Chasing the Unicorn

Cylake

Cylake is asking enterprise security buyers to put the cloud era in reverse, then finance the hardware, software, data, and AI required to make the turn.

That sounds backward until the customer is a bank, agency, defense program, hospital, or infrastructure operator whose telemetry cannot leave its control. Cloud-first brought faster updates while leaving some of the institutions with the most to protect choosing between modern capability and sovereignty.

Cylake has raised $245M through a convertible note before its product reaches beta. Lightspeed Venture Partners led the financing, with Picture Capital and Redpoint Ventures participating. Total disclosed funding is now $290M after a $45M Seed led by Greylock Partners.

The founding team knows what sits inside that promise. Founder and CEO Nir Zuk founded Palo Alto Networks. Co-founder and Chief Development Officer Wilson Xu led engineering there. Co-founder and Chief Architect Udi Shamir helped start SentinelOne. CTO Jesse Ralston completes a leadership team rebuilding the stack as one system.

Cylake's appliance combines compute, storage, a shared data lake, security functions, policy, detection, automation, and AI inside the customer's environment. The company says machine learning will watch the unified stream while language models help investigate, explain, and respond. The customer keeps the infrastructure and data under its own control.

That turns a subscription decision into a systems-engineering job. Cylake has to ship hardware, support air-gapped environments, handle endpoint telemetry, and prove that a complete platform can replace fragmentation without becoming one very large new dependency.

Lightspeed partner and co-founder Ravi Mhatre is joining the board. Rakesh Loonkar of Picture Capital is joining as an observer. Cylake says it has more than 40 employees and unnamed design partners, with beta expected by the end of 2026 and general availability in 2027.

The note's valuation and conversion economics were not disclosed. Neither were customer names, revenue, deployments, or independent benchmarks. The capital can build the machine; regulated institutions will judge it inside protected rooms where reliability and control carry more weight than a launch slide.

Cylake now has the money to bring the cloud's development speed back behind the customer's walls. The next part unfolds between an ambitious architecture and security teams that have spent years refusing shortcuts for very good reasons.

VideoGen

Generative video keeps winning the demo at the first clip. The customer still needs the campaign, the translation, the training library, the next edit, and the second version that should remember what the first one already learned.

VideoGen is raising $3.3M for that longer job. The Seed round includes Y Combinator, Rebel Fund, Lobster Capital, and Stretford End Capital, with Mento VC, Pioneer Fund, and Decacorn VC participating. No lead investor or valuation was disclosed.

Anton Koenig and David Grossman started the company in 2023 with $30K saved from internships. They bootstrapped for the first year, joined Y Combinator's S24 batch, and built an AI video platform that now reports more than 5M users across more than 190 countries.

That reach deserves a clean qualifier. VideoGen says it serves thousands of customers, from individuals and small businesses to Fortune 500 teams, but the company has not disclosed paid conversion, retention, revenue, or enterprise contract value. Five million users is a distribution signal. The business hiding inside that number still has to become visible.

VideoGen's product thesis starts where the clip usually stops. Its agents take a brief or uploaded media through structured workflows into a fully editable video. Shared context and reusable assets help teams make related videos without rebuilding every decision, while 200+ voices, 50+ languages and accents, one-click translation, browser editing, a Text-to-Video API, and MCP integration extend the work across teams and software.

That distinction matters as generation gets easier to access. Marketing teams do not buy a model demo; they need campaign variants that survive review. Learning teams need translated material that can change when the policy does. Developers need an interface that can produce again without turning every API call into a fresh creative negotiation.

VideoGen says the capital will scale the team and grow the brand. That is a practical mandate for a small company with a global top of funnel. The next phase is less about proving that people will try AI video and more about earning repeat work when the first output needs a revision, a new market, another channel, or a teammate who was not in the room when the brief began.

The models will keep improving. VideoGen's wager is that the durable customer relationship forms around everything required to turn their output into work a company can edit, reuse, govern, and ship.

Celero Communications

Celero Communications closed a $100M Series B 295 days ago. In semiconductor time, that is barely enough to finish the meeting notes.

The company says it used that window to tape out and validate a 2nm coherent DSP with advanced analog technology. Now Celero has raised a $275M Series C at a valuation above $3B, taking total capital to $415M.

Atreides Management, Valor Equity Partners, and CapitalG co-led the round. Sutter Hill Ventures and Maverick Silicon returned. Atreides Managing Partner and CIO Gavin Baker will join the board.

Co-founders Nariman Yousefi, CEO and President, and Oscar Agazzi, CTO, built Celero around a problem the AI infrastructure market keeps making more expensive. Accelerators can only behave like one computing system when data moves between them with enough bandwidth, reach, power efficiency, and fiber efficiency.

PAM4 has carried much of the short-reach data-center load. Traditional coherent technology travels farther, but its telecom power and cost structure does not automatically fit dense AI clusters. Celero is targeting scale-up, scale-out, and scale-across networks, from 1.6T links toward next-generation 3.2T interconnects.

That is the architecture. The Series C is about surviving everything that comes after it.

Validated silicon matters because a semiconductor cannot ship as a simulation. It still has to become a qualified product with stable performance, acceptable yield, ecosystem support, customer design wins, and a production schedule that hyperscalers can trust.

Celero says it validated what it calls the industry's first 2nm coherent DSP with advanced analog technology. The wording deserves precision: Marvell announced 2nm coherent DSP products in March. Celero's milestone is its claimed validation of a specific AI-optimized architecture, not proof that competitors missed the node.

Celero has not disclosed customers, revenue, yield, foundry, independent benchmarks, or a volume-production date. It is the work this capital now has to expose.

Yousefi and Agazzi spent careers across Broadcom, ClariPhy, Inphi, and Marvell building earlier generations of networking silicon. Their advantage is knowing where elegant designs meet ugly production realities.

Their new investors are giving them enough money to meet those realities quickly. The next Celero milestone will matter when 2nm performance leaves the validation bench, enters a customer's qualification process, and starts carrying the traffic that made the round possible.

ClaimQI

A suspicious claim is easy to flag. A payer still has to explain which rule it broke.

That second sentence is where healthcare payment integrity gets expensive. A score can start an investigation, but a denial or recovery decision has to survive policy language, coding logic, provider questions, internal review, and an auditor who would like more than “the model felt weird about it.”

Chad J Wells and Janna Hart built ClaimQI, Inc. around that gap. ClaimQI says its LLM translates coverage policies and codes into deterministic, executable rules, while claims data and hard determinations stay outside the language model. The output is designed to show the rule, explanation, and source behind the decision.

That architecture is a commercial choice as much as a technical one. Health plans do not buy ambiguity at enterprise scale. They buy a decision they can defend across Medicare Advantage, Medicaid, Medigap, and self-funded ERISA claims without turning every case into an archaeological dig through policy documents.

ClaimQI closed a Seed round of just under $500K on September 8, 2026. CQuence Health invested alongside four unnamed private investors and will provide strategic support. A separate $150K Nebraska Department of Economic Development Innovation and Prototype grant remains pending final approval, so it does not belong in the closed financing total.

The capital will support the ClaimQI Platform and Payer Portal, with a formal launch for Medicare payers at the Healthcare Payment & Revenue Integrity Congress in Chicago. ClaimQI plans to expand across payer markets where coverage rules, billing codes, and financial consequences collide every day.

CMS estimated $23.67B in gross Medicare Part C improper payments for FY2025, a 6.09% error rate based on payment year 2023. CMS also makes an important distinction: improper payments are not a fraud estimate. They can include overpayments, underpayments, missing documentation, and technical failures.

ClaimQI has not disclosed revenue, customers, valuation, investor allocations, or independently audited savings and accuracy. SOC 2 certification is described as in progress. Those blanks matter because this product's explanation becomes part of the buyer's accountability. The Seed gives ClaimQI room to put its design choice in front of real claims teams: let AI translate the complexity, keep the decision anchored to explicit rules, and make every recovered or prevented dollar carry its paperwork with it.

Lightsage

The old sales funnel has a blind spot: a customer that never clicks.

An AI coding agent can discover an API, compare tools, read the docs, pick an SDK, attempt authentication, and walk away without producing the lead, demo request, or attribution trail a growth team expects.

Jun Liang LEE, CEO, and Sean Er, CTO, built Lightsage around that missing journey. The San Francisco company raised $4M in Seed funding led by Nexus Venture Partners, with backing from operators across Salesforce, Postman, Apollo, DocuSign, and the developer ecosystem.

The company's argument is bigger than better visibility inside answer engines. An agent still has to understand the product, navigate documentation, choose the right tooling, authenticate, use the API, and finish the job. Every failure along that path can erase a customer before conventional analytics knows one existed.

Lightsage calls the category Agent-Led Growth. Its platform runs simulations across answer engines and coding agents, gives them integration tasks, and observes where the journey breaks. The cause might be discovery, docs, authentication, an endpoint, an SDK, or agent tooling that fails across products.

That makes this a growth story with an ownership problem hiding inside it. Marketing can improve discovery. Developer relations can sharpen examples. Product can repair onboarding. Engineering can fix authentication and endpoints. The agent crosses all 4 functions in minutes, while the org chart still hands the failure from one queue to another.

Lightsage names Firecrawl, Reducto, Daytona, Rime, and Tinyfish as early customers. Forbes reported that some saw a 10%-20% increase in sales to AI-agent customers, according to Jun Liang Lee. With no public sample, period, or methodology, the figure remains an early company-reported signal.

The $4M will deepen evaluation, analytics, attribution, and optimization across APIs, SDKs, CLIs, MCP servers, and Agent Skills. Lightsage also plans to grow its technical and commercial teams, then move from developer tools into B2B software, infrastructure, and payments.

Abhishek Sharma and Nexus are backing a change in the unit of software growth. The old unit was a human visit that became a signup. Lightsage wants to measure a delegated task that becomes successful product use.

That leaves Jun Liang Lee and Sean Er with a peculiar assignment: teach software companies to recognize a customer they cannot see yet, then prove that fewer agents disappear between a recommendation and a working product.

Arctos / Atlanta Falcons

The NFL waited decades before admitting private equity, then wrote a rule that keeps the investor silent.

That is the tension inside Arctos Partners' reported agreement to buy 10% of the Atlanta Falcons at a $10.6B enterprise valuation.

The number is loud. The ownership position is designed to be quiet.

CNBC reported that the stake would be acquired in two tranches over 18 months. The transaction still needs NFL approval, with a vote expected in October, so this is an agreement to buy, not a completed acquisition.

The $10.6B figure is also a valuation, not a disclosed purchase price. A clean 10% multiplication gets you to $1.06B, but the actual consideration, seller mix, adjustments, and tranche terms remain private. Arthur Blank owns just under 73%, and reporting has not established whether the full interest comes from him, existing limited partners, or both.

That uncertainty does not weaken the market signal. It explains the structure.

The NFL's private equity policy allows approved funds to own a combined maximum of 10% of one team. Each position must be at least 3%. A fund can invest in up to 6 franchises, must initially hold the stake for 6 years, and receives no voting power.

The league created liquidity without handing over the controls.

For Falcons ownership, the deal can turn a fraction of an appreciating asset into cash while Arthur Blank keeps the franchise. For Arctos, now part of KKR Solutions, Atlanta would become a 4th NFL position after the Chargers, Bills, and Browns.

For the rest of the league, the most important output may be the price. A passive 10% interest can still establish what sophisticated capital is willing to pay around a franchise, even when no controlling stake changes hands.

That distinction matters. Minority valuations do not carry the control premium of a full sale, and fans should not expect a private equity fund to start making football decisions. The NFL built the policy to prevent exactly that.

What it does create is a growing market at the edge of ownership. Families and controlling owners gain another source of liquidity. Funds gain access to scarce sports assets. Every approved transaction adds another data point for the next conversation.

The October vote will determine whether Arctos joins the Falcons' ownership table. The harder question will keep moving after approval: how many owners decide that selling a silent slice is the cleanest way to let the market speak about everything they still control?

Bluecore Energy

A port spends years electrifying cranes, trucks, warehouses, and ships, then discovers the clean-power problem arrived at the waterline.

Bluecore Energy HQ is building for that shoreline. Founder and CEO Kofi Asante has raised an oversubscribed $50M Seed led by Silverton Partners for compact nuclear reactors designed for floating barges. Axios and The Register say the round includes Bluecore's $10M Pre-Seed, making total disclosed capital $50M.

Nine months from inception, the company has secured a terminal, delivered a barge, and begun an electrically heated prototype at the Port of Long Beach. It holds no nuclear fuel or radioactive material. The team is testing sensors, controls, monitoring, and water flow before a licensed machine exists.

Bluecore's initial design targets about 10 MWe. The company says one unit could match the electricity use of roughly 15,000 homes, run for years between refueling, and pair with other units for port-scale demand. An eventual system could sit offshore and send power back by subsea cable.

That mobility changes the map without making nuclear easy. The reactor must survive engineering review, maritime classification, fuel procurement, manufacturing, federal licensing, local approval, and public scrutiny. The barge adds vessel safety, jurisdiction, offshore operations, emergency planning, and grid-connection questions.

The $50M will fund those unglamorous proofs: hardware, testing, validation, regulatory work, manufacturing preparation, hiring, and fuel procurement. Bluecore says formal engagement with the Nuclear Regulatory Commission and U.S. Coast Guard has begun. The agencies have a coordination framework for civilian maritime reactors, while MARAD and the Port of Long Beach are separately exploring SMRs, microgrids, shore power, and new propulsion systems.

Silverton joins returning backers including Slauson & Co., Harlem Capital, Precursor Ventures, and Chris Larsen, alongside new institutional and individual investors. The syndicate is financing speed, but nuclear speed is measured by how quickly one institution can trust the evidence produced for the next.

Bluecore has no operating reactor, commercial license, named customer, or public deployment date. It does have a rare seed-stage test: turn a visible barge, a non-fueled prototype, and early regulator engagement into decisions strong enough to carry power from the water into one of the world's busiest port systems.

Solstice Oncology

The most important part of Solstice Oncology's $225M launch is not the size of the round.

It is the timing of the treatment.

The Boston biotech is building around a clinical question that immuno-oncology has rarely tested at scale: can a potent immune therapy do more before surgery, while the tumor is still present and the immune system has a target to recognize?

Solstice launched with a $225M Series A led by RA Capital Management, with Canaan Partners, Forbion, and other investors participating.

The capital will advance porustobart, a next-generation CTLA-4 antibody licensed from Harbour BioMed, into Phase 2 development.

Its first study will pair porustobart with pembrolizumab in stage II or III mismatch repair-proficient or microsatellite-stable colon cancer. These tumors have historically responded poorly to checkpoint inhibitors.

The 100-patient US and European trial was not yet recruiting on September 9. Solstice expects data in the second half of 2027.

There is an early clinical signal behind the program, but it needs careful framing.

Harbour BioMed reported a 30% response rate, or seven responses among 23 patients, in a Phase 1 study of porustobart plus a PD-1 antibody. Those patients had late-line microsatellite-stable metastatic colorectal cancer without liver metastases.

That is encouraging in a difficult setting. It is also a small, selected cohort, not proof that the same result will translate to treatment before surgery.

Porustobart is designed to combine CTLA-4 checkpoint blockade with stronger depletion of regulatory T cells inside tumors. Solstice also points to a shorter half-life than first-generation CTLA-4 medicines, which could make exposure more manageable. Both ideas now need prospective clinical evidence.

The financing gives the company room to generate it.

Founded in February 2026, Solstice is led by CEO Caroline J. Loew, PhD. David Feltquate, MD, PhD, is CMO, and Maiken Keson-Brookes, LLB, LLM, is COO and general counsel.

A March Form D showed about $121.3M sold in the offering. That appears to be an earlier Series A close, not additional funding.

The next readout has to answer a focused question: can porustobart's design and the neoadjuvant setting produce enough benefit to justify using a powerful immune therapy in patients whose tumors are still operable?

That is what this unusually large first round is buying: a clinical answer, not just a longer runway.

Harmoni

A machine built during World War II can still make a good part. Asking it to brief an AI about the live job, part, operator, revision, and last bad run is where the future starts charging integration fees.

Harmoni is building for that gap. Its tablets use RFID to recognize workers and jobs, connect ERP and machine data, and deliver instructions, programs, checksheets, and support requests.

Co-founders Adam Ellis, CEO, and David Caputo have raised a $10M Series A led by Bessemer Venture Partners. The company also introduced HAL, an AI layer designed to inherit Harmoni's operational context instead of asking a worker to reconstruct it for every prompt.

That context is the product before AI becomes the headline. The machine knows its cycle, the ERP knows the order, engineering owns the revision, and quality has the checksheet. The operator still notices when those records stop describing the same job.

Harmoni says HAL can use the active part, machine, process, history, and engineering requirements to help workers understand instructions, diagnose issues, identify risks, and route action. A wrong answer here becomes scrap, downtime, rework, or a late delivery.

Business Insider reported roughly 40 customers, with 10 more expected in September, and said aerospace and defense is Harmoni's largest segment. WessDel has used the system since 2024 after workers spent close to an hour each day at three reporting kiosks.

Harmoni says its automations return more than 200 productive hours per operator each year. The company-reported figure is not an independent benchmark, but it gives the Series A a useful ambition: recover skilled time without treating the people or machines on the floor as obsolete.

Deloitte and The Manufacturing Institute estimate U.S. manufacturing could need 3.8M additional workers by 2033, with 1.9M positions potentially unfilled. More automation is the obvious answer. Making it cooperate with the installed reality is the harder work.

Kent Bennett and Bessemer are backing an operational layer for that reality. The $10M will support product development, HAL, go-to-market expansion, and team growth. Earlier financing is documented, but its amounts, the valuation, and lifetime capital remain undisclosed.

Harmoni now has to carry live context across factories where equipment, processes, knowledge, and tolerance for error vary by the shift. Each deployment should leave the next operator with less administrative work and a clearer answer before metal, time, or trust gets cut away.

MCatalysis

The cleanest fuel on the spreadsheet still loses when the purchasing team has to explain why it costs more. Climate intent can open the meeting. Unit economics decides whether the order survives it.

MCatalysis has closed a $5M Seed round to attack that problem from inside the machinery already making and moving fuel. HL Energy Ventures and Oxford Science Enterprises co-led the financing, giving the company capital to build an industrial pilot at Axel'One in Lyon, France.

Founder and CEO Michael Irwin and CTO XIANGYU Michael JIE are working from an Oxford-developed microwave-catalysis foundation. Their process is designed to convert industrial offgas, plastics, and biomass into drop-in fuels that can move through existing infrastructure.

That word, drop-in, is doing commercial work. A new fuel brings equipment decisions, qualification, operating risk, distribution changes, and a buyer who must defend the premium if the economics drift. MCatalysis wants to keep the molecule familiar while changing how it gets made.

The company says conventional refining and early synthetic alternatives rely on extreme temperatures, heavy energy inputs, and multi-stage facilities. Its microwave-driven catalysts are designed to reduce that burden. Oxford researchers, including Michael Jie, have published peer-reviewed work in the field. Commodity-cost fuel remains a company target, not industrial proof.

Lyon is where that thesis meets plant conditions. The $5M funds an industrial pilot at Axel'One inside the Vallée de la Chimie, while R&D stays in London. The handoff runs from catalyst performance to throughput, uptime, product quality, feedstock variation, and a cost curve a customer can use.

MCatalysis separately cites a recent $2.25M UK Research and Innovation award. That grant supports the scale-up story, but it is not part of the $5M venture round. The company also announced an undisclosed Seed financing led by HLEV in 2025 and has not said whether that earlier capital sits inside this close.

HLEV has backed MCatalysis from the beginning. OSE adds an investor built around turning Oxford science into companies. They are betting that compatible fuel and modular refining can remove enough friction for economics to carry the transition.

The announcement names no customers, offtake agreements, or verified production cost. Those absences define the work moving into Lyon, where a catalyst has to become a plant and the green premium has to disappear from the buyer's explanation.

Implicity

Arnaud Rosier's first product requirement came from a physician's morning: five device portals, five passwords, and five displays.

Remote cardiac monitoring had solved the distance problem. Pacemakers, defibrillators, and loop recorders could send data without another clinic visit. The bill arrived as more transmissions, more alerts, more systems, and the responsibility to decide which signal needed attention.

That frustration became Implicity. Arnaud Rosier, an electrophysiologist, remains CEO and co-founder. David Perlmutter is co-founder, Chief Strategy Officer, and Head of R&D. Louay Gargoum is CTO.

Now Implicity has raised $40M in growth equity led by IRIS, with Five Arrows investing alongside it. IRIS reports the same round as €35M, the European currency presentation of one financing.

The money will expand Implicity's U.S. commercial team, support product work for American health systems, strengthen its European position, and fund predictive and diagnostic AI, including next-generation heart-failure algorithms.

Remote monitoring is standard care for patients with cardiac implantable electronic devices, yet the guidance behind that standard reads like an operations manual: enrollment, connectivity, triage, documentation, alert management, and EHR coordination. Visibility creates value only when a clinic can turn it into a decision.

Implicity aggregates and normalizes device data across manufacturers, integrates with EHR workflows, and helps teams separate routine noise from events that deserve action. The company reports 250+ medical centers and 120,000+ patients monitored daily across the U.S., France, and Germany.

SignalHF received FDA clearance in 2023, and Implicity's next-generation ILR ECG Analyzer followed in July 2026. A French observational study of 69,394 ICD patients also associated a universal monitoring system with lower adjusted mortality and hospitalization hazards than manufacturer-specific monitoring.

The study was observational, and its disclosures identify Arnaud Rosier as Implicity's CEO and a major shareholder. Health-system trust depends on evidence that carries its limits in plain sight.

IRIS and Five Arrows are funding a harder commercial job than another dashboard. Implicity has to become the layer a hospital trusts between fragmented device feeds and clinical attention, then keep earning that position as patient count, alert volume, and regulatory responsibility move together.

Encoded Therapeutics

In February 2025, Encoded Therapeutics cut 29% of its staff to preserve runway and put ETX101 at the center of the company. That decision came before the strongest clinical signal, pivotal dosing, and the $275M Series F now financing the next stretch.

The South San Francisco biotech closed with GV and an undisclosed healthcare fund as co-leads. Participants included ARCH, Braidwell, Farallon, Illumina Ventures, Invus, Janus Henderson, Matrix, Nolan, RTW, SoftBank Vision Fund 2, and Venrock.

Co-founders Kartik Ramamoorthi, Ph.D., CEO, and Stephanie Tagliatela, CSO, built Encoded Therapeutics around a gene-therapy problem. Dravet syndrome is commonly caused by SCN1A loss-of-function variants, but the gene is too large for a standard AAV payload. ETX101 uses an AAV9 vector to deliver a regulatory element designed to increase the patient's own SCN1A expression in inhibitory neurons.

Encoded reported updated open-label Phase 1/2 POLARIS data from 21 participants at the European Epilepsy Congress. At Month 12, the small groups with available follow-up showed median seizure reductions of roughly 79% at dose level 3 and 89% at dose level 4. The company also reported developmental gains and no treatment- or procedure-related serious adverse events through the August 3 cutoff.

Those interim, company-reported results come from small, nonrandomized cohorts. They support the move into pivotal development without settling efficacy, durability, safety, or commercial value.

The $275M funds that handoff. Encoded will support the pivotal ENDEAVOR Part 2 study in infants and young children, expand evaluation through age 18, scale internal GMP manufacturing, and move ETX301 toward a 2027 IND submission for post-amputation neuroma pain.

Manufacturing belongs inside the investment story. A one-time genetic medicine needs reproducible product, regulatory-ready controls, clinical execution, and capacity that can survive the jump from a few study participants to a commercial process.

Encoded disclosed a $104M Series C in 2019 and a $135M Series D in 2020. Those rounds and this Series F total $514M. Other financing history is incomplete, so no exact lifetime figure is defensible.

The round keeps 2 clocks moving: clinical evidence on seizure and developmental outcomes, and manufacturing evidence that a one-time therapy can be made consistently beyond the trial site. Families living with Dravet syndrome have already supplied the patience. Encoded and its investors have financed the proof they still owe them.

Beep

A driverless shuttle leaves the curb carrying more than passengers. It carries a route promise, an agency contract, a wheelchair ramp, command-center oversight, and a rider who does not care which vendor owns the algorithm.

That operating burden sits behind Beep, Inc.'s newly closed Series B. Mobileye led the round, existing investors participated, and the financing brings Beep's total reported capital to approximately $130M.

The amount raised in this closing was not disclosed. The announcement does not say $130M was raised in this round, and a June SEC filing showing $7.1M sold from a $20M offering does not establish the final September close either.

Beep lives inside the distance between a vehicle that can drive itself and a transit service people can depend on. AutonomOS coordinates vehicles, infrastructure, routes, demand, supervision, and real-time service management while Beep handles the planning, deployment, and operations around the network.

Mobileye has spent decades teaching vehicles to interpret the road. As it expands toward robotaxi operations, leading Beep's round puts capital behind the operator that must turn autonomous driving into schedules, rider support, safety processes, and public accountability.

Kevin Reid, Beep's cofounder, CEO, and Chairman, now leads a company with operating evidence beyond the pilot deck. Beep began operating Jacksonville Transportation Authority's NAVI service under a five-year contract, with 14 electric Ford E-Transit vehicles using Oxa's driving system. ATL Spoke entered service as Atlanta's first autonomous-transit pilot, and Beep is working with MOIA America on a planned U.S. rollout of autonomous ID. Buzz vehicles.

Those deployments use different systems and partners. The durable asset may be the operating layer that can absorb those differences without asking a rider or transit agency to become the integrator.

The new capital is meant to scale Beep's mobility-as-a-service offering, deepen public-sector and technology partnerships, and extend AutonomOS. It cannot skip procurement, validation, accessibility, maintenance, community trust, or service reliability. It gives Beep more room to make that work repeatable.

Autonomous transit will be judged at ordinary moments: the vehicle arrives, the connection holds, the rider gets help, the agency sees the network, and the service survives another Tuesday. Mobileye's investment puts more weight behind Beep at precisely that handoff.

Wint

Most buildings learn about a water problem from the damage. The ceiling stains, equipment stops, a project slips, and only then does the pipe become a data source.

Wint Water Intelligence has raised a $36M Series D to move that decision earlier. LIP Ventures and returning investor Inven Capital co-led the round, financing product development and wider market reach for Wint's AI-powered water-intelligence platform.

CEO Alon Geva and co-founder and Chief Strategy Officer Yaron Dycian are building across a strangely fragmented budget. Water touches facilities, sustainability, insurance, construction schedules, energy use, and business continuity. Inside many organizations, those consequences belong to different teams until a leak introduces everyone.

Wint monitors flow patterns across building water systems, learns what normal looks like, and flags anomalies that may signal a burst pipe, stuck toilet, slow leak, or waste. Connected valves can shut water automatically or remotely. Local analysis, backup power, and cellular, Wi-Fi, and Ethernet options are designed to keep protection working when connectivity disappears.

The insurance layer makes this more interesting than another sensor story. HSB, part of Munich Re, distributes Wint and backs eligible construction projects with a performance warranty of up to $250K. HSB says a joint analysis with Wint across 66 sites found 73% fewer claims and 90% lower monetary losses at protected locations.

The 66-site study is useful evidence, not a promise for every building.

Wint says its systems saved 1.15B gallons of water in 2025, prevented an estimated $100M in damage across 1,300+ incidents, and avoided 39,000+ metric tons of carbon emissions. Those totals are company-reported. They also explain why the product is expanding from leak response into Water Insights and AI analysis of temperature and boiler health.

Inven co-led Wint's $35M Series C in 2023. Insight Partners led a $15M Series B in 2022, and Grosvenor made an undisclosed strategic investment after testing the platform in 2025. The disclosed B, C, and D rounds total $86M; the exact lifetime number remains higher but unknown.

The Series D gives Wint more room to become the shared operating layer between owners, contractors, insurers, and sustainability teams. Every new property adds different pipes, usage patterns, shutdown rules, and people authorized to act. Wint now has to make all of that variation learnable before the next ceiling supplies the evidence.

Bulwark Dynamics

The phrase “last-mile delivery” sounds harmless until the last mile is a contested beach with no pier, no crane, unreliable GPS, and no sensible reason to expose a crew.

That is the operating problem Bulwark Dynamics has chosen. The Menlo Park defense startup raised a $6.8M Seed round to develop and produce Caravel, an autonomous landing craft built to cross shallow water, beach itself, discharge cargo, and retract without shore infrastructure.

The named investors include SBI US Gateway Fund LP, Archetype Ventures Inc., Nēnē Ventures, and Onomichi Dockyard. No lead, valuation, or terms were disclosed. Bulwark also records an earlier undisclosed pre-seed, so $6.8M is the current transaction, not a defensible lifetime total.

The sharper part of this round sits beside the cap table. Bulwark signed an MOU with Onomichi Dockyard to work toward serial production of the Caravel series in Japan. The companies are also discussing a dedicated site for unmanned-vessel production.

Nhat Lieu, founder and CEO, Yuta Shiina, co-founder and CTO, and Isaac Anderson, co-founder and lead engineer, are making the shipyard part of the autonomy stack. Maritime software does not move a pallet after the hull reaches sand. Navigation, beaching, offload, retraction, maintenance, and production have to survive the same mission.

Bulwark says a 15-foot prototype completed an unmanned transit, beach landing, cargo offload, and retraction in early 2026. It is developing 35-foot and 55-foot variants, while the Seed capital supports production capacity, operational testing, and hiring in the U.S. and Japan.

The announcement is still a starting line. It does not disclose a government contract, production order, customer roster, deployment date, or independently audited performance record. An MOU can align intent; it cannot launch a fleet by itself.

The financing thesis is unusually concrete. Distributed operations across Indo-Pacific shorelines make fixed logistics nodes visible and vulnerable. Bulwark is betting that smaller autonomous craft, built with allied shipyard capacity close to the theater, can make resupply less concentrated and more replaceable.

The next proof sits between prototype speed and industrial repetition. Caravel has to keep landing, unloading, and leaving as the build moves into a production system shared across two countries. The beach will grade the autonomy. The shipyard will decide how many times the lesson can be repeated.