into health, insurance, commerce, property, climate, and security.

Fiat Ventures announced its first $25M fund in 2022 with Invesco Private Capital as anchor LP. Fund II increases the available capital, but the more interesting expansion is in the job FGV wants the manager to perform: source through operating work, help build distribution, lead or co-lead more rounds, and connect founders with institutions that can become customers or partners.

That model deserves attention without the victory lap. FGV reports more than 40 portfolio companies and more than $250M in revenue driven after investment, but Fund II still has to turn platform access into repeatable company outcomes and fund performance.

Founders will learn whether the operating system behind the check creates leverage or another set of meetings. LPs will learn whether proximity improves selection without compromising independence. FGV Capital has raised enough money to scale the experiment. The harder asset to compound is trust.

Fasset

Stablecoins make a transfer legible to software. They do not make a bank, regulator, liquidity provider and local customer agree on the route. That work still belongs to institutions, contracts and corridor-by-corridor operations.

Fasset just put $68M behind that work. The company announced a Series C led by Japan's SBI Group at a stated $1B private valuation, 3 months after a $51M Series B. Fasset has now raised $119M in 2026 and says lifetime funding exceeds $150M.

Mohammad Raafi Hossain, Co-Founder and CEO, and Daniel Ahmed, Co-Founder and COO, built Fasset around a blunt constraint: access to money and assets still depends heavily on geography. Own Network connects banks, payment companies, telecom providers, liquidity partners and settlement systems across more than 100 banking corridors.

Stablecoins can sit underneath that flow without forcing every customer to think like a crypto trader. Fasset says AI chooses among rails, currencies and liquidity sources based on cost, speed and availability. The complexity moved; it did not disappear.

That is where SBI becomes more useful than a unicorn headline. SBI first invested in Fasset in May and announced a partnership with SBI Remit in June. Now the group is leading the Series C while both sides work on routes connecting Japan with Asia and other emerging markets. This is capital attached to distribution and financial infrastructure.

The scale is serious, with an asterisk. Fasset reports more than $40B in annualized transaction volume, 3M+ wallets, 1,000+ enterprises and customers across 125 countries. Mohammad Raafi Hossain told CoinDesk that revenue grew about 6x and Fasset has been profitable for 12 consecutive months.

Fasset did not disclose revenue, profit, margin, active-wallet counts or corridor economics. The $40B flow proves money is moving, not how much value Fasset keeps after liquidity, compliance and banking partners take their share.

More corridors can create a larger network, but each market adds licences, counterparties, currencies and failure modes. Agentic routing can choose a path; it cannot negotiate away the rules around it.

The capital buys Fasset room to connect more institutions and test more routes. It does not make regulation uniform or liquidity cheap in every market. Those differences are the business.

The $1B valuation says SBI accepted a price. The next proof is whether Fasset can turn institutional access and company-reported volume into repeatable economics without making the customer carry the complexity. Stablecoins can shorten settlement. Fasset still has to make the whole route hold.

Discernis

Litigation teams do not get extra time because the evidence arrived buried inside 15 million documents. The clock keeps moving while reviewers search for the few facts capable of changing a case.

Discernis is building for that pressure. The New York City legal intelligence company has raised a $2.5M Seed round led by Newfund, with Triple Impact Capital, Remarkable Ventures, and C2 Ventures participating.

Founder and CEO Robert Kwiatkowski, PhD, holds a Columbia computer science doctorate and worked on AI at AWS. Discernis is applying that training discipline to litigation and investigations, where a convincing answer means very little if the system missed the document sitting 400,000 files away.

Most document tools narrow the field. They sample a collection, rank likely hits. Discernis says its system reads the entire collection, analyzes connections across documents, and works from the review protocol a legal team already wrote.

That distinction turns speed into a heavier obligation. If software claims to understand the whole record, attorneys will expect it to explain why evidence matters, preserve the chain of reasoning, and survive scrutiny from clients, regulators, judges, and opposing counsel.

Discernis reports that its AI reviews more than 50,000 documents per hour and is already used by Am Law firms on active matters. Those are company-reported claims, not an audited victory lap. They still frame the commercial opening: legal teams are under pressure to review more data without letting cost, deadlines, or confidentiality turn accuracy into a luxury.

The deployment model may decide whether that opening becomes a market. Discernis can run on-premises or inside a customer's secure cloud, with inference performed inside that environment and no call to a third-party commercial AI API. For firms handling privileged files, government investigations, and outside counsel restrictions, data control is part of the product, not procurement paperwork stapled on later.

The $2.5M will support further product development and expansion across large law firms and corporate legal departments. Capital can deepen the technology and put more sellers in front of buyers. It cannot shorten the trust cycle by decree.

Discernis now has to prove that whole-collection analysis is not only faster than traditional review, but repeatable, explainable, and defensible across matters with different facts, infrastructure, and risk. Legal AI will not be judged by the cleanest demo. It will be judged by the ugly case, the bad deadline, and the document nobody knew to search for.

Vanguard / Altruist

Independent advisors can own the client relationship and still depend on somebody else for custody, clearing, software, service, and pricing. Vanguard's agreement to acquire Altruist puts that dependency under a new owner.

The companies signed a definitive agreement on August 26. Closing is expected later this year, subject to customary conditions and regulatory approvals. Official terms were not disclosed. Axios reports $4.6B in cash; Reuters says The Wall Street Journal put the value near $4B.

Vanguard is buying the operating layer beneath an independent advisor's client relationship, not another fund wrapper.

Altruist combines a self-clearing brokerage with account opening, trading, portfolio management, billing, reporting, and the client experience. The company says more than 6,000 advisors use the platform and that the number of RIA firms served grew 112% year over year.

Jason Wenk founded Altruist in 2018 after living with the industry's paper, transfer, pricing, and capacity problems inside advisory firms. The company has since raised more than $600M, including a $152M Series F led by GIC in 2025 at a company-stated $1.9B valuation.

Vanguard already knew the asset. It first invested in Altruist in 2020, and former Vanguard chairman and CEO Bill McNabb sits on Altruist's board. Salim Ramji is turning that six-year relationship into ownership.

The stated logic is an advice-capacity problem. Many Vanguard fund investors work with advisors. More people could use advice than the industry can currently serve. Altruist gives Vanguard custody infrastructure, advisor workflows, and proximity to independent firms. Vanguard gives Altruist capital capacity, investment expertise, and a much larger institutional base.

The structure matters. After closing, Altruist is expected to keep its leadership, brand, advisor focus, and standalone operating model. That promise recognizes what Vanguard is paying to preserve: speed, product judgment, and trust with firms that built their identity around independence.

Advisors will judge the promise through pricing, service, product velocity, custody reliability, data use, and whether Altruist can stay close to their operating reality while owned by an institution with its own direct investor relationships.

For Jason Wenk, the decision was about what institution Altruist could become after the founder is no longer the center of it. For Vanguard, it is a decision to compete closer to the advisor and the client.

The deal puts scale behind Altruist. Keeping independence useful will be the harder integration.

Blank Street

Blank Street's $105M deal puts 2 different kinds of money across the counter. $75M goes into the company, while $30M buys shares from existing holders.

That distinction matters because only the first pile has to open stores, build supply, hire teams, stretch the menu, and prove a New York and London brand can travel west without losing its shape. The second gives shareholders liquidity.

An August 24 SEC filing records a fully sold $104,999,981 equity offering to 6 investors, with the first sale on August 7. Axios reports General Atlantic led the transaction, and Financial Times-derived reporting places the post-money valuation near $650M.

Existing investors General Catalyst, Left Lane Capital, and Tiger Global also participated, according to that reporting. The primary financing is being described as a $75M Series C, while the filing itself does not name investors, label the round, or state a valuation.

Issam Freiha and Vinay Menda started Blank Street with a coffee cart at Brooklyn's Wythe Diner in 2020. Issam Freiha is now CEO, Vinay Menda is president, and their small-format model spans a reported 106 stores across the US and UK. The planned California openings turn that footprint into an operating test across new real estate, labor markets, supply routes, and customer habits on the West Coast.

The original bet was economic before it was aesthetic. Smaller footprints could reduce fixed costs, while quick service and a mobile app added convenience.

Blank Street now has to prove that system at a different scale. The company opened in Studio City on August 12, with Beverly Hills, West Hollywood, and Malibu planned as part of its California push.

The menu is moving too. Financial Times-derived reporting says Blank Street wants to grow beyond morning coffee through matcha, snacks, and ice cream. That is a frequency strategy: give the same customer another reason to walk in after lunch.

General Atlantic is underwriting whether a recognizable brand can turn more than 100 locations into a repeatable system across markets and dayparts.

The reported $650M valuation creates a demanding scorecard. Store count is visible, but the harder evidence remains private: same-store sales, contribution margins, customer retention, labor productivity, and whether California units earn the right to keep multiplying.

Blank Street has already shown it can make a cup travel through culture. The $75M question is whether the company can make the economics travel with it, while the other $30M reminds everyone that growth and liquidity can share a headline without doing the same job.

Faro AI

A clinical protocol is treated like a document because that is what humans can circulate. Software has a more demanding question: what did every line mean to the rest of the trial?

Faro has raised a $37.3M Series B to answer that question with structured data and AI agents. Merck Global Health Innovation Fund and S32 co-led the round. General Catalyst, Northpond Ventures, Polaris Partners, PTX Capital, and Zetta returned, while Ankona Capital joined as a new investor.

Clinical development is a chain of scientific, medical, regulatory, and operational decisions. Endpoints, visits, eligibility rules, and timing choices can change patient burden, site workload, cost, and the systems required to run the study.

An AI model can produce clean prose and still misunderstand the trial.

Scott Chetham and Ross Jaffe founded Faro around the idea that protocol design should begin as structured information rather than a static file. Patrick Leung, CTO, is extending that foundation into AI agents that can reason across the intent, relationships, and constraints inside clinical development.

Faro is not asking pharmaceutical teams to trust a chatbot with a regulated workflow. It is building the ontology and data models beneath the agent, so the system has something more reliable than a pile of documents to reason over.

The company says its platform is used by 6 of the world's 10 largest pharmaceutical companies. Recursion has separately disclosed using Faro for protocol design and downstream automation, and Faro announced a Bristol Myers Squibb collaboration this year. That does not prove every agentic workflow works at scale, but it puts Faro in the rooms where proof will be demanded.

Faro has now disclosed at least $75.6M in financing. The Series B will expand its agentic AI capabilities and support customer deployments.

The commercial opportunity is moving beyond faster protocol writing. Structured study definitions can feed document authoring, EDC builds, site budgets, vendor specifications, and workflows that still lose time to re-entry, reconciliation, and translation between teams.

That raises the standard. In clinical development, automation has to preserve context, traceability, and human oversight while the work changes hands. A fluent mistake can travel farther than a slow one.

Faro now has the capital to prove its structured foundation can carry more of the development process. The next test is whether its agents can move faster without making the people accountable for the trial lose sight of why each decision was made.

Voya Energy

A data center can secure the land, the racks, and the GPUs before it secures permission to power them. The machine is ready. The grid calendar is still clearing its throat.

Voya Energy has raised a $35M Series A led by Energy Impact Partners, with John Doerr, Mantis VC, StepStone, Founders Fund, Overmatch, and Seven Stars participating. The capital will support technology development, 2027 deployments, and manufacturing scale-up in 2028.

Richard Wang, Matt Horton, and Steven Kaye are building around low-grade scrap aluminum. Voya forms it into pellets and feeds them into a generator where aluminum reacts with air and water. A proprietary electrolyte and metal-air cell convert that chemical energy into electricity without combustion or point-of-use air emissions.

The byproduct is aluminum trihydrate, or ATH, which Voya plans to route into industrial markets. The longer-term ambition is a closed loop that turns ATH back into fuel-grade aluminum. That remains a roadmap, not an operating result.

The packaging fits a physical customer problem. A power-constrained site may not have years to wait for transmission, a gas pipeline, or more grid capacity. Voya targets 2 MW from a 20-foot generator container paired with a 10-foot fuel container holding roughly 100 MWh.

At scale, Voya targets 100 MW of generation and 10 GWh of stored energy on one acre. The company reports conversion efficiency near 57%, versus roughly 35% for diesel, and says nearly a dozen companies are preparing pilots.

Those numbers describe the wager. They do not settle it.

Bloomberg reports that Voya has an in-house prototype, but no commercial-scale proof. The unnamed pilots still need to produce buyer-grade evidence: uptime, maintenance, delivered power cost, fuel logistics, permitting, and byproduct economics. The pilots will keep the score.

Richard Wang built Cuberg through its acquisition by Northvolt. Matt Horton has operated across Voltera, Rivian, Proterra, and Propel Fuels. Steven Kaye has spent 20 years in advanced materials and energy storage and holds more than 60 patents. Difficult chemistry becomes a business only when manufacturing, service, and customer operations stop treating it like an experiment.

Voya raised a $13M seed before this round, bringing disclosed funding to $48M. The new capital buys the company a chance to turn aluminum from a compact energy story into dependable site infrastructure.

The first deployment will have to prove more than a generator. It will have to prove the entire loop.

Teragen Energy

Natural gas, hydrogen, public clean-energy capital, and deep-science venture money rarely arrive at the same answer. Teragen Energy just gave them one $6M reason to keep arguing around the same table.

The Southborough, Massachusetts company closed an oversubscribed Pre-Seed round co-led by BEVC and Energy Capital Ventures. AP Ventures, AIC Ventures, the Massachusetts Clean Energy Center, and UntroD Capital Asia participated.

That syndicate does not represent one tidy view of the energy future. Energy Capital Ventures invests around natural gas. AP Ventures specializes in hydrogen and industrial decarbonization. MassCEC develops public clean-energy markets. BEVC backs hard science.

They can disagree about the destination while recognizing the same problem: large power users need firm capacity faster than grids can always provide it.

Ruofan Wang, PhD, CEO and Co-Founder, co-invented Teragen's solid oxide fuel cell architecture at Berkeley Lab. Joshua Persky, CTO and Co-Founder, brings more than 20 years in product engineering and manufacturing scale-up. Emerson Reiter, CCO and Co-Founder, brings the utility and project-development side.

Their machine converts natural gas or hydrogen into electricity through high-temperature electrochemistry, without combustion. Teragen says its metal-supported design can improve cost, power density, response, and ruggedness while producing near-zero local pollutants. Optional configurations add energy storage or carbon capture.

The U.S. Department of Energy cites scenarios in which data centers could consume 9.5% to 15.3% of U.S. electricity by 2030. A buyer does not get paid for winning an argument in the interconnection queue. The facility has to be online, reliable, permitted, and economical.

Teragen is still early. The announcement names no commercial pilot, customer, module rating, efficiency figure, installed cost, stack life, or degradation rate. The company is moving from prototypes toward first pilot projects, which means this round is funding the proof rather than standing in for it.

The $6M will expand testing and manufacturing infrastructure, grow engineering, scale the technology, and accelerate commercialization. That is where a clever architecture meets repeatable hardware, permitting, procurement, and uptime.

Teragen has attracted investors who see different fuel pathways but the same power constraint. The next test is whether one machine can preserve that optionality after a real customer starts measuring every hour it runs.

Transfyr

Anna Marie Wagner and Renee Wegrzyn are financing the space between a correct protocol and a successful experiment.

The protocol can list every reagent, instrument, temperature, and timestamp. It still may not capture the movement, room condition, equipment behavior, or judgment that made the work succeed in one lab and fail in another.

Transfyr has launched with a $25M Seed round led by General Catalyst. Lux Capital, Breakout Ventures, Factory, Neo, SV Angel, MVP Ventures, Underscore VC, Lyda Hill, and angel investors also participated.

Anna Marie Wagner, co-founder and CEO, previously led AI and corporate development at Ginkgo Bioworks. Renee Wegrzyn, PhD, co-founder and chief innovation officer, was the inaugural director of ARPA-H. Now they are building infrastructure for what usually disappears between scientific possibility and real-world execution.

Transfyr combines sensors with multimodal AI models to observe laboratory work as it happens. The system captures operator actions, environmental context, equipment telemetry, and process dependencies, then turns that physical record into machine-readable knowledge.

That is a different product from another electronic lab notebook or AI writing tool. Transfyr wants to isolate variability, diagnose failures, improve training and technology transfer, and translate manual work into automation-ready instructions.

AI can read the scientific record at extraordinary scale, but it cannot learn from what science never recorded. Robots can repeat instructions precisely, but precision does not rescue an incomplete instruction. The missing layer is the context between the written procedure and the physical event.

Public programs are moving toward the same problem. The National Science Foundation announced $380M for 20 AI-enabled programmable cloud-laboratory teams, while a Massachusetts grant supports BioBuilder's work with Transfyr on AI-enabled lab training.

None of that proves the commercial model. Transfyr has not disclosed customer names, revenue, deployments, or audited gains. The $25M now has to finance the work that makes physical AI real: robust sensors, clean data pipelines, lab deployment, security, biological-risk controls, and models that separate signal from ordinary motion.

The prize is scientific work that can move between people, laboratories, and machines without losing the context that made it work. The obligation is proving that observing more creates better decisions, not merely a larger archive.

Science already has a publication record. Transfyr is trying to give it an execution record.

PATH

PATH gets paid at checkout. Its environmental argument starts with the refill that happens later.

That distinction sits underneath PATH's new round. The amount was not disclosed. Agility Ventures, the corporate venture arm of Agility, is the only participant the company named. The money is meant to expand U.S. and international distribution, increase brand investment, support product innovation, and strengthen supply-chain and operating capacity.

The real sale starts after checkout.

PATH sells purified still, sparkling, and alkaline water in aluminum bottles designed to be reused and recycled. The bottle enters the market as a convenience purchase. Its environmental case gets stronger only when the buyer keeps it, refills it, and lets it replace another single-use bottle.

Distribution creates that opportunity. It does not complete the behavior.

Shadi Bakour, co-founder and CEO, built PATH with co-founders Amer Orabi, COO and President, and Ali Orabi, CMO. The company traces the idea to 2014, launched in 2015, and produced its first aluminum bottled-water format in 2016. By 2022, PATH said it had reached more than 50,000 retail doors and raised a $30M Series A led by Altos Ventures.

This 2026 round is different in one important way: PATH is withholding the size, formal series, lead investor, and full syndicate. What it is disclosing is the work.

More stores. More countries. More consumer awareness. More product development. A supply chain capable of supporting all of it.

That makes Agility Ventures an interesting name in the room. PATH did not publish an investor thesis, so nobody should write one for the firm. Still, a corporate venture investor connected to a global supply-chain business is participating in a round whose announced priorities include distribution and operational optimization. The overlap is worth noticing without pretending it is a strategy memo.

The harder proof starts after expansion.

A refillable package still moves through a business built around velocity, shelf turns, availability, and repeat purchases. PATH has to grow units without letting refillability become a label that disappears once the bottle leaves the store. International growth adds new retailers, logistics, price points, consumer habits, and local recycling systems to that equation.

The brand has already shown it can make aluminum packaging recognizable across retail, hospitality, sports, and co-branded partnerships. This capital gives PATH more room to make that proposition travel.

The next operating question is whether the behavior travels with it.

Arintra

A hospital can deliver the right care and still lose the argument over what happened in the chart.

The physician documents the encounter. The coder translates it. The payer applies its rules. Between those steps, a visit can become an incomplete claim, a denial, or revenue nobody realizes went missing.

Arintra has raised a $25M Series B led by Define Ventures, bringing total funding to $51M. Existing investors Peak XV Partners, Yale New Haven Health Center for Health Care Innovation, Endeavor Health Ventures, Y Combinator, Counterpart Ventures, Ten13, and Spider Capital also participated.

That syndicate is financing more than faster coding. Nitesh Shroff and Preeti Bhargava, PhD, are building one intelligence layer that starts with the code, stays inside the EHR, and carries its reasoning into documentation improvement, denial appeals, payer insight, and DRG validation.

Revenue cycle software has spent years breaking the work into point solutions. Each tool may solve its task while the provider still owns the gaps between them. Arintra calls its answer revenue assurance: connect care delivered to revenue earned, with an audit trail for every coding decision.

The company says its platform supports 23+ specialties across ambulatory, emergency, diagnostic, and inpatient settings, processing more than $5B in annual claim value.

One customer result has a name behind it. Mercyhealth told Healthcare IT News that Arintra supports more than 50,000 charts per month there, helped increase revenue by 5.1%, and cut pre-accounts-receivable days by roughly 50%. Coders shifted toward complex cases, denials, compliance, and provider education instead of chasing routine volume.

UC Davis Health says its teams can audit results approximately 50% faster. KLAS Research published an Emerging Company Spotlight on Arintra in April, while cautioning that findings can move as more provider interviews enter a small sample.

That caveat belongs beside the traction. Coding is not a forgiving place for AI to be charmingly wrong. Every captured dollar must remain compliant, every recommendation must survive an audit, and every specialty adds clinical and payer-specific complexity.

The Series B will fund more enterprise deployments, broader specialty coverage, and additional revenue-cycle work. Care happens once, but getting paid requires several systems and teams to agree on what the record means.

Arintra has capital to make that agreement more automatic. The next proof is whether one intelligence layer can remain explainable as the claims, specialties, payer rules, and consequences keep multiplying.

Unrivaled

Most athletes meet enterprise value on the far side of a contract negotiation. Unrivaled put it into the cap table.

Napheesa Collier and Breanna Stewart founded the 3-on-3 league in 2023 around a blunt idea: the players creating the product should own part of the property. Two seasons later, that idea has attracted an oversubscribed Series C led by Ten Pillars Sports Fund, backed by UC Investments.

Axios reports the round reached $106M, while Unrivaled says it surpassed an initial $100M target. The financing values the league at $650M, up from $340M after its Series B in September 2025.

The more interesting number sits on the other side of the ownership table. Unrivaled says players remain its largest shareholder group, with an equity pool now worth nearly $200M and up more than 550% since inception.

That value is company-reported, not an independent appraisal. Athlete equity is no longer a launch promise; it is becoming a material claim on an operating sports business.

Unrivaled has business evidence behind the valuation story. The league says Season 2 revenue reached $45M, up from $27M in Season 1, while merchandise revenue rose 130%, ticket revenue climbed 204%, and sold-out tour stops in Philadelphia and Brooklyn drew 39,751 people combined.

The television picture is less tidy. The Associated Press reports ratings declined in Season 2 while the schedule competed with the Winter Olympics and College Football Playoff. A league can grow event demand, direct audience, and revenue while still having to earn a larger, more reliable broadcast habit.

That is where Alex Bazzell and the new capital have work to do. The plan is a much better Miami arena experience, 8 tour stops in 2027, and a broader NIL program for top college players.

Returning backer Bessemer Venture Partners joined Ten Pillars alongside Jenny Just, Carmelo Anthony, Geno Auriemma, the Berman Family, Ashton Kutcher, Alex Morgan and Trybe Ventures, Dan Rosensweig, and Trae Young.

That roster can open doors, but it cannot play the possession. The $106M now has to turn a compelling athlete-ownership thesis into repeatable arena economics, tour operations, media value, player recruitment, and fan behavior without diluting the promise that made Unrivaled Basketball distinct.

The valuation says investors believe a player-owned league can become a durable sports property. The next seasons decide whether shared ownership remains part of the operating system when the crowds, costs, and expectations all get bigger.

SiFly

Thirty to 40 minutes of flight can make a multirotor useful while forcing every long mission to plan around the return, recharge, and second launch.

SiFly just raised a $20M Series A led by Shield Capital, with Qudit, BBK Capital, Alumni Ventures, and others participating. The round is meant to move the Santa Clara company from endurance proof into production, delivery, and deployment.

Founder and CEO Brian L. Hinman built the Q12 around a different operating window. SiFly says the electric VTOL aircraft can fly for 3 hours in forward flight, carry up to 10 pounds, and has completed thousands of flights across hundreds of testing hours. One test lasted 3 hours, 11 minutes, and 54 seconds, which the company says earned a Guinness World Record in its weight class.

The record gets attention. The mission economics decide whether the business works.

A public-safety team does not buy flight time as a trophy. It buys coverage over an incident. A utility does not care that a battery is impressive in isolation. It cares how many miles can be inspected before the aircraft, pilot, and work order have to reset. Agriculture turns endurance into acres only when the system survives weather, scheduling, maintenance, and the operating day.

That is why DronePort matters beside the Q12. SiFly is building one node to support multiple aircraft and consolidate launch, recovery, charging, mission control, and maintenance. The company says that model can replace dense networks of one-drone docks with broader regional coverage.

The financing will scale Q12 production, expand U.S. manufacturing and supply-chain capacity, support initial customer deployments, and advance DronePort field validation. SiFly also plans to build the customer-operations, go-to-market, and regulatory capabilities that hardware companies usually discover are more expensive than the prototype.

Shield Capital's thesis is straightforward: longer endurance can mean fewer aircraft, docks, operators, and battery swaps for the same work. Its investment also lands inside a market looking for more capable U.S.-made autonomous systems across public safety, infrastructure, agriculture, and national security.

Now comes the unforgiving part. SiFly has to turn test hours into production yield, backlog into delivered aircraft, and a record flight into lower cost per mission. The Q12 has already shown it can stay in the air. That proof will arrive in ordinary operating days, not a test record. The Series A is paying for SiFly to prove the operation can stay together around it.

Breedr

A calf can leave the ranch with a new owner, a new ration, and half its history trapped in the prior operation's paperwork. The next business sees the animal, but not always the decisions that shaped it.

Breedr has raised a $27M Series B to keep that record moving. Partech led through its impact fund, with Latitude, the growth-stage fund of seed investor LocalGlobe, and returning investor Outsiders Fund participating. The round brings total funding to $46.6M, while Partech's Arnaud Minvielle and Latitude's Remus Brett join the board.

Ian Wheal grew up on a cattle and sheep farm in South Australia. Ian Wheal founded Breedr in 2018 because producers can do careful work for years and still struggle to prove the quality created before the next buyer.

The software gives individual cattle a digital record from conception through processing. Genetics, weight, health, movements, medication, and carcass feedback can follow the animal through the 3 or 4 operations common in a beef lifecycle.

That continuity makes Breedr more than a farm-management app. The company also runs a data-verified marketplace and an embedded cattle finance fund. One layer records the animal, another helps it trade, and the third puts capital behind the inventory.

Breedr says more than 2M cattle are on the platform across 10 global supply chains, with close to $500M of livestock transacted in 2026. Customers range from family ranches to operations handling more than 100,000 cattle a year. Those are company-reported metrics, but they make the next question unavoidable.

Who gets the value when the record gets better?

The rancher needs better breeding and pricing decisions. The feedyard needs sourcing and performance visibility. The processor needs predictable supply. The lender needs to understand the animal behind the collateral. Breedr needs all of them to keep contributing data.

That pressure is arriving while the U.S. cattle base remains tight. USDA counted 86.2M cattle and calves on January 1, 2026, including 27.6M beef cows, down 1%. The 2025 calf crop fell 2% to 32.9M head. When supply cannot expand quickly, better information per animal becomes a commercial problem, not a software demo.

Ian Wheal and CTO Liam Farrell now have capital to grow the team, add producers, expand connected supply chains, and capture more data such as genomics. More records will make the system larger. Better handoffs have to make it valuable.

Breedr is building the memory layer for an industry that changes owners more easily than habits. The next proof is whether every participant earns enough to keep the record alive.

Gatik

Gautam Narang, Arjun Narang, and Apeksha Kumavat spent nine years narrowing autonomous freight to a stubbornly ordinary job: move goods between distribution centers and stores, then do it again tomorrow without becoming the reason a shelf is empty.

Gatik now has another $200M to prove that operating discipline can travel. Qatar Investment Authority and Koch Disruptive Technologies co-led the Series D, with Millennium Management, ARK Invest, Intact Private Capital, and other investors participating. The company did not disclose a valuation. The capital is for expansion, not a new research thesis.

The capital follows a commercial record that looks different from the usual autonomous-vehicle highlight reel. Gatik reports more than $600M in contracted revenue, 85,000 fully driverless orders completed, and 99% on-time delivery. Those are company-reported numbers, but they describe the scoreboard freight customers actually use.

The founders made a useful choice early. Gatik focused on medium-duty trucks running regional routes between facilities, where cargo is predictable, networks repeat, and customers can measure whether capacity showed up.

That lane has widened. TechCrunch reports dozens of fully driverless commercial trucks and a 41-truck PepsiCo deployment in Dallas, Phoenix, and Northwest Arkansas. Public customers also include Walmart, Loblaw, Kroger, and Tyson Foods. Isuzu supplies the vehicle platform; NVIDIA provides in-vehicle compute.

Freight pays for a truck arriving safely and predictably enough that a network can depend on it. Gatik Driver has to navigate highways and surface streets, adjust to changing stops, and fit inside customer systems that were not built for an autonomy demo. One excellent run is a milestone. Thousands of uneventful runs are a business.

Gatik says the round will finance more trucks, customers, cities, engineers, and operations staff. Management plans to move from dozens of driverless trucks toward thousands over time. KDT first led Gatik's $85M Series B in 2021, so its return carries the weight of an investor that watched the model mature. Intact Private Capital says it tripled its commitment.

The next proof will be harder than the last. Gatik must add vehicles and markets without loosening the safety case, service reliability, customer integration, or economics that made its focused model attractive. Autonomous freight has spent years asking people to imagine the future. This $200M round is attached to a less glamorous demand: make tomorrow morning's delivery boring enough that the customer stops thinking about the truck.

Alice

For nearly a decade, Alice's team went looking for the internet's worst behavior before it reached somebody else's platform. Apax is now putting $140M behind the idea that those receipts can become infrastructure for the AI era.

The company, formerly ActiveFence, tracked fraud, extremism, coordinated manipulation, and other abuse across the open web. That work produced Rabbit Hole, a dataset of real-world adversarial and harmful content. Alice uses those patterns to attack AI models before release, tune runtime defenses, and keep testing systems in production.

That distinction matters. An AI assistant can produce a bad answer. An AI agent with access to email, files, databases, and software tools can turn a bad answer into an operating event.

Noam Schwartz, CEO and co-founder, built Alice with Iftach Orr, CTO and co-founder, Alon Porat, Chief Customer Officer and co-founder, and co-founder Eyal Dykan. Apax Digital Funds led the round, joined by Samsung, SentinelOne, Maj Invest, MoreTech, and Phoenix Insurance. Resolute Ventures, Grove Ventures, CRV, Highland Europe, Vintage Investments, Norwest, NFX, and Claltech returned.

The financing brings Alice's total funding to $280M and puts Apax Digital on the board. Alice did not disclose the round type or valuation.

Alice says it is approaching $100M in ARR, its AI business grew more than 500% over the last 2 years, and more than 150 researchers study how models can be manipulated or fail. The company also says it protects more than 3B people online and works with 8 of the 10 leading AI model labs, including named relationships with Anthropic, Google, and Cohere.

WonderBuild stress-tests systems before launch. WonderFence applies runtime guardrails to prompts and responses. WonderCheck keeps testing production systems for drift, regressions, and new attacks. Rabbit Hole supplies the ugly source material connecting all 3, creating a loop from observed abuse to testing, defense, and monitoring.

The new capital will deepen that platform, expand the team and data behind Rabbit Hole, and scale go-to-market work with model labs and enterprises. The investment thesis is clear: attackers learn continuously, so AI defenses cannot behave like a compliance screenshot taken the week before launch.

The harder question arrives after the funding announcement. Proprietary data is valuable only if it keeps producing better tests, faster controls, and fewer dangerous surprises as models change and agents gain authority. Alice now has $140M more to prove that the worst behavior it studied online can help secure the software being trusted to act.

Generalist AI

A robot can watch a 3-to-12-second demonstration and still be a long way from surviving an 8-hour shift. Generalist AI is raising capital inside that gap.

An SEC Form D shows $198.2M sold to 32 investors in a planned $208.2M equity offering. Axios reports 8VC led the financing. TechCrunch reports it extends Generalist AI's Series B to $600M and values the company at $3B. Those are reported terms; Generalist AI and 8VC did not confirm them to the publications.

The speed is the story. Generalist AI announced $400M in June at a $2B valuation, led by Radical Ventures, and said that financing pushed total capital past $500M. In less than 3 months, the Series B and reported valuation have each grown 50% while the company moved from GEN-1 to GEN-1.5.

Pete Florence, CEO, Andy Zeng, Chief Scientist, and Andrew Barry, CTO, are building the intelligence rather than the robot. Their bet is that one embodied foundation model can learn across different machines, environments, and physical tasks instead of forcing every deployment to begin with new programming.

GEN-1.5 gives that thesis a useful number. Generalist AI says the model averaged 59% success across 10 simple, short-horizon tasks after one demonstration lasting 3 to 12 seconds, with no gradient updates. Give it 5 minutes of data per task and 10 gradient steps, and the reported average rises to 83%.

The company also supplied the sentence the market needs to keep: the tasks are simple and short-horizon, and the success rates remain modest. That caveat does not weaken the research. It separates a legitimate technical result from the usual robotics demo that reaches production already wearing a cape.

Generalist AI's June plan offers the clearest public account of where the money goes: next-generation models, a larger physical-data engine, more compute and training infrastructure, and work with industries that can put the systems into everyday use. The company is financing the loop between physical experience, better models, and commercial deployments.

That loop now carries a $3B reported expectation. Customers do not buy a benchmark, a valuation, or the possibility that robot learning has found its GPT-3 moment. They buy repeatability, safety, uptime, integration, and economics when the lighting changes, the object moves, or the shift runs long.

Generalist AI has made teaching a robot look dramatically cheaper in practice. The new capital makes the next question much more expensive: whether those seconds of instruction can survive months of work.

Outer Bio

AI can rank a compound before lunch. Human skin may need weeks to reveal whether the idea was useful.

That gap is the business Outer Biosciences has spent years building around. Its Yuna platform keeps living, full-thickness human skin viable and measurable for four weeks, then uses the data to improve the next machine-learning prediction.

TechCrunch reports that Outer Bio has raised roughly $23M to date. Wing Venture Capital announced a lead investment, while Outer Bio also names Initialized Capital, Sozo Ventures, Hawktail, Lightspeed, Third Kind, Liquid 2, and SV Angel as backers. The latest amount and series remain undisclosed. This is cumulative funding, not a clean new $23M round.

Michael Polansky, CEO, co-founded Outer Bio with Kyung-Jin Jang, CSO, Hinojosa, CTO, and Stanley King II, CBO. Their problem is less glamorous than the AI headline. Standard ex vivo skin remains usable for days, while collagen remodeling, chronic inflammation, and cellular senescence can take weeks.

Yuna extends that window. Outer Bio says the system has generated more than 10TB of proprietary data across 300-plus donors and 10,000 treatments. Each compound-sample pair produces more than 30,000 data points; its virtual library spans 5.9M compounds across 11 databases.

The loop is physical. A model prioritizes a compound. Living tissue tests it. Longitudinal measurements record what happened, and the answer returns to the model. AI supplies a hypothesis; biology keeps the score.

Outer Bio's preprint describes three four-week tests: psoriasis-like inflammation reduced by a known JAK inhibitor, senescence-related signatures attenuated in aged tissue, and UVB injury mitigated by sunscreen. Yuna still cannot recruit circulating immune cells and lacks full vascular perfusion.

That honesty is part of the commercial test. Beauty and pharma buyers need evidence that survives formulation, safety work, manufacturing, and a customer's skin.

Outer Bio earns revenue from research collaborations and plans to license or sell ingredients rather than build a consumer brand. Its roadmap points toward more throughput, deeper sensors, broader biology, and product-development capability. None is a disclosed use-of-proceeds schedule.

The $23M gets attention. The real asset is the experimental clock: four weeks of living tissue, repeated across donors, teaching software where it was wrong. Turn that loop into ingredients partners can commercialize, and Outer Bio will have built something AI cannot scrape from the internet.

Stability AI

A record label signing a venture check is also negotiating the future of its catalog.

Stability AI's $76M Series B brings Electronic Arts, Sony Music Group, Universal Music Group, Warner Music Group, AMD Ventures, and Pacific Alliance Ventures onto the cap table. Coatue, Greycroft, Kadmos Capital, Sean Parker, and Eric Schmidt returned.

The roster matters because the money and market are coming from the same room. EA is co-developing tools for game artists and developers. Universal and Warner are shaping music products around licensed data, artist control, and commercially safe workflows. Sony arrives as a new investor.

These investors own catalogs, employ creators, ship games, run production pipelines, and negotiate rights. They know where a clever demo becomes an expensive operational problem.

The round funds Stability AI's creative-production suite, applied research, and professional services. It also gives the company a harder standard to meet. Strategic investors need the model to fit a workflow, respect rights, survive legal review, and improve production without asking the artist to surrender control.

Prem Akkaraju has been rebuilding around that commercial reality since becoming CEO in June 2024. Stability AI says it has raised $232M under current leadership across 2 equity rounds and convertible notes. That is not a lifetime total. No new valuation or lead investor was disclosed.

Emad Mostaque and Cyrus Hodes co-founded Stability AI, and Stable Diffusion helped make open generative imaging a mass-market behavior. Mostaque left the CEO role and board in March 2024. The next chapter has less room for mythology and more obligation to build a durable business.

Stable Audio 3.0 shows the direction. Stability AI says the models were trained on fully licensed data, with open-weight versions and an enterprise path. The product argument is now control, customization, licensing, indemnification, and a tool that can live inside somebody else's production process.

That distinction matters in music. A buyer is judging more than output quality. It is asking whose data trained the model, what rights travel with the result, and who carries the risk when the work leaves the studio.

That is why this financing is more interesting than the amount. These investors can help define the product, open distribution, pressure-test the rights model, and become reference customers. They can also expose weak promises faster than a conventional venture syndicate.

Capital bought Stability AI time. Its strategic partners just made the next test much less theoretical.

OmicsBank

A radiology image, a pathology slide, a lab result, and a clinical note can belong to the same patient without becoming one usable record. The patient is whole. The data estate is not.

OmicsBank is building inside that separation. The San Francisco company has raised $2.25M in Seed funding from Redesign Health to expand clinical data infrastructure across the U.S., the Middle East, and Southeast Asia.

Co-founders Sumit Sinha and Vijay Goel started OmicsBank in 2025. Their software sits inside hospital systems, standardizes and de-identifies fragmented records, links clinical events with imaging and multi-omics data, and keeps governance attached to the work.

The product is not a database somebody can assemble by scraping the public internet. OmicsBank says its network reaches more than 90 hospitals and diagnostic labs, with data from more than 12M patients, 30M+ imaging studies, 6M+ pathology slides, and hundreds of thousands of biospecimens and sequencing files. Those are company-reported figures, not audited outcomes.

The scale matters only if records preserve enough context around a clinical decision. Volume without linkage gives a model more material, not more understanding.

South Asia represents roughly 25% of the world's population but less than 2% of participants in genome-wide association studies, according to peer-reviewed research. Models trained on narrow populations can become technically impressive while remaining clinically incomplete.

OmicsBank is trying to turn that representation gap into infrastructure. Pharma and biotech teams can use linked records for drug development and real-world evidence. Frontier AI teams can use multimodal data for training, evaluation, and reinforcement-learning environments. Hospitals get a way to query archives that were generated locally but rarely organized as one longitudinal patient record.

Redesign Health is financing the institutional work required to earn access one hospital at a time. Orbit Ventures led OmicsBank's 2025 pre-seed. The new capital will support U.S. health-system partnerships, network growth across Asia and the Middle East, and GPU-enabled research environments.

OmicsBank has not disclosed valuation, revenue, named customers, or independently validated outcomes. Expansion will test whether data quality, consent, provenance, and governance remain intact as more institutions and jurisdictions enter the network.

Healthcare AI keeps getting judged by what a model can produce. OmicsBank is betting that the harder advantage begins earlier, with what a hospital can safely organize, permit, and trust.

Zencore

John Forstrom and Guilhem Tesseyre built Zencore around a staffing choice that services firms usually make harder as they scale: keep experienced architects and engineers close to the client.

Superstep Capital is now financing the next test of that model. The digital-services private equity firm made a strategic investment in Zencore on August 25. The amount, ownership stake, valuation, and transaction structure were not disclosed, so this is a strategic investment, not a mystery round that needs an imaginary number attached.

Zencore is a Google Cloud Premier partner founded in 2021 by former Google Cloud leaders. The company reports hundreds of engagements across North America and EMEA, 9 specializations, and 4 Partner of the Year awards. Zencore says it remains independent.

The money arrives with an assignment. Zencore plans to expand leadership, delivery, go-to-market, alliances, and forward-deployed engineering while launching 3 ZenAI Factory offerings. The software, data, and operations factories apply agents to development, reporting, infrastructure monitoring, diagnosis, and incident remediation.

Enterprise buyers have moved beyond asking whether an agent can produce a clever demo. They need to know whether it can touch source code, production data, and infrastructure without creating a fresh operating problem for every task it removes.

Stuart Coleman and Superstep have spent their careers around digital-services economics. Their thesis favors specialized, founder-led firms and brings operating help across talent, alliances, go-to-market, and delivery. Those are the pressure points that appear when expertise has to become a repeatable business.

John Forstrom described the plan as the same team and focus with more capacity. The tension sits inside “more.” A consultancy can hire faster than it can transfer judgment, and package delivery faster than it can prove that a framework survives customer permissions, legacy systems, and production consequences.

The ZenAI Factory gives John Forstrom and Guilhem Tesseyre a clearer commercial unit than open-ended AI consulting. It also creates a harder promise: improve releases, produce trustworthy reporting, and resolve incidents without becoming one.

Superstep is backing focus in a market that rewards scale. Zencore now has to show that those forces can share the same delivery model. The next phase will be measured in the handoff between architecture and operations, where the senior people who won the room still need to stay close enough to the systems when the room gets bigger.

Concordare

One protocol amendment can be a small change on paper and a fresh round of rebuilding everywhere else.

A clinical research team may translate that update into source documents, EDC fields, CTMS schedules, billing grids, and site instructions. Too much of the operation still tests whether every system interpreted the same PDF the same way.

Concordare Trials has disclosed an undisclosed funding round led by Surface Ventures, with Techstars, Innovation Works, Gaingels, and individual investors participating. The company published no amount, valuation, or formal round label, so the useful story sits in what the capital is being asked to change.

Founder and CEO Zach Sawaged spent roughly a decade across research sites, sponsors, and technology vendors before building around the protocol itself. Most clinical software enters after people have already rebuilt the study logic. Concordare wants to move upstream.

Apta extracts a protocol PDF into the Clinical Trial Protocol Model, or CTPM. The model can feed source documents, EDC builds, CTMS configurations, and billing grids. The study team validates it before anything downstream is generated.

That human gate matters. Concordare published a third-party structured quality assessment on one protocol covering 3,171 elements: 99.62% precision, 97.47% recall, and 98.53% F1. It also disclosed 12 incorrect elements and 82 missing ones, a more useful picture than the percentage alone.

Concordare says Rutgers University uses the platform across multiple teams and claims the workflow can save 6-8 weeks of site activation. Both remain company-reported, but they frame the question Gyan Kapur is underwriting.

If the protocol becomes structured once, sites and sponsors may stop rebuilding the same requirements inside every system. Amendments could propagate through a shared model instead of restarting the translation exercise, work that decides whether a trial opens on time.

The funding supports product, team, and site-partnership growth. Concordare also announced an undisclosed pre-seed in February; the August materials do not say whether this is separate, an extension, or fuller detail on that financing. Precision in the capital record belongs beside precision in the protocol.

Concordare now has to make one digital representation survive more institutions, protocol types, security reviews, and amendments without turning shared infrastructure into a shared mistake. The protocol already tells the trial what to do. Concordare is trying to make every downstream handoff agree on what it heard.

Visa / BioCatch

A payment can carry the right password, come from a familiar device, pass authentication, and still be exactly what the criminal wanted.

That is the problem Visa is paying $2.4B in cash to move closer to. The company signed a definitive agreement to acquire BioCatch from funds advised by Permira and other shareholders, subject to regulatory approvals. Closing is expected by the end of Visa's fiscal Q2 2027.

BioCatch studies what happens before money moves. Its models analyze thousands of behavioral, device, network, application, and transactional signals, including how someone types, swipes, handles a device, hesitates, or behaves under pressure. It is recognizing when a legitimate customer has been manipulated, when a trusted device is compromised, or when a mule account is preparing to receive funds.

Visa says BioCatch protects 1.8B devices and 760M users, serves more than 350 banking clients across 21 countries, and works with more than 100 of the world's largest banks. Permira says the company was analyzing more than 19B digital sessions each month by July 2026, while revenue and gross profit grew roughly 3x during its ownership.

That scale matters because Visa already sits inside a different part of the fraud stack. Its network sees transactions. Featurespace, acquired in 2024, strengthens real-time payment monitoring and risk scoring. BioCatch adds a view into behavior, device trust, account opening, scams, and money-mule activity before the authorization becomes the crime scene.

Andrew Torre, Visa's President of Value-Added Services, framed the deal around stopping fraud before it reaches payment. That fits a business that generated nearly $11B in Value-Added Services revenue in fiscal 2025 and works with roughly 14,500 financial institutions. Visa is buying another place in the decision chain where its clients can say no.

Gadi Mazor says BioCatch will keep its leadership team and reporting structure and join Visa's Value-Added Services organization. Bank clients will care about that continuity, but they will also watch whether BioCatch remains as focused and network-agnostic once it belongs to one of the world's largest payment networks.

Fraud teams have spent years trying to separate a good transaction from a bad one. BioCatch built its value around a harder distinction: a real person can perform the action while a criminal controls the outcome. Visa wants that distinction closer to the network, where behavior, device intelligence, and payment data can meet before the money disappears.

Onos Health

Aetna is using Onos Health's behavioral-health AI platform. CVS Health Ventures, the venture arm of Aetna's parent, has now invested. Customer evidence and strategic capital are meeting inside the same healthcare system.

Onos Health raised a $17M Series A led by Costanoa Ventures, with Flare Capital Partners and CVS Health Ventures participating. It follows a $6.3M venture round in October 2025. No valuation was announced.

Behavioral health can look legible in a claims file while care quality remains buried in clinical notes. A claim shows what was billed. It cannot explain whether treatment followed clinical guidelines, whether a member is progressing, or whether a costly care pathway is the right one.

That gap pushes health plans toward retrospective review after cost and quality signals have had time to compound.

Akshay Agrawal, CEO, Josh Levitan, CPO, and Suhaas Prasad, CTO, built Onos to combine claims, utilization data, clinical documentation, and quality guidelines. The platform surfaces patterns and treatment gaps while keeping decisions with clinicians. Onos says its outputs link back to the source record.

The company reports that more than 70% of behavioral-health quality signals sit inside unstructured documentation. It also reports 35% better clinical-standard adherence, 75% better clinical-review efficiency, and more than a 6% reduction in program costs within 12 months. Onos says 3 of the 6 largest U.S. health plans use the platform.

Those are company-reported results, not independently audited outcomes. They are still specific enough to define what this round has to prove.

Aetna's role sharpens the commercial signal. A named health-plan customer is using the product while the venture arm of its parent invests in the vendor. Repeatability across other plans, provider networks, diagnoses, and states remains open.

The difficult part is proving the system can reduce review without becoming an opaque utilization barrier with better branding. Behavioral health already carries too much friction between members, providers, and payers.

Clinical intelligence has to surface evidence earlier, make the reasoning inspectable, and leave judgment with the people accountable for care.

The $17M gives Onos more room to accelerate adoption. It also gives the company a sharper obligation: reproduce the reported outcomes across more health-plan environments without confusing automation with authority.

The next chapter sits inside plan-provider conversations, where shared evidence must change the care path without becoming another layer everyone works around.