The Scoreboard Stays Up

Reflection AI

Open intelligence still arrives with a very closed invoice.

Reflection announced a $2B Series B in October 2025 at an $8B post-money valuation. NVIDIA reportedly supplied about $800M, making it the largest disclosed participant in a broad investor group.

That accounting matters. Reflection did not raise a separate $800M round from NVIDIA. The chipmaker's check sat inside a $2B financing built for one of technology's most expensive jobs: assembling the compute, data, research talent and reinforcement-learning infrastructure required to train frontier models.

Misha Laskin and Ioannis Antonoglou used their Google DeepMind experience to build Reflection around a different ownership argument. Enterprises, governments and sovereign entities should be able to inspect, customize and operate powerful models without sending their intelligence through somebody else's closed system.

The ambition sounds open. The supply chain is concentrated enough to make a customs officer blush.

NVIDIA was helping finance a future customer whose research plan would consume enormous amounts of accelerated computing. Reflection later signed for more GB300 capacity through SpaceXAI, while NVIDIA remained investor, infrastructure supplier and beneficiary of the demand created by the lab it helped capitalize.

That loop became visible last week when Reflection introduced Beam, its first open-weight model. The company describes Beam as a 501B-parameter mixture-of-experts system with 23B active parameters, built for coding, reasoning and agentic work.

Reflection says Beam was pretrained on 23.8T tokens and its reinforcement-learning run generated more than 100M rollouts on 10,500 NVIDIA GB300 GPUs over 4 weeks. Those are company-reported figures. Beam remains in final red-teaming and evaluation before Reflection releases the weights, technical report, model card and developer artifacts.

This funding story is easier to judge than it was a year ago. The $800M was never the product. It bought time inside a brutally expensive handoff from research thesis to compute contract to trained model.

Misha Laskin and Ioannis Antonoglou now have to keep that handoff moving. Buyers and developers will ask whether Beam performs, deploys safely and creates enough value to support the machinery underneath it.

NVIDIA financed part of that machinery. Reflection has put a model at the other end. The next evidence belongs to the people who run Beam outside Reflection's building.

Isomorphic Labs

Max Jaderberg now helps run a company where the promising molecule has to leave the computer. Isomorphic Labs made its former Chief AI Officer President in January, putting the researcher who helped build its engine closer to the whole business of making medicines.

That business may command a remarkable price. Bloomberg reported on October 8 that the Google DeepMind spinout is in early funding talks at a valuation of at least $40B, potentially $50B. The financing has not closed, terms could change, and the new round's size and investors remain undisclosed.

Founder and CEO Demis Hassabis has already assembled serious backing. Thrive Capital led Isomorphic's $2.1B Series B in May, with Alphabet, GV, MGX, Temasek, CapitalG and the UK Sovereign AI Fund participating. The company raised $600M in its first external round in March 2025. Those investors belong to completed financings, not a confirmed October lineup.

The useful detail is what Isomorphic is building around the model. IsoDDE, its drug design engine, searches for molecules against biological targets and desired properties. The company describes one computational design example running over two to four days, followed by making selected compounds and testing them in the lab. That sequence still ends with scientists handling actual material.

Isomorphic has internal oncology and immunology programs alongside collaborations with Novartis, Eli Lilly and Johnson & Johnson. A pharmaceutical partner brings target knowledge, development experience and its own decisions about which programs deserve more work. The engine has to become useful inside that relationship.

Clinical expertise is joining the machinery too. CMO Ben Wolf's appointment accompanied a Cambridge, Massachusetts presence in 2025. Jaderberg's move from AI leadership to President puts the computational and organizational ambitions under the same operating brief.

On October 7, it joined the Biohub-convened Virtual Biology Initiative, an effort to build accessible biological data for predictive AI. The training material needs engineering too.

A bigger valuation discussion gives that brief a very public audience. Isomorphic is trying to carry AI, data, chemistry and medical development inside one company, while its partners work on difficult targets. The lab result still has to become a candidate someone is prepared to take further toward patients, and the people around Jaderberg have to build an organization that can keep carrying that work.

Universal Quantum

Sebastian Weidt and Winfried Hensinger put a seam in their quantum computer. Universal Quantum is building smaller chips that can be joined together, making the connection between them part of the machine's future.

That choice now has more than $100M in Series A funding behind it. DCVC and Firgun Ventures co-led the round announced on October 8, financing the Sussex company's modular trapped-ion architecture and its international expansion.

Weidt, CEO, and Hensinger, Chief Scientist and Chairman, founded the University of Sussex spinout in 2018 after co-authoring a 2017 blueprint. Their engineering question has stayed remarkably consistent: how do you add more useful qubits without making the machinery around them impossible to manage?

Charged atoms held by electric fields do the quantum work. Universal Quantum's iQPU brings the chip and classical controls together; UQConnect transfers qubits between modules, while UQLogic uses microwave fields to control operations. The company specifies 70K operation, so cooling remains part of the job.

A modular design gives engineers somewhere to divide the work. It also gives them interfaces to make reliable. Chip fabrication, control electronics and error correction have to agree with one another before a larger machine can earn its place in a customer's computing environment.

Zak Romaszko heads Quantum Chips. The company's technical team has customer development work underway: the German Aerospace Center confirms the Toccata and Legato prototype projects. Universal Quantum reports a $77M DLR contract, a separate commercial relationship from this equity round.

The investor group reaches well beyond the co-leads: EDBI, Artal advised by Invus, Integral GlobalTech, Main Sequence, Hostplus, NGS Super, IAG Capital, EdenBase, Eigenstate Holdings, SiteGround Capital, Dave Baszucki, Nauta Capital and The Venture Collective. Baszucki participated personally; Roblox is not named as a corporate investor.

Singapore puts a practical location inside the expansion plan. Universal Quantum has announced an R&D centre built around chip development and packaging capabilities, alongside further work in the US, Japan and Hamburg. The intended applications include materials and drug discovery; large-scale, error-corrected utility remains the engineering destination.

More capital lets the founders carry more of that work in parallel. The chips still have to meet, and the teams in Sussex, Hamburg and Singapore have to make those meetings repeatable.

Wally

The hardest part of a $249 dental membership is not convincing someone to try it. It is making the fifteenth location feel as trustworthy as the first, then doing it 85 more times.

Wally Health has raised a $25M Series A led by Maveron, with Bling Capital and Jack Abraham participating. The company also announced a strategic partnership with EMS North America.

Co-founder and CEO Tyler Burnett built Wally with co-founders Stipe Latkovic and Chelsea Acosta Patel around a blunt consumer complaint: dental care can be expensive, uncomfortable, hard to price, and easy to postpone until something hurts.

Wally charges $249 per year for unlimited cleanings and diagnostics without requiring insurance. Members use Swiss AIRFLOW cleaning technology and Guided Biofilm Therapy, digital X-rays, 3D intraoral scans, saliva testing, and AI-assisted detection.

The company reports more than 50,000 members across 15 locations in New York City, Philadelphia, Jersey City, and Chicago. Members visit an average of three times per year. Wally also reports a 4.9-star rating and 97% hygienist retention.

Those are company-reported operating metrics, not independently audited clinical outcomes. They still reveal what this round is buying: repeat behavior from patients and a workforce willing to stay inside the model.

The American Dental Association found that 41% of working-age adults visited a dentist in 2023, compared with 28% of people without dental insurance. Wally told Axios that roughly 60% of its members are uninsured. The customer is buying a predictable way to return before pain writes the schedule.

Wally plans to grow from 15 locations to 100 by the end of 2027, beginning with Washington, D.C., Atlanta, Austin, and Miami. The capital will also support hiring and continued investment in the software that coordinates scheduling, operations, and clinical workflows.

That expansion creates an unforgiving handoff. A clinical network has to reproduce simplicity through leases, equipment, training, local compliance, staffing, diagnostics, and thousands of appointments where software must support judgment rather than impersonate it.

Maveron's consumer thesis fits the bet. Wally is trying to make preventive dentistry behave like a service people choose repeatedly, not an insured event they tolerate occasionally.

The Series A gives Wally more doors to open. The valuable proof will accumulate inside them, one returning member and one retained clinician at a time.

TrueLayer

Italy is investing in a payment network that two Italians started in London. TrueLayer's newest backer is following the engineering talent as well as the company address, with Milan doing work that reaches bank customers across Europe.

TrueLayer announced a $27M investment on October 8, led by CDP Venture Capital alongside existing and new investors. Co-founder and CEO Francesco Simoneschi and co-founder and CTO Luca Martinetti started the business in 2016. Its ambition is to make paying straight from a bank account an ordinary part of online commerce.

The company says its Pay by Bank network now serves more than 30M European consumers across 22 countries. Around a quarter of its workforce, 70 people, is based in Italy. TrueLayer expects to keep hiring there through 2027, giving CDP's European payment-infrastructure argument a place where engineers actually show up to work.

That argument eventually lands in a much smaller space: the payment screen. A merchant has to offer the option, a customer has to trust the bank connection, and the transaction has to finish without turning a purchase into homework. Ownership of the rails becomes commercially useful when someone chooses to travel on them.

TrueLayer has been widening the occasions for that choice. Its June Bank on File launch brought recurring bank payments to Trading 212, IG Group, InvestEngine and East Lothian Housing Association. This week, it announced instant personal-account top-ups in the SumUp App. An investment contribution and an account top-up give the network different ways into a customer's routine.

The product work is surprisingly human. TrueLayer's design team describes replacing payment jargon with the simple request to link a bank account for future payments. Customers consent to an ongoing relationship, so clarity about what they authorize becomes part of the infrastructure itself.

The financing follows a $50M extension of TrueLayer's Series E announced in 2024. Today's $27M is a separate investment, arriving as the company adds more everyday uses for bank payments and puts its European expansion agenda in front of an Italian backer.

CDP is backing that work as Europe looks for more control over its payment systems. The investment's series and valuation were not disclosed. For Simoneschi and Martinetti, the useful connection runs from Italian engineering desks to businesses willing to put Pay by Bank in front of their customers, then back through the next payment those customers choose to make.

Hightouch / Strata

Kelsey Mullaney's next job puts an integration founder inside the revenue conversation. After buying Strata, Hightouch is bringing its CEO into revenue programs and go-to-market strategy, close to the customers who need those connections to work.

Hightouch announced the acquisition on October 9. Mullaney built Strata with Julian Mclain and Ilya Brin after the three met as early employees at Iterable. They had seen the same request recur: a marketer wanted to use customer data, and an engineer had to build the route.

That request has a long afterlife. The connection needs to survive the next change to an API, a field or a tool. Buying another specialist application also buys another maintenance obligation, usually assigned to someone with a very full calendar.

Strata lets teams describe the data they want moved and its destination. Its agents generate the integration and keep it running as systems change. The product site describes API schemas, transformations, rate limits, bidirectional flows and dry-run testing. Quite a lot of work sits behind a marketer's apparently simple instruction.

Hightouch co-CEOs Tejas Manohar and Kashish Gupta, alongside co-founder and CTO Joshua Curl, have been building toward a marketing platform where agents can plan and execute with current customer and business context. Custom integrations determine how much of that context actually arrives.

Curl identifies broader coverage as the reason for the purchase. Hightouch expects Strata to help accelerate its integration roadmap, ease onboarding and give customers more flexibility as their stacks evolve. Those benefits are the company's plan; the announcement gives no measured post-acquisition results.

Hightouch raised $150M at a $2.75B valuation in April, led by Goldman Sachs and Bain Capital Ventures. That financing belongs to its wider AI marketing buildout. The Strata purchase price remains undisclosed, and the company has given no product migration timetable.

Mullaney brings knowledge of the integration problem into the commercial job of explaining how Hightouch will solve it. A buyer asking to connect another tool is also asking who will maintain that connection after the implementation team has moved on.

As Hightouch expands the systems its marketing agents can reach, that second question follows Mullaney into the customer conversation. It reaches past the first successful sync and into the next change made by somebody else's software.

Gallatin AI

A sustainment plan can be correct at breakfast and dangerous by lunch.

A route closes. Consumption outruns the forecast. A vehicle goes down. Medical demand moves somewhere nobody expected. The spreadsheet still has all its cells. The mission has already changed.

Gallatin AI has raised a $50M Series A to make that gap smaller. 8VC, Silent Ventures, and other investors backed the round, giving the company capital to expand engineering, accelerate product development, and broaden its AI-enabled logistics capabilities.

Woody Glier, Daniel Buchmueller, and Brian Ballard built Gallatin around an unglamorous truth: every advanced military system eventually becomes a sustainment problem. Autonomous vehicles, sensors, aircraft, software, and weapons still need fuel, parts, medical supplies, munitions, transport capacity, and people who can decide what moves first when the plan breaks.

Navigator is Gallatin's decision-support layer for that work. The platform connects fragmented data and workflows, forecasts demand, flags shortfalls, and recommends supply and movement plans that logisticians can adjust as conditions change. Gallatin says it is already used by Army and Air Force customers across tactical, operational, and strategic planning, with exercises spanning Japan, Alaska, and Europe.

The Defense Innovation Unit confirmed a Gallatin prototype contract under the Joint Sustainment Decision Tool program for U.S. Army Europe and Africa. Gallatin has also disclosed work with III Armored Corps, Army PORTAL, Palantir environments, and an Arctic rotation in Alaska.

The funding arithmetic deserves the same precision. Gallatin reports $70M raised in total, all equity. Its public $15M seed and $50M Series A account for $65M; the reviewed disclosures do not explain the remaining $5M or name a Series A lead or valuation.

That qualification does not shrink the operating signal. Gallatin is trying to move military logistics from periodic reporting toward a system that keeps rebuilding the decision as reality changes. The hard part will be proving Navigator can stay useful when data is messy, routes are contested, communications degrade, and multiple commands need the same supplies for different reasons.

The future force will carry more autonomy into the field. It will also create more machines to fuel, repair, load, route, and recover. Gallatin's next phase lives inside that handoff, where software has to help the people responsible for keeping every other system moving.

Stord

An independent brand puts its name on a delivery promise carried out by people the customer will never meet. Stord is building a business around making that promise easier to keep while the brand keeps the relationship.

Sean Henry, CEO, and Jacob Boudreau, CTO, co-founded the Atlanta company in 2015. Their answer joins fulfillment, shipping and commerce software, so inventory, orders and the experience after checkout can operate together. The customer sees a parcel. The merchant needs the whole operation to agree.

Stord has secured a $400M credit facility led by Citi, with Morgan Stanley, JPMorgan, First Citizens, Citizens and KeyBank participating. It follows approximately $250M in Series F financing at a $3B valuation in May. The latest announcement adds borrowing capacity; Stord has not disclosed how much it has drawn.

The facility will support network expansion, robotics and automation at Stord Labs, and further software and AI investment. That combination fits a company whose growth involves both code and buildings. A better routing decision still needs inventory in the right place and a warehouse team able to execute it.

At its Atlanta lab, Stord tests robotics and AI against fulfillment workflows on the same warehouse and order management systems used in production. The commercial opportunity is to carry an improvement across a shared operating environment, giving individual brands access to work they would struggle to fund and maintain alone. A merchant can keep its storefront and customer data while asking Stord to handle a more complicated logistics job.

Stord reports nearly 100 fulfillment locations worldwide, more than 1,000 brands and over $15B in annual merchandise volume. Those figures describe the network and the commerce moving through it. They should not be confused with Stord's own revenue or with 100 company-owned warehouses.

Bill Zerella is joining as CFO after finance leadership at ACV Auctions and Fitbit. Mark Wayland is joining as CRO after Box and Tanium. Finance and enterprise selling now sit beside the founders' operating and technical responsibilities as the company expands what merchants can buy from one provider.

For a brand trying to grow its own storefront, the attraction is practical: reliable delivery without assembling a logistics company on the side. Stord's expansion will reach that merchant through an order routed correctly, a package leaving on time and a customer willing to come back to the brand that made the promise.

Pivot Energy

Solar construction capital only scales if yesterday's projects stop occupying tomorrow's borrowing capacity.

Pivot Energy has closed a $173M term loan facility covering 51 community-scale solar projects, 135 MWdc across six states. First Citizens Bank, Huntington National Bank and BankUnited provided the takeout financing.

The facility refinances operating projects out of Pivot's construction warehouse and consolidates its first three portfolios. It is Pivot's first term loan, which sounds minor until you follow what happened to the warehouse.

Construction facilities fund assets while they are being built. Leave completed projects there too long and every operating array keeps consuming capacity intended for the next site.

Bret Labadie, CFO, and his team are moving 51 projects into permanent debt and reopening the revolving facility for more distributed solar and storage. No new panel appears at closing. The balance sheet becomes capable of financing the next one.

That matters in a community solar market that reached 10.1 GWdc of cumulative capacity in 2025 while annual installations fell 25%. Deployment can wobble even as electricity demand and grid pressure keep expanding.

Three banks committing $173M to a diversified, six-state portfolio does not erase policy, interconnection or operating risk. It shows that distributed assets can be aggregated into a package large enough for institutional term debt rather than financed forever as a collection of construction projects.

Founder Rick Hunter started Pivot in 2009. Tom Hunt, CEO, later led its shift from a developer managing less than 10 MW into an ECP-owned independent power producer with more than 4.3 GW completed or under development.

The financing machinery grew with it. In 2024, Pivot arranged a $450M warehouse led by First Citizens with ATLAS SP Partners, plus structured equity from HASI, to support 300 MWdc across 96 projects. This term loan is the next handoff: three portfolios leave construction capital so the warehouse can keep moving.

Distributed solar is also a repeat-financing business. Developers assemble smaller assets, move each cohort through construction, prove the operating portfolio and return borrowed capacity to work.

Pivot's $173M close turns that sequence into infrastructure. The next evidence will arrive project by project, as released warehouse capacity becomes another group of local power assets rather than another line trapped in a financing announcement.

Matchpoint Therapeutics

Novartis holds an option on a Matchpoint Therapeutics program. Nextech Invest and Norwest are financing the next stretch of work around another one, with a $150M Series B closed on October 7, 2026.

That gives CEO Andre Turenne and CSO Jared Cumming two development relationships to carry forward. MPT-062, an oral T-cell modulator for autoimmune and inflammatory diseases, is slated to enter clinical development in 2027. A separate program continues with Novartis funding under a July 2025 agreement.

Matchpoint came out of work by scientific founders Nathanael Gray, Tinghu Zhang, Jianwei Che and Edward Chouchani, drawing on Stanford, Dana-Farber and Harvard Medical School. Their field asks an unusually consequential question about attachment: can a small molecule hold onto a disease-causing protein precisely enough to make a useful medicine?

The Watertown, Massachusetts company's Advanced Covalent Exploration platform combines cellular chemoproteomics, machine learning and an evolving library of covalent compounds. The discovery machinery looks for useful binding sites and checks selectivity across proteins. A stronger bond becomes interesting when the team can control where it forms.

The new syndicate includes Invus, BB Biotech, T1D Fund and BOLD Longevity Growth, alongside returning Access Biotechnology, Atlas Venture, Sanofi Ventures, Digitalis Ventures and other existing investors. Nextech Managing Partner Kanishka Pothula and Norwest Principal Brian Matesic were announced to join the board as MPT-062 moves toward clinical work.

Matchpoint previously disclosed a $30M seed and a $70M Series A led by Sanofi Ventures. Add the Series B and those equity rounds total $250M. The Novartis agreement has different arithmetic: up to $60M in upfront payment and research funding, with up to $1B in total potential payments plus royalties. Those conditional amounts belong to the agreement, with receipt and future milestones still separate questions.

The dedicated clinical budget now has a named job: carry MPT-062 toward proof of concept while advancing the broader pipeline. In parallel, Matchpoint leads the Novartis program through development-candidate selection; option exercise would hand global development and commercialization rights to its partner.

Inside the same company, the scientists are preparing one program for clinical development and another for a possible partner handoff. Each path will ask something different of the discovery work that brought Matchpoint this far.

Phinity Labs

Silicon has an expensive editing process. At Phinity Labs, Sonya Jin and Aadi Nashikkar are building toward a future where a chip team can explore more designs before deciding which one deserves to become physical. Those discarded options have a cost, too.

The San Francisco company announced a $5.2M Seed on October 8, led by Uncork Capital, with Moxxie Ventures and angels including Jeff Dean participating. Pear VC led the earlier pre-seed through PearX, and Uncork also names Pear among the seed backers. That capital is backing an applied research lab whose ambition reaches from architecture discovery through tapeout.

Chip engineers have to make choices about power, speed and space while the software their hardware will run keeps changing. A design that looks elegant on a screen still needs to satisfy physical constraints. Exploring another possibility consumes time from people who already have a formidable amount to verify.

Phinity is putting a learning loop into that work. Agents propose designs, evaluate them against physical feedback and revise them. The interesting commercial possibility is a wider menu of useful architectures, with more of the search happening before an engineering team commits its resources to a particular answer.

The cofounders came into this from model training. Sonya Jin worked on synthetic data for NVIDIA's chip-design and code-reasoning models and graph ML at AWS AI Lab. Aadi Nashikkar brought experience in synthetic data, post-training and a previous startup exit. They have assembled a team combining AI research with silicon-design experience, putting learning systems beside people who know what it takes to ship hardware.

Phinity says it works with major foundation-model labs and has reached an eight-figure annualized run rate. Those customers are unnamed, and the figure is company-reported. Uncork describes commercial work running alongside development of Phinity's own agents: the business can earn while confronting the engineering problems its research needs to understand.

There is already serious AI inside chip design. Synopsys and Cadence sell tools for optimizing power, performance and area. Phinity's homepage sets 2028 as its target for taking a requested chip to a GDS layout file in weeks. For the team choosing its next custom architecture, that ambition reaches into the budget meeting, where an interesting design still needs someone willing to pay for exploring it.

Precision Chroma

A scientist feeding a chromatography column has already made a series of purification decisions. Precision Chroma wants to let the column take on more of that work, starting with the material that comes straight out of a bioreactor.

The Christchurch company has closed a $7.3M Seed round led by M Ventures, with founding investor Bridgewest Ventures NZ returning and Icehouse Ventures and Aspire NZ Seed Fund joining. CEO and co-founder Sean Feast is taking research from the University of Canterbury toward the company’s first research and process-development products.

That audience matters. The scientists designing a purification process know what has to happen before a valuable protein reaches the column. Conventional biologics workflows typically use centrifugation and depth filtration to remove cells and debris before Protein A chromatography. Each operation adds equipment, consumables and another opportunity to lose product.

Precision Chroma changes the architecture inside the bed. Its 3D-printed agarose structures have ordered passages designed to let cells and debris move through while therapeutic proteins bind. Direct affinity capture from an unclarified harvest could remove those preceding clarification steps. The design keeps familiar affinity chemistry, operating conditions and existing chromatography systems in the conversation.

A scientist evaluating a new separation method has fewer changes to understand when the equipment and chemistry remain recognizable. Precision Chroma is concentrating the novelty in the structure that does the separating.

M Ventures' Investment Director Oliver Hardick knows the journey from research to a bioprocessing business. He founded Puridify in 2013 and led it to acquisition by GE Healthcare in 2017. In this announcement, he points to the gap between upstream productivity and downstream purification, alongside the adoption challenges facing new chromatography technologies.

The $7.3M will support manufacturing scale-up, larger-format columns, team growth and deeper work with biopharmaceutical manufacturers. Precision Chroma's product page anticipates samples of its ProA Complete Capture Columns in late 2026; GMP manufacturing applications sit further along the development path.

For Sean Feast, the next exchange puts Canterbury's engineered structures in front of scientists with their own cell cultures, target antibodies and purification decisions. Their work will help establish where those first columns belong in the process.