Spark 2026: Where Are AI Returns? Is the Proof in the Privates?
Global Thematics & Technology
Spark showed the AI payoff we expect seems to be materializing in privates, with startups across verticals reporting ROI for customers & rapid scaling. Trends: Open-weight models seem to be here to stay, value is shifting around the model, moats are forming where physical/digital worlds meet.
Key Takeaways
- Our Spark Private Company Conference brought together founders and CFOs across 100+ private companies. Privates are one signal of where the market is headed.
- We left convinced that 1) substantial AI value and ROI is being generated by private companies, and 2) as a result, startups are scaling faster than before.
- A core theme was the importance of open-weight models in unlocking Pareto optimal outcomes for AI adopters/enablers. We think this is a durable trend.
- Other takeaways were the value of everything around the model (harness/data), 3Ps, shift to agents, intersection of the digital/physical worlds, edge compute.
We left Spark more convinced in potential value creation across the AI ecosystem. Last week, we hosted over 100 private companies at our Spark Private Company Conference. Our biggest takeaway? Counter to the recent negative market moves, we left more bullish on the potential for value creation in this AI cycle. Today, substantial AI investment risk is being taken in the public markets, and investors are waiting for returns to justify this spend. Our conversations with companies across the startup ecosystem suggest to us three things: that 1) demand is here, 2) value is being created across the private landscape, and 3) as a result, privates are scaling more quickly than ever before, with companies like Fireworks reportedly growing ARR from $100M to $1B in eight months, Mercor claiming to have scaled ARR to $4B over 17 months, Cognition saying it has reached $1B of ARR in under two years, and Genspark saying it has generated $250M of ARR within 12 months of launch (cited by companies in firesides).
We are also seeing tangible ROI being generated across verticals spanning agentic AI, defense tech, physical AI, and infrastructure. Examples include, according to the companies, Cognition driving 30-40% productivity gains at customers adopting Devin, Replit helping a bank re-write legacy accounting software for under $100K against a $10-20M systems-integrator quote, Pallet processing documents for a customer at one-fifth of human cost with 99.5% accuracy, and Genspark helping cut a marketing cycle from 6 weeks to 2.5 hours at a customer. These are just a handful of examples, and in our view, suggests that AI demand is there...and companies are moving to meet it.
We note, of course, that recent negative market moves could prove warranted given the trillions of dollars of capex being spent in the market. Yet while capex and financing are a risk, we remain bullish on the cycle.
Keep reading for our thematic takeaways from the conference and thoughts from our teams.
For key MS Thematic Research relating to what we heard at the conference, see Public Meets Private: Key Morgan Stanley Thematic Research
For a list of attendees, see Getting Up to Speed: Overview of Our Private Company Attendees.
1. Open-weight models seem to be here to stay and are key to powering the AI innovation cycle.
We leave more convinced the world is hybrid model. We've been vocal on the importance of open-weight models for AI adoption (see work here, here, here), and it was hard to find a company that didn't talk about the strategic importance of open-weight models in unlocking Pareto optimal outcomes for its businesses or customers. We view open-weight models as a key enabler of the cycle, helping to drive greater enterprise adoption (via Jevons Paradox) and sustainable economics for AI application companies. Multiple companies talked about how Chinese model competition is driving frontier prices lower, which they see as a net-positive for all adopters, even those who don't use Chinese models (see our thoughts on this here). To be clear, we believe lab prices are also coming down because of model efficiencies, but regardless, this is a net good for the cycle – both to enterprise adoption via Jevons Paradox and margins of AI application vendors.
The importance of model routing is only increasing and driving ROI for companies. Companies highlighted that no single model wins every task. Legora – vertical AI for law – highlighted how different models are good for different law tasks (like Gemini and fact-checking) and how it benchmarks 18 different models across 200 criteria to route requests to the right model. Genspark dynamically routes tasks across 70+ models to optimize performance and inference costs, and says that its purpose-built slide-generation model outperformed GPT-4o, quoting costs of ~$0.30 per million tokens vs. ~$10. This is reportedly creating material cost savings – Replit said model routing helps generate frontier-level performance at ~40% of standard inference cost and is helping to drive margins higher, while Cognition said Devin outperformed Claude on a bug-detection benchmark at ~30–40% lower cost through model routing.
2. Models are table stakes, not moats, and true value comes from everything around the model.
As models converge, durability is moving to what surrounds them: domain workflows, proprietary data, evals, post-training and implementation. Legora, Replit, Cognition, and Fireworks all make the argument that the moat is owning the customers' workflow rather than just model access. Legora's differentiation and claimed 73% win rate is attributed to its legal expertise, manifesting, according to the company, both in terms of better product quality and better implementation (hired 180 former lawyers as 'forward deployed lawyers'). EliseAI makes the point that regulated industries require knowing when not to use AI, which requires context and domain expertise. EliseAI's agents decide when to quote an apartment price, but retrieve it from a verified table rather than generate it, highlighting the value of domain-specific controls. Enterprise trust, auditability, and agent governance are emerging as critical competitive moats.
Inferencing platforms and post-training have become even more important, as customers look for model customization to improve ROI on their AI spend. We're seeing more companies post-training open-weight models to unlock the insights within their organizations and maintain data security. We're also hearing companies increasingly concerned about labs training on their data (noted by Baseten and our own checks). We believe the inferencing platforms are sticky because of their post-training capabilities and expertise, as most companies don't have the in-house expertise to post-train at scale. Fireworks said customized models can cut costs 5-10x versus shared frontier-model APIs. This in turn could increase the value of model evaluation platforms like Mercor that are necessary for post-training and model selection.
3. People, Power, and Politics continue to be constraints.
We have written extensively about the 3P's (People, Power, and Politics – see the latest here, here, and here). Key startups in the power/data center space noted challenges with gaining access to power, as well as the fact that labor force and political challenges are only getting more difficult. One such company with ~2K employees and ~20K construction jobs commented that availability of electricians, plumbers, and welders are all constrained, and investment in skill trades education is necessary. Inference-aaS platforms also noted that getting access to power in the US is difficult. Cognition also noted that data center politics is a key macro risk it is watching. We continue to believe this increases the value of power players like BE, Solaris, and the PSPs.
4. Agents are everywhere.
Agents are moving beyond assistance and coding into the execution of everyday work. We heard this across housing, healthcare, customer service and finance. EliseAI described aiming to automate 20% of building operations labor through agents handling lease renewals and maintenance requests, as well as expanding into patient intake and scheduling. Legora said one deal-desk employee can now support 200+ sellers, with agents handling approximately 95% of routine inquiries. Cresta highlighted agents collecting payments and acting alongside human representatives during customer conversations, while Mercury said Command connects payroll setup, payment approvals and ledger posting from one prompt. We believe the TAM for agents is increasingly becoming that of labor – Cognition put Devin’s cost at approximately $4 per human-equivalent engineering hour versus its $35/hr benchmark for a software engineer, and guarantees customers backfilled credits if ROI falls below $10/ per human-equivalent hour. These task-specific comparisons make the ROI opportunity concrete as agents compete for labor spend.
Selling outcomes does not mean outcome-based pricing is already the norm. We came into Spark focused on software’s opportunity to address labor budgets and how vendors capture that value. The enterprise-agent panel highlighted customer preference for predictable contracted consumption. Cresta noted that outcomes can be difficult to define and audit, and rewarding call deflection can hurt retention. EliseAI made the point that auditable cost savings close contracts more readily than hard-to-prove revenue uplift. Net, we left more convinced of the customer value, while the share vendors capture, and how they charge for it, is still evolving.
5. Moats are being created at the intersection of the digital and physical worlds.
In a world where software is easily coded, companies are finding moats at the intersection of physical operations and software. Stord – a fulfillment and commerce software company powering fast deliveries and seamless returns – said its moat is their physical logistics footprint, especially in a world where software is easy to build. Varda – which builds recoverable spacecraft that manufacture pharmaceuticals in microgravity – said it created a business based on the convergence of two physical-world shifts: 1) large rockets began being reused, and 2) pharma was developing more complex drugs that benefit from microgravity. Flock Freight – which brokers truckload freight shipments between shippers and trucking companies to combine partial truck loads – claims a moat from combining a dense physical shipment network and an optimization software that pools partial loads to fill trailers, which saves shippers ~20% vs dedicated truckloads, pays carriers ~20% more, and wins them a higher GM profile vs. peers (~30% vs. ~10% for traditional brokerage. Anduril talked about how its secret sauce isn't their hardware but the integration with software.
In physical AI, manufacturing is the ultimate bottleneck to scale. The Hadrian, Genesis AI, and Nuro panel suggested the constraint is shifting from model capabilities alone to building and deploying hardware at scale. Genesis AI sees no obvious technical barrier to creating general purpose robotics models for the model architecture itself. The gating factor is manufacturing in the West, where electronics expertise exists but CNC and structural-part know-how is thin, and where most of its BOM still comes from China and can’t be unwound in the next 2-3 years. Genesis is betting on direct-drive hands over tendon-driven designs, which they believe match human muscle but can be a "manufacturing nightmare", and expects a hand that can run 24/7 for a month next year. Hadrian attacks the same gap from the upstream factory side. Machines across the A&D supply chain are productive only 10-15% of the time (about 1 in 10 minutes), versus 65%+ at Hadrian, which runs many customers’ parts rather than a single OEM’s (a model it calls Manufacturing as a Service or MaaS). Nuro, going driver-out in the Bay Area this year, argued lidar’s complementary data justifies its integration cost, and expects robotaxi insurance to fall below human-driven rates as safety data builds. We believe the winners will pair software with manufacturing capacity, making the physical-world moat in #5 increasingly a production moat. Many understand that lack of data is the ultimate limiter to creating smarter robotics models, but the 'data probes' (robots) still need to be manufactured, and today only China can produce them at scale.
6. Watching developments in AI hardware specialization and edge compute.
Our top takeaway from our conversations with semiconductor and hardware companies at Spark is the growing importance of AI optimization through specialization. As AI diffuses through the digital and, eventually, physical economy, we expect to see more new and different types of AI workloads, much like the chatbot to agent transition underway today. That growing workload diversity is creating opportunities in XPUs, software, and networking between the cloud and the edge. For example d-Matrix described its disaggregated inference solution with Nvidia, in which attention runs on GPUs while the memory bound expert layers run on d-Matrix’s specialized accelerator. Lightmatter and Upscale AI each made the case that moving data efficiently between GPUs, CPUs, networking, and storage will only become more important over time. Specialization is also becoming more important in software. Liquid AI says it is building specialized models for the edge, tailored to specific CPU/NPU architectures, while emphasizing smart routing between the edge and the data center. SiMa.ai made a related point for physical AI, arguing that it is primarily a software story: its agentic SDK reportedly cuts customer integration time from 6-9 months to under a day.
Some AI workloads need intelligence closer to where data is generated. In our preview, we highlighted latency and data sovereignty as drivers of on-premise and on-device inference. While we remain early, commercial deployments are beginning to demonstrate scale. Liquid AI said its 600M-parameter model powers in-car intelligence across ~1.7M Mercedes-Benz vehicles in North America—an early proof point that local inference is moving beyond prototypes. Verkada reported ~2.8M deployed devices and described AI-enabled trespass detection and automated warnings, emphasizing edge processing for low-latency responses. This illustrates how intelligence can enable intervention rather than passive monitoring.
Public Meets Private: Key Morgan Stanley Thematic Research
Morgan Stanley Research has an extensive corpus of Thematic and Private Company Research spanning the themes we saw at Spark. Below, please find key resources to get up to speed on these topics.
1. The Rise of the AI Factory.
As we wrote here, we are seeing new business models emerge around how to rethink the data center and coordinate increasingly complex infrastructure — power, compute, networking, cooling, and data — as part of a broader AI factory. We see 4 categories of companies in this theme: 1) Power (or accelerating access to it), 2) Compute (increasing useful compute delivered per unit of power, given key bottlenecks), 3) Integration of Power + Compute (redesigning power and compute as a single integrated system), and 4) Reimagining the Data Center. Venture funding for this universe of startups has accelerated sharply in 2026, reaching $33.5 billion by early October, up 2x from 2025. Power and Compute — or the scarcest inputs in the ecosystem (see our work on the power shortfall here and here) — have accounted for more than 80% of 2026 capital. However, we see capital broadening to Integration of Power + Compute and Re-imagining the Data Center.
How to get up to speed on these themes...see relevant team research below:
- US Thematics: Venture Vision: Data Center to AI Factory
- Global Thematics & US Semiconductors: What Does the Power Shortfall Mean for Semis? (5 Oct 2026)
- Powering AI: Racking Up the Power Shortfall (21 Sep 2026)
- Semiconductors: How can the AI ecosystem work around memory bottlenecks? (5 Oct 2026)
- Cerebras Systems: Unique capability in fast inference drives large economic opportunity; Initiate at OW (8 Jun 2026)
- Technology: Bringing More Light to Scale-Up Networks: An Updated Scale-Up Primer (13 Jul 2026)
- Video | Spring Training: Telecom & Networking Equipment and Cybersecurity (26 May 2026)
- Global Technology: AI Transceivers: Growth Dominates Disruption (26 Feb 2026)
- Telecom & Networking Equipment: Into the Spotlight: Optical Market Opportunities (23 Feb 2026)
2. AI's Move from Assistant to Agent.
Consumer and enterprise AI agents have come into focus after the launch of Muse, Google's Gemini AI agent, and others. On the enterprise side, our focus is on how AI-native software vendors are monetizing agents that execute work on a customer's behalf, a step beyond the assistant products that defined the first wave of enterprise AI. That step moves software's addressable market toward labor budgets, which are larger than IT budgets. There are three themes that will determine how much of that opportunity accrues to software vendors. The first is value capture, meaning how vendors frame the share of displaced labor cost they can charge for, and whether that share settles closer to 5% or 50% once customers keep their portion as margin. The second is pricing model evolution, as vendors move off per-seat toward usage- and outcome-based models. We will be listening for how outcomes are defined and verified in contracts, and what that means for revenue visibility. The third is unit economics, where gross margins depend on how efficiently vendors route each task to the lowest-cost capable model and shift workloads onto smaller or proprietary models.
- Microsoft: Battle of the AI Stack: Built for a Multi-Model World (7 Oct 2026)
- Meta Platforms Inc: 23 Days of Musing...Top Investor Questions and Our Answers (30 Sep 2026)
- Internet: The Morgan Stanley Agentic Playbook (18 Sep 2026)
- Software: The Moat & The Journey – Assessing Software Durability + Growth in the Age of AI (21 Jul 2026)
3. Model Customization and the Value of Inference-aaS Platforms.
As enterprise AI moves from experimentation into production and the world becomes multi-model, we expect a larger share of value in the model stack to accrue to the layers that serve and customize models. There's a few aspects of this debate that are important. The first is the rising importance of the orchestration layer, where platforms decide which model handles each request and how that workload scales across hardware. The value of this only rises as companies look to use both open and closed-weight models in order to rationalize token spend. We want to understand whether this layer can sustain pricing as model providers and hyperscalers offer competing serving options. The second is efficiency in deployment, as customers look to get more output from less compute through smaller distilled models and higher GPU utilization. The open question is how much of the resulting cost decline platforms keep as margin and how much they pass through to customers. The third is the growing value of post-training relative to pre-training, as customers use fine-tuning and reinforcement learning on proprietary data to adapt open models to specific tasks. We will be listening for whether post-training creates durable differentiation and switching costs. These themes also bear on the agent vendors above, since inference is a primary component of their cost of revenue and efficiency gains at this layer flow directly into their gross margins. The fourth is token cost monitoring and budgeting. Companies are increasingly looking for offerings to monitor their token spend and look more in a more granular fashion into ways to rationalize spend, and we're interested in how startups are addressing this.
- Microsoft: Battle of the AI Stack: Built for a Multi-Model World (7 Oct 2026)
- Internet and Software: Why Amazon Bedrock and Microsoft Foundry Should Become Household Names (7 Oct 2026)
- Global Thematics: Are Enterprises Really Adopting Open-Weight Models? (12 Oct 2026)
- Software: The Moat & The Journey – Assessing Software Durability + Growth in the Age of AI (21 Jul 2026)
4. Security Gets Rebuilt for the Agent Era.
Security in the age of AI has become an increasingly important topic, given the litany of agent breaches seen over the past few months. We believe increasing autonomous agent attacks, agents taking on new behaviors, and an increasing speed/frequency of breaches require new technologies to identify and protect against these threats. This is against an increasingly attractive spending backdrop — according to IDC, the ~$300bn cybersecurity market (software and services) is set to grow at a 15% CAGR over the next couple of years, a number we think could accelerate to 23% in our base case as magnified threats heighten the need for next-generation security.
- Cybersecurity & Global Thematics: Attack of the Agents: Cybersecurity Scenarios In 3 States of the World (20 Sep 2026)
- Cybersecurity: Agents on the Loose: Cyber Faces an Identity Crisis (20 Aug 2026)
- Software: Security in Runtime: SIEM Market Evolution (1 Jun 2026)
5. Physical AI finally becoming mainstream.
The implications seem very large. Over the next 12-18 months, we expect a number of large new entrants into the physical AI debate, including large private labs as well as companies in the Mag 7, which should both increase competition among the startups and bring substantial credibility into the theme. To be more specific, we think the recent shift among the labs from competing purely on who is at the edge of the frontier, to instead focusing on productization will eventually lead into the physical world in search of large TAM expansion opportunities (LLMs > coding agents > agentic platforms > consumer devices > devices that move, i.e., robotics). We believe the ripple effects are highly underappreciated.
- Robotics: RoboFest California Preview: Bots by the Bay (16 Sep 2026)
- Robotics: Actuators: Robot Muscle Needs American Hustle (13 Sep 2026)
- Physical AI: The Morgan Stanley Robot Almanac (7 Dec 2025)
6. Autonomy and Software Are Redefining Defense.
The line between commercial technology and national security is blurring, as geopolitical tension, drone warfare, and the speed of modern threats push governments toward software-defined, AI-enabled systems and away from traditional platforms. We believe the shift toward autonomy, real-time intelligence, and lower-cost, scalable systems is creating a new class of defense and public safety technology companies.
- Defense: Dollars, Disruptors, and De(Consolidation); Identifying Investment Opportunities (6 Mar 2026)
- Defense Tech: Anduril: a NeoPrime(r) (12 Jan 2026)
- Robotics: The Robot Almanac (Vol. 5): Space & Defense (18 Dec 2025)
7. Healthcare Tech: AI Targets the Administrative Burden.
Healthcare is a regulated, labor-intensive industry where a clear AI use case is administrative work, such as documentation, scheduling, intake, insurance verification, and revenue cycle. We see this as a close parallel to the shift from assistant to agent in software, with labor budgets rather than IT budgets as the addressable pool. Our team has written about the very large opportunity in healthcare revenue cycle management to tackle very significant inefficiencies facing the system today, with automation potential is highest in prior authorization, coding, and denials.
- Transforming Healthcare: AI Is Altering The Competitive Landscape In Revenue Cycle Management (5 Oct 2026)
- Transforming Healthcare: Private Company Research Series: Latest Checks On Ambient AI Scribe and LLM Use In Health Systems (20 Jul 2026)
- Transforming Healthcare: Private Company Research Series: Ambient AI Scribes Take Healthcare By Storm (24 Nov 2025)
8. Fintech Is Being Rebuilt Around Agents and New Rails.
We see three themes in fintech. First, agentic finance: As AI agents increasingly access data, manage money, and transact on behalf of consumers and businesses, we believe the infrastructure for payments, identity, and settlement has to adapt. Second, digital assets and stablecoins are moving from the periphery toward mainstream payments and capital markets infrastructure. Third, vertical fintech companies are embedding payments, credit, and insurance directly into industry workflows, often serving customers that incumbents have underserved. Key debates are how fraud and compliance costs evolve as AI raises the stakes on identity, and whether AI-driven automation can meaningfully lower the cost to serve.
- Digital Assets: Digital Assets: From Disruption to Convergence (8 Sep 2026)
- Global Banks & Exchanges: Digital Rails, Real Economics: Digital Assets & the Future of Wholesale Banking (2 Jun 2026)
Getting Up to Speed: Overview of Our Private Company Attendees
Company Profiles — By Fireside
The valuation information provided here is included for illustrative purposes only. The information is based on publicly available information (funding rounds, company disclosure, third-party vendors, etc.) and was not produced or endorsed by Morgan Stanley Research.
Anduril
| Field | Anduril |
|---|---|
| Description | Autonomous defense systems and weapons |
| Founded | 2017 |
| Last raise / valuation | ~$5b Series H (May 2026); ~$61b valuation |
| Key investors | Thrive Capital, Andreessen Horowitz (leads) |
| Operating scale | ~$2.2b revenue (2025) |
The valuation information provided here is included for illustrative purposes only. The information is based on publicly available information (funding rounds, company disclosure, third-party vendors, etc.) and was not produced or endorsed by Morgan Stanley Research.
- Company description: Anduril develops defense software and hardware (incl. autonomous vehicles, weapons, and sensors) primarily for US and allied governments. Its AI-enabled Lattice software integrates data from sensors, drones, and robotics to enable customers to detect, track and intercept unmanned aircraft systems (UAS) and autonomous drone systems. Beyond surveillance and counter-drone systems, Anduril also produces combat aircraft, missiles, solid rocket motors, underwater vehicles, and defense-ready compute, and communications infrastructure.
- Product/ service verticals: Lattice: Horizontal software operating platform connecting Anduril's hardware. It ingests sensor data from Anduril hardware, creates a common operating picture, enables command and control, and tasks autonomous systems. Its software development kit (SDK) also allows third-party hardware and applications to integrate with the platform. Sensing, surveillance & counter-UAS: Persistent UAS sensing and base and border protection through Sentry towers, Wisp infrared sensing, Spyglass/ Spark radar and Pulsar electromagnetic warfare. Anduril also produces UAS interceptors such as Anvil and Roadrunner. Autonomous systems & precision strike: Air strike through Ghost, Bolt, and Altius; autonomous strike via Fury; low-cost long-range strike via Barracuda family. Maritime underwater vehicles for surveillance, seabed monitoring, payload delivery and strike via Dive-LD, Dive XL, Copperhead and Seabed Sentry sensors. Tactical infrastructure: Deployable compute, communications, networking and soldier-level interfaces via products such as Menace, Voyager, EagleEye and Lattice Mesh. Solid rocket motors (SRMs): Produces SRMs, a critical missile component, for its own weapons and as a merchant supplier to third parties.
- Listed company relationships: Competes with AeroVironment (NASDAQ: AVAV) in portable strike drones and Kratos (NASDAQ: KTOS) in autonomous combat aircraft; its solid-rocket-motor business challenges Northrop Grumman (NYSE: NOC) and L3Harris (NYSE: LHX). RTX/Raytheon (NYSE: RTX) competes with Roadrunner-M in counter-drone interception but also partners with Anduril on advanced missile propulsion. Boeing (NYSE: BA) uses Anduril motors in its developmental Army interceptor, while Saab (NASDAQ Stockholm: SAAB B) selected Anduril to develop and produce motors for the Boeing–Saab Ground-Launched Small Diameter Bomb. Palantir (NASDAQ: PLTR) partners on military-data and command-and-control integration and leads the Army battlefield-intelligence program for which Anduril supplies hardware. L3Harris integrates night-vision equipment into Anduril’s soldier systems. Rheinmetall (XETRA: RHM) partners on European versions of Fury aircraft and Barracuda missiles; HD Hyundai Heavy Industries (KRX: 329180) partners on autonomous surface vessels; and Archer Aviation (NYSE: ACHR) co-develops the Thunder hybrid-electric aircraft, which remains under development. Rocket Lab (NASDAQ: RKLB) is contracted to provide three hypersonic test launches under a ~$30m agreement funded by Anduril.
- Key near-term milestones: 1) Initial surface-launched Barracuda missile deliveries for the US Army targeting 1H27; 2) Delivery of 600+ Bolt-M targeting April 2027; 3) Long Beach campus (~1.18m sqft) opening targeting 2027; 4) Air-launched Barracuda-500 deliveries targeting 2027.
Verkada
| Field | Verkada |
|---|---|
| Description | Cloud-based physical security platform |
| Founded | 2016 |
| Last raise / valuation | CapitalG-led investment (December 2025); ~$5.8b |
| Key investors | CapitalG (Alphabet, NASDAQ: GOOGL) (lead); NVIDIA (NASDAQ: NVDA) (July 2026) |
| Operating scale | ~$1b annualized bookings (December 2025); ~31k customers; ~2.4m devices |
The valuation information provided here is included for illustrative purposes only. The information is based on publicly available information (funding rounds, company disclosure, third-party vendors, etc.) and was not produced or endorsed by Morgan Stanley Research.
- Company description: Verkada is a physical-security company. Its cloud-based management platform, Command, consolidates disconnected security systems, centralizing monitoring and incident investigations across buildings.
- Product/service verticals: Command connects cameras with artificial intelligence (AI) footage search, door controllers and intercoms for entry, alarms and environmental sensors for alerts, and Workplace for visitor management. On-camera storage, combined with cloud management, removes the need for separate video recorders.
- Listed company relationships: NVIDIA (NASDAQ: NVDA) Partner on AI video-search development using Cosmos models, announced July 2026; CDW (NASDAQ: CDW) Reseller of Verkada licenses; Motorola Solutions (NYSE: MSI) Competitor through Avigilon Alta.
- Key near-term milestones: 1) Sound Systems speakers and audio players targeting Oct 2026; 2) Transportation fleet camera availability in the US and Canada targeting November 30, 2026.
Island
| Field | Island |
|---|---|
| Description | Enterprise browser and workplace security |
| Founded | 2020 |
| Last raise / valuation | ~$400m Series F (September 2026); ~$6.4b |
| Key investors | Evolution Equity Partners (lead); Sequoia Capital, Coatue, Insight Partners |
| Operating scale | Millions of users; ~1,000 employees |
The valuation information provided here is included for illustrative purposes only. The information is based on publicly available information (funding rounds, company disclosure, third-party vendors, etc.) and was not produced or endorsed by Morgan Stanley Research.
- Company description: Island sells enterprise-browser and workplace-security software that controls application access and data sharing.
- Product/service verticals: Enterprise Browser uses Chromium’s open-source foundation to enforce access, copy/paste and download controls locally. Enterprise Platform expansion extended shared policies to consumer browsers. Enterprise Network combines on-device checks with selective cloud inspection, reducing reliance on routing traffic through centralized security systems. Enterprise AI governs prompts and agent permissions, automates workflows and enables deployment of internal AI applications.
- Listed company relationships: Amazon (NASDAQ: AMZN) Customer; CrowdStrike (NASDAQ: CRWD) Integration partner, with Island feeding browser logs into Falcon security analytics; Palo Alto Networks (NASDAQ: PANW) Competitor via Prisma Browser.
- Key near-term milestones: N/A
Baseten
| Field | Baseten |
|---|---|
| Description | AI model inference platform |
| Founded | 2019 |
| Last raise / valuation | ~$1.5b Series F (June 2026); ~$13b |
| Key investors | Altimeter, Conviction, Spark Capital (leads); Sands Capital, Wellington (co-leads) |
| Operating scale | 20x revenue and 40x inference-volume growth yoy (March 2026) |
The valuation information provided here is included for illustrative purposes only. The information is based on publicly available information (funding rounds, company disclosure, third-party vendors, etc.) and was not produced or endorsed by Morgan Stanley Research.
- Company description: Baseten is a model-serving platform for AI inference, enabling developers and enterprises to deploy open-source, custom and fine-tuned AI models with optimized runtimes, high availability and developer-oriented workflows, either on Baseten-managed infrastructure or on the customer's cloud. It focuses on the infrastructure required to take trained AI models from development into production at scale.
- Product/ service verticals: Dedicated Inference: Runs customers’ proprietary, fine-tuned or open-source models. Sitting between models and underlying GPU/cloud infrastructure, Baseten optimizes execution and manages deployment, scaling, traffic routing, monitoring and reliability. Deployable on Baseten Cloud, customer-cloud or hybrid configurations. Model APIs: Provides pre-hosted, optimized models through APIs, without customers managing deployment. Training/post-training: Training Jobs runs customers’ existing training scripts on managed GPUs. Loops, its reinforcement learning training toolkit, remains in early access and supports moving trained models directly into production inference. Baseten for Model Labs: Enables closed-weight model developers to distribute and monetize models through branded APIs and Baseten’s model library, with access controls, billing and supporting infrastructure.
- Listed company relationships: NVIDIA (Nasdaq: NVDA) – Supplier of GPUs and Dynamo inference software with chips accessed through cloud providers; Alphabet (Nasdaq: GOOGL) – Supplier / Competitor on cloud compute and fine-tuned model hosting respectively; Amazon (Nasdaq: AMZN) – Competes on managed inference via SageMaker AI; HubSpot (NYSE: HUBS) – Customer of inference platform
- Key near-term milestones: 1) 3x headcount this year; 2) Next Base Labs research fellowship cohort starting Jan 2027.
Legora
| Field | Legora |
|---|---|
| Description | AI software for legal teams |
| Founded | 2023 |
| Last raise / valuation | ~$550m Series D (March 2026); ~$5.5b |
| Key investors | Accel (lead); Atlassian (NASDAQ: TEAM), NVIDIA (NASDAQ: NVDA) via NVentures (extension) |
| Operating scale | ~$200m+ ARR; ~2,100 firms; ~130k monthly users (September 2026) |
The valuation information provided here is included for illustrative purposes only. The information is based on publicly available information (funding rounds, company disclosure, third-party vendors, etc.) and was not produced or endorsed by Morgan Stanley Research.
- Company description: Legora provides AI software to law firms and in-house legal teams. It combines internal documents and legal sources to reduce manual research, review and drafting, while retaining lawyer oversight.
- Product/service verticals: Agent executes multi-step assignments; Tabular Review answers standardized questions across document sets in source-linked tables. Legal Research retrieves cited legal authorities; Word Add-in drafts documents and proposes tracked edits. Shared firm precedents, reusable instructions and document-system integrations connect these tasks into repeatable workflows.
- Listed company relationships: Salesforce (NYSE: CRM) Customer adopting Legora across its global legal organization, announced September 2026; Alphabet (NASDAQ: GOOGL) Supplies cloud/models and partners on a beta Gemini Enterprise for Legal connection announced August 2026; Thomson Reuters (NASDAQ: TRI) Competitor in research, review and drafting through CoCounsel Legal/Westlaw; RELX (NYSE: RELX) Competitor via LexisNexis’ Lexis+ with Protége.
- Key near-term milestones: 1) Legal-research ontology targeting Q4 2026
Mercor
| Field | Mercor |
|---|---|
| Description | Expert human data for AI |
| Founded | 2023 |
| Last raise / valuation | ~$350m Series C (October 2025); ~$10b; in talks at ~$20b (July 2026) |
| Key investors | Felicis (lead); Benchmark, General Catalyst, Robinhood Ventures (NASDAQ: HOOD) |
| Operating scale | ~$2b ARR (June 2026) |
The valuation information provided here is included for illustrative purposes only. The information is based on publicly available information (funding rounds, company disclosure, third-party vendors, etc.) and was not produced or endorsed by Morgan Stanley Research.
- Company description: Mercor supplies human expertise and training data to artificial intelligence (AI) laboratories and enterprises. Specialists translate professional judgment into tasks and scoring rules, helping customers train and test models on complex workflows rather than generic questions.
- Product/service verticals: Human Data recruits and manages expert teams, monetized through hourly billing and a percentage-based fee tied to expert pay. Its “Open Box” approach provides direct expert access and lets customers retain data on their own platforms. AI Productivity Index (APEX) benchmarks measure professional-task performance. Mercor Enterprise uses the same expert network and evaluation infrastructure to deploy custom agents. Deeptune adds sandboxed, simulated reinforcement learning, training agents in dynamic environments with tools and interfaces.
- Listed company relationships: NVIDIA (NASDAQ: NVDA) – Reported customer; Innodata (NASDAQ: INOD) – Competitor
- Key near-term milestones: N/A
Cognition
| Field | Cognition |
|---|---|
| Description | Autonomous AI software-engineering agents |
| Founded | 2024 |
| Last raise / valuation | ~$2b Series E (September 2026); ~$48b |
| Key investors | Andreessen Horowitz, Accel (leads); Founders Fund, General Catalyst |
| Operating scale | ~$1b ARR (September 2026) |
The valuation information provided here is included for illustrative purposes only. The information is based on publicly available information (funding rounds, company disclosure, third-party vendors, etc.) and was not produced or endorsed by Morgan Stanley Research.
- Company description: Cognition builds autonomous AI software engineers for enterprise engineering teams.
- Product/service verticals: Devin Cloud is its flagship product that can autonomously plan, write, test and ship production code inside the customer’s existing codebase. Related Devin interfaces include Desktop for IDE interfacing, CLI for command-line interfacing, and Review for code-review. Fusion is an execution architecture inside Devin.
- Listed company relationships: Amazon (NASDAQ: AMZN) – Strategic distribution partner; Cognizant (NASDAQ: CTSH) – Partner and Devin customer; Microsoft (NASDAQ: MSFT) - Competitor via GitHub Copilot’s cloud agent
- Key near-term milestones: N/A
Liquid AI
| Field | Liquid AI |
|---|---|
| Description | Compact on-device AI models |
| Founded | 2023 |
| Last raise / valuation | ~$250m Series A (December 2024); ~$2b |
| Key investors | AMD (NASDAQ: AMD) (lead) |
| Operating scale | ~48.7m model downloads |
The valuation information provided here is included for illustrative purposes only. The information is based on publicly available information (funding rounds, company disclosure, third-party vendors, etc.) and was not produced or endorsed by Morgan Stanley Research.
- Company description: Founded as an MIT spinout, Liquid AI develops compact AI foundation models for automakers, device manufacturers and enterprises. Its models run locally on devices, addressing cloud latency, privacy and power constraints.
- Product/service verticals: Liquid Foundation Models (LFMs), currently LFM2.5, process text, images and speech on-device; the Liquid Edge AI Platform (LEAP) customizes and embeds them in applications. Liquid Context supplies permissioned user memory to Liquid Agent, a task-executing assistant offered for manufacturer evaluation.
- Listed company relationships: Shopify (NASDAQ: SHOP) – Customer under a multi-year, multimillion-dollar license, with search in production by November 2025; Mercedes-Benz Group (XETRA: MBG) – Partner for multi-year embedded-speech for selected North American vehicles; Qualcomm (NASDAQ: QCOM) – Partner optimizing Context and Agent for Snapdragon Hexagon; AMD (NASDAQ: AMD) – Series A lead investor and partner optimizing LFMs for Ryzen AI processors.
- Key near-term milestones: 1) Mercedes-Benz in-vehicle deployment targeting 2H26; 2) LFM integration into MacPaw’s Eney assistant for macOS targeting later in 2026.
SiMa.ai
| Field | SiMa.ai |
|---|---|
| Description | Edge AI chips and software |
| Founded | 2018 |
| Last raise / valuation | ~$150m Series C (September 2026); ~$1.45b |
| Key investors | Fidelity, Amplify Partners (co-leads); Dell Technologies (NYSE: DELL) via Dell Technologies Capital |
| Operating scale | ~150 customers (September 2026) |
The valuation information provided here is included for illustrative purposes only. The information is based on publicly available information (funding rounds, company disclosure, third-party vendors, etc.) and was not produced or endorsed by Morgan Stanley Research.
- Company description: SiMa.ai supplies chips and software modules to robotics, drone, automotive and industrial-equipment manufacturers. It runs AI models locally, reducing cloud dependence within device power and latency constraints.
- Product/service verticals: Modalix machine-learning system-on-chip (MLSoC) runs vision and language workloads; Palette prepares and deploys models; open-source Palette Neat adds natural-language application development. Third-generation accelerator IP can be integrated into high-volume customers’ chips.
- Listed company relationships: Synopsys (NASDAQ: SNPS) – Partner on an automotive chip-design blueprint; Micron Technology (NASDAQ: MU) – Supplies low-power memory for Modalix modules; TSMC (NYSE: TSM) – Manufactures Modalix on its 6-nanometer process; NVIDIA (NASDAQ: NVDA) – Competes through Jetson Orin NX and Nano in embedded AI; SiMa modules are compatible with existing carrier boards.
- Key near-term milestones: 1) Third-generation hardware targeting 1H28 focusing on machine learning IP, chiplets, and system-on-chips (SoCs).
d-Matrix
| Field | d-Matrix |
|---|---|
| Description | In-memory AI inference chips |
| Founded | 2019 |
| Last raise / valuation | ~$275m Series C (November 2025); ~$2b |
| Key investors | BullhoundCapital, Triatomic Capital, Temasek (co-leads); Microsoft (NASDAQ: MSFT) via M12 |
| Operating scale | ~250+ employees (November 2025) |
The valuation information provided here is included for illustrative purposes only. The information is based on publicly available information (funding rounds, company disclosure, third-party vendors, etc.) and was not produced or endorsed by Morgan Stanley Research.
- Company description: d-Matrix supplies AI inference hardware to cloud providers and enterprises. Its chips compute within memory, designed to reduce data movement, response latency and power consumption when running trained models.
- Product/service verticals: Corsair is its flagship AI inference accelerator platform, using memory-centric compute to run the inference stage of pre-trained AI models at low latency. JetStream networking connects accelerators across servers; Aviator software compiles and executes models; SquadRack integrates all three into rack systems. The acquisition of Wallaroo.ai in August 2026 adds a deployment orchestration software.
- Listed company relationships: NVIDIA (NASDAQ: NVDA) – Partnered to integrate next-generation, stacked-memory Raptor accelerators into NVLink Fusion/MGX racks, targeting 4Q27 availability; Astera Labs (NASDAQ: ALAB) – Developing custom connectivity for D-Matrix-integrated NVIDIA racks; TSMC (NYSE: TSM) – Manufactures Corsair on its N6 process; Alchip (TWSE: 3661) – Corsair design/ manufacturing partner and Raptor co-developer.
- Key near-term milestones: 1) Raptor tape-out–final design handoff for fabrication targeted by year-end 2026, ahead of planned NVIDIA rack integration.
Replit
| Field | Replit |
|---|---|
| Description | AI app-building platform |
| Founded | 2016 |
| Last raise / valuation | ~$400m Series D (March 2026); ~$9b |
| Key investors | Georgian (lead); Andreessen Horowitz, Coatue |
| Operating scale | ~60m users (September 2026); ~$150m ARR (September 2025) |
The valuation information provided here is included for illustrative purposes only. The information is based on publicly available information (funding rounds, company disclosure, third-party vendors, etc.) and was not produced or endorsed by Morgan Stanley Research.
- Company description: Replit provides AI software-creation tools for entrepreneurs and enterprise teams. Its platform combines coding and deployment, reducing manual setup for prototypes, internal tools and customer-facing applications.
- Product/service verticals: Agent 4 writes, tests and deploys applications; Replit Design generates interactives mockups and live apps. Database is a fully-managed SQL database that can be added to Replit App to provide persistent data storage.
- Listed company relationships: Alphabet (NASDAQ: GOOGL) – Gemini-model supplier; Microsoft (NASDAQ: MSFT) – Partner via integrating with Azure and Rayfin and deploying Replit apps into Fabric; Visa (NYSE: V) – Customer; Wix (NASDAQ: WIX) – Competitor via Base44.
- Key near-term milestones: 1) New London office headcount could reach 100 by early 2027 (vs. ~20 in September 2026).
Lightmatter
| Field | Lightmatter |
|---|---|
| Description | Photonic interconnects for AI chips |
| Founded | 2017 |
| Last raise / valuation | ~$400m Series D (October 2024); ~$4.4b |
| Key investors | T. Rowe Price-advised funds (lead); Fidelity, GV; Lockheed Martin (NYSE: LMT), HPE (NYSE: HPE) |
| Operating scale | ~310 employees (March 2026); evaluation platforms shipping |
The valuation information provided here is included for illustrative purposes only. The information is based on publicly available information (funding rounds, company disclosure, third-party vendors, etc.) and was not produced or endorsed by Morgan Stanley Research.
- Company description: Lightmatter develops photonic (light-based) interconnects for AI chipmakers and hyperscale data centers. Optical links address the bandwidth, power and distance limits of electrical connections, allowing larger processor clusters to communicate efficiently.
- Product/service verticals: Passage L200 stacks optical and electronic chips, spreading connections across chip surfaces rather than only edges; Passage M1000 is a photonic interposer (light-routing base) beneath multiple chips. Forthcoming Passage L20 places optics beside processors, with bidirectional links that halve fiber requirements; Guide modules supply the laser light.
- Listed company relationships: NVIDIA (NASDAQ: NVDA) – NVLink Fusion partner (June 2026) adapting Lightmatter optics to connect custom processors to NVIDIA switches; GlobalFoundries (NASDAQ: GFS) – Fabrication partner for M1000; Marvell (NASDAQ: MRVL) – Competitor in optical interconnects via Celestial AI.
- Key near-term milestones: 1) Guide DR liquid-cooled laser-module sampling targeting Q4 2026; 2) Passage L20 CPX evaluation-kit shipments targeting Q1 2027
Upscale AI
| Field | Upscale AI |
|---|---|
| Description | AI data-center networking silicon |
| Founded | 2025 (launched) |
| Last raise / valuation | ~$190m Series A-1 (June 2026); ~$2b |
| Key investors | Premji Invest (lead); NVIDIA (NASDAQ: NVDA), Intel (NASDAQ: INTC), Qualcomm (NASDAQ: QCOM) |
| Operating scale | ~200 employees (March 2026); pre-revenue |
The valuation information provided here is included for illustrative purposes only. The information is based on publicly available information (funding rounds, company disclosure, third-party vendors, etc.) and was not produced or endorsed by Morgan Stanley Research.
- Company description: Upscale AI develops networking silicon, systems and software for AI data centers. Its open-standard fabric connects accelerators, memory and storage, targeting network bottlenecks in large, synchronized AI clusters.
- Product/service verticals: SkyHammer is its scale-up switch silicon, linking accelerators, memory and storage within a rack at low, deterministic latency. Scale-out Ethernet systems built on NVIDIA Spectrum-X switch silicon connect racks and clusters. Both run on an AI-optimized version of SONiC, the open-source network operating system.
- Listed company relationships: NVIDIA (NASDAQ: NVDA) – Investor (June 2026) and supplier of Spectrum-X silicon for scale-out systems; Intel (NASDAQ: INTC) and Qualcomm (NASDAQ: QCOM) – Series A investors; Broadcom (NASDAQ: AVGO), Cisco (NASDAQ: CSCO) and Arista Networks (NYSE: ANET) – Incumbent competitors in AI datacenter switching.
- Key near-term milestones: 1) Product launch event on October 8, 2026; 2) First product shipments targeting late 2026 to early 2027
Fireworks AI
| Field | Fireworks AI |
|---|---|
| Description | AI model inference and customization cloud |
| Founded | 2022 |
| Last raise / valuation | ~$1.5b Series D (July 2026); ~$17.5b |
| Key investors | Atreides, Index Ventures, TCV (leads); NVIDIA (NASDAQ: NVDA), AMD (NASDAQ: AMD) |
| Operating scale | ~$1b+ ARR; ~40t+ tokens daily (July 2026) |
The valuation information provided here is included for illustrative purposes only. The information is based on publicly available information (funding rounds, company disclosure, third-party vendors, etc.) and was not produced or endorsed by Morgan Stanley Research.
- Company description: Fireworks AI provides cloud infrastructure to customize and run AI models. Enterprises and developers use it for coding assistants and industry-specific applications without managing underlying compute.
- Product/ service verticals: Inference runs models on shared or dedicated GPU capacity, priced per token or GPU hour. Managed Training and the Training API customize models with customer data and feed directly into Inference. Nexus routes coding requests across open and proprietary models, with spending controls.
- Listed company relationships: Microsoft (NASDAQ: MSFT) – Distribution partner via Microsoft Foundry, billed through Azure; NVIDIA (NASDAQ: NVDA) – Investor and GPU supplier via cloud providers; AMD (NASDAQ: AMD) – October 2025 multiyear agreement for MI325X/MI355X access through cloud providers; Amazon (NASDAQ: AMZN) – AWS cloud and co-sell partner, but competes via Bedrock custom-model inference.
- Key near-term milestones: 1) 3–4x compute footprint expansion targeting October 2026.
Dataminr
| Field | Dataminr |
|---|---|
| Description | Real-time AI event and risk intelligence |
| Founded | 2009 |
| Last raise / valuation | ~$300m financing (September 2025), valuation undisclosed; last priced: ~$475m Series F (March 2021) at ~$4.1b |
| Key investors | Fortress, NightDragon (leads, 2025); Eldridge, Valor Equity Partners |
| Operating scale | ~$200m ARR (April 2025); 18 national militaries |
The valuation information provided here is included for illustrative purposes only. The information is based on publicly available information (funding rounds, company disclosure, third-party vendors, etc.) and was not produced or endorsed by Morgan Stanley Research.
- Company description: Dataminr provides real-time event and risk intelligence to enterprises, governments and newsrooms. Its AI turns public text, imagery and sensor data into alerts, reducing manual monitoring of threats to people, assets and operations.
- Product/service verticals: First Alert Advanced and Dataminr Advanced for Corporate Security, launched August and September 2026 respectively, add corroborating sources, context and near-term forecasts for government and corporate security teams. Dataminr for Cyber Defense (built on acquired ThreatConnect) combines external threat intelligence with internal security data; Dataminr for News flags breaking stories.
- Listed company relationships: Microsoft (NASDAQ: MSFT) – Integration partner feeding Dataminr alerts into Sentinel security workflows; Amazon (NASDAQ: AMZN) – Cloud host and distribution channel via AWS Marketplace; Mastercard (NYSE: MA) – Competitor in cyber-threat intelligence via Recorded Future.
- Key near-term milestones: N/A
Flock Safety
| Field | Flock Safety |
|---|---|
| Description | Public-safety cameras and software |
| Founded | 2017 |
| Last raise / valuation | ~$275m raise (2025); ~$7.5b |
| Key investors | Round lead undisclosed; Andreessen Horowitz (led March 2025), Tiger Global |
| Operating scale | ~$500m+ ARR (1H26); 12k+ customers (July 2026) |
The valuation information provided here is included for illustrative purposes only. The information is based on publicly available information (funding rounds, company disclosure, third-party vendors, etc.) and was not produced or endorsed by Morgan Stanley Research.
- Company description: Flock Safety supplies public-safety cameras, sensors and cloud software to police departments, businesses, schools and neighborhoods. Its platform links vehicle sightings and incident feeds to generate investigative leads for responders.
- Product/service verticals: License plate recognition (LPR) cameras capture plates and vehicle attributes; video cameras stream incidents; acoustic sensors locate gunfire; docked drones provide aerial views. FlockOS combines feeds on a command-center map; FreeForm searches footage using ordinary language. Its cross-agency network extends searches beyond a customer’s cameras, subject to customer-controlled sharing permissions.
- Listed company relationships: Motorola Solutions (NYSE: MSI) – Partner integrating Avigilon video with Flock plate alerts; Axon (NASDAQ: AXON) – Key listed competitor via Outpost LPR cameras and Fusus real-time crime center software; Amazon (NASDAQ: AMZN) – Cloud supplier via AWS.
- Key near-term milestones: 1) Mandatory search-misuse monitoring (Audit Assistance) and case codes for law-enforcement searches targeting year-end 2026
Vantor
| Field | Vantor |
|---|---|
| Description | Satellite imagery and geospatial intelligence |
| Founded | 1992 (as DigitalGlobe) |
| Last raise / valuation | No standalone raise; Maxar take-private (May 2023) at ~$6.4b EV |
| Key investors | Advent International (owner); British Columbia Investment Management (minority) |
| Operating scale | 3.5m+ sq km daily imaging capacity; 1.2m+ platform users |
The valuation information provided here is included for illustrative purposes only. The information is based on publicly available information (funding rounds, company disclosure, third-party vendors, etc.) and was not produced or endorsed by Morgan Stanley Research.
- Company description: Formerly Maxar Intelligence, Vantor provides satellite imagery, 3D mapping and AI-enabled geospatial software to defense, intelligence and commercial customers.
- Product/service verticals: WorldView satellites collect imagery on demand; Vivid supplies maps and three-dimensional (3D) terrain. Tensorglobe fuses sensor feeds using AI and powers Sentry automated site monitoring; Raptor matches drone imagery to 3D terrain for GPS-denied navigation.
- Listed company relationships: BAE Systems (LSE: BA.) – Selected June 2026 to build next-generation Vantage satellites; CACI (NYSE: CACI) – Selected September 2026 to supply Pulse optical payloads; Alphabet (NASDAQ: GOOGL) – Partner integrating Google Earth AI into Tensorglobe, including for classified environments; Planet Labs (NYSE: PL) – Competitor in high-resolution satellite imaging and site monitoring through SkySat/Pelican.
- Key near-term milestones: 1) Initial Pulse 40cm-class imaging satellite deployments targeting 2027–2029; 2) Vantage 20cm-class satellites entering service before 2030.
Varda Space Industries
| Field | Varda Space Industries |
|---|---|
| Description | In-orbit pharmaceutical manufacturing and reentry |
| Founded | 2021 |
| Last raise / valuation | ~$251m Series D (September 2026); ~$1.6b |
| Key investors | Lux Capital, Natural Capital (leads); Founders Fund, Khosla Ventures, General Catalyst |
| Operating scale | Six missions flown; 28 launches booked through 2029 |
The valuation information provided here is included for illustrative purposes only. The information is based on publicly available information (funding rounds, company disclosure, third-party vendors, etc.) and was not produced or endorsed by Morgan Stanley Research.
- Company description: Varda builds recoverable spacecraft that manufacture pharmaceuticals in microgravity and return them to Earth. The same capsules provide hypersonic reentry testing for US defense customers.
- Product/service verticals: W-Series spacecraft crystallize drug compounds in orbit and return them in detachable capsules, without a crewed station. Pharmaceutical services span ground screening, orbital processing and analysis, with future drug royalties targeted. Government contracts use ~Mach 25 reentries to test sensors and heat shields; government work is ~70% of 2027 payload demand.
- Listed company relationships: United Therapeutics (NASDAQ: UTHR) – Multi-mission pulmonary-drug formulation partner, announced May 2026; SpaceX (NASDAQ: SPCX) – Launch provider (W-6 on Transporter-16, March 2026) and competitor via its Starfall reentry program; Rocket Lab (NASDAQ: RKLB) – Former spacecraft-bus supplier, since replaced by Varda’s in-house design; Redwire (NYSE: RDW) – Competitor in orbital drug crystallization via SpaceMD.
- Key near-term milestones: 1) Nine cumulative commercial reentries targeting end-2026; 2) First human clinical dosing of a microgravity-enabled formulation targeting ~2030.
Stord
| Field | Stord |
|---|---|
| Description | Outsourced e-commerce fulfillment and software |
| Founded | 2015 |
| Last raise / valuation | ~$250m Series F (May 2026); ~$3b |
| Key investors | Existing investors (no single lead); Kleiner Perkins, Founders Fund, Franklin Templeton (NYSE: BEN) |
| Operating scale | 1,000+ brands; ~$15b+ annual GMV; ~100 fulfillment sites (May 2026) |
The valuation information provided here is included for illustrative purposes only. The information is based on publicly available information (funding rounds, company disclosure, third-party vendors, etc.) and was not produced or endorsed by Morgan Stanley Research.
- Company description: Stord provides outsourced fulfillment and commerce software to consumer brands selling online and through retailers. Its shared order-and-inventory platform connects warehouse operations and sales channels, reducing coordination across disconnected logistics providers and systems.
- Product/service verticals: Fulfillment covers storage, picking, packing and shipping through third-party carriers. Its order management system (OMS) routes orders to fulfillment sites; warehouse management system (WMS) manages inventory and warehouse workflows; Consumer Experience software adds delivery estimates, tracking and returns. Software integrates with Stord’s physical network but can also run customers’ own warehouses. StordAI Chat, launched in beta in March 2026, uses artificial intelligence (AI) to answer order and inventory questions. Revenue combines logistics fees and software subscriptions.
- Listed company relationships: UPS (NYSE: UPS) – Fulfillment partner since Stord acquired Ware2Go from UPS (May 2025); Procter & Gamble (NYSE: PG) – Fulfillment customer through Native; Amazon (NASDAQ: AMZN) – Competitor via Multi-Channel Fulfillment for off-Amazon orders.
- Key near-term milestones: N/A
Ayar Labs
| Field | Ayar Labs |
|---|---|
| Description | Optical interconnect chiplets for AI |
| Founded | 2015 |
| Last raise / valuation | ~$500m Series E (March 2026); ~$3.75b; ~$150m extension (September 2026) at >$5b (reported) |
| Key investors | Neuberger Berman (lead); NVIDIA (NASDAQ: NVDA), AMD (NASDAQ: AMD), MediaTek (TWSE: 2454), Alchip (TWSE: 3661), Wiwynn (TWSE: 6669), Intel (NASDAQ: INTC) via Intel Capital |
| Operating scale | ~200 employees |
The valuation information provided here is included for illustrative purposes only. The information is based on publicly available information (funding rounds, company disclosure, third-party vendors, etc.) and was not produced or endorsed by Morgan Stanley Research.
- Company description: Ayar Labs develops optical interconnects for AI chipmakers. Its silicon-photonics chiplets move data between processors using light rather than copper, addressing bandwidth and power bottlenecks in large-scale AI training and inference.
- Product/ service verticals: Ayar offers two complementary hardware products. TeraPHY is a silicon-photonics chiplet designed to bring optical connectivity directly into the processor package, converting electrical data into light signals for high-bandwidth communication across servers and racks. SuperNova is an external laser source that supplies light to TeraPHY.
- Listed company relationships: NVIDIA (NASDAQ: NVDA) – Strategic investor and partner integrating Ayar’s co-packaged optics (CPO) into NVIDIA rack-scale architecture; AMD (NASDAQ: AMD), MediaTek (TWSE: 2454) and Alchip (TWSE: 3661) – Strategic investors; Lumentum (NASDAQ: LITE) – External laser supplier under a 2022 agreement; Marvell (NASDAQ: MRVL) – Competitor via Celestial AI.
- Key near-term milestones: 1) Optical component qualification for volume production targeting end-2027, ahead of customer product ramps in 2028–2029.
Nuro
| Field | Nuro |
|---|---|
| Description | Licensed autonomous-driving technology |
| Founded | 2016 |
| Last raise / valuation | ~$203m Series E (August 2025); ~$6b |
| Key investors | Uber (NYSE: UBER), NVIDIA (NASDAQ: NVDA), Baillie Gifford; Fidelity, T. Rowe Price, Tiger Global |
| Operating scale | ~100 test vehicles in Bay Area and Houston (August 2026) |
The valuation information provided here is included for illustrative purposes only. The information is based on publicly available information (funding rounds, company disclosure, third-party vendors, etc.) and was not produced or endorsed by Morgan Stanley Research.
- Company description: Nuro licenses autonomous-driving technology to automakers and mobility operators for robotaxis, delivery fleets and consumer vehicles. Having shifted to licensing in 2024, it lets partners deploy driving automation without building the full system in-house.
- Product/service verticals: Nuro Driver combines AI software, sensors and onboard compute for driverless (Level 4) operation within defined conditions. Driver Assist targets hands-off, eyes-on consumer driving; Nuro Toolkit adds simulation, validation and remote-fleet tools.
- Listed company relationships: Uber (NYSE: UBER) – Investor (~$500m committed including milestone payments) and exclusive ride-hailing and fleet-operations partner for the planned robotaxi service; Lucid (NASDAQ: LCID) – Integrates Nuro into Gravity and future Midsize vehicles under Uber’s April 2026 purchase commitment for at least 35,000 vehicles; NVIDIA (NASDAQ: NVDA) – Investor and supplier of DRIVE AGX Thor onboard compute.
- Key near-term milestones: 1) Bay Area robotaxi launch with Uber targeting late 2026; 2) Houston launch targeting mid-2027
Hadrian
| Field | Hadrian |
|---|---|
| Description | Automated aerospace and defense manufacturing |
| Founded | 2020 |
| Last raise / valuation | ~$1.37b Series D (August 2026); ~$7.87b |
| Key investors | WCM, Washington Harbour, Valor Equity Partners, 137 Ventures, Baillie Gifford (leads); RTX (NYSE: RTX) via RTX Ventures |
| Operating scale | ~3m sq ft capacity across four facilities |
The valuation information provided here is included for illustrative purposes only. The information is based on publicly available information (funding rounds, company disclosure, third-party vendors, etc.) and was not produced or endorsed by Morgan Stanley Research.
- Company description: Hadrian is a software-enabled contract manufacturer and factory operator that produces precision metal components to specifications for aerospace and defense customers. Its proprietary Opus software coordinates automated machining, production scheduling and inspection. Customers purchase finished parts or engage Hadrian to build and operate dedicated production lines or entire factories, including within their own facilities.
- Product/ service verticals: Precision Components: Manufacture precision parts, including components produced using computer numerical control (CNC) machining, for aerospace and defense customers from a product’s prototype to production stage. Manufacturing-as-a-Service: Provide dedicated manufacturing capacity or inspection cells, deployed either in Hadrian facilities or customer sites, supporting customer-specific production requirements. Factories-as-a-Service: Designs, builds and operates dedicated production infrastructure for a customer or program, to service entire production cycle
- Listed company relationships: RTX (NYSE: RTX) purchases components through Collins Aerospace and Raytheon, including Javelin missile parts, and is a strategic investor through RTX Ventures. Lockheed Martin (NYSE: LMT) partners under a December 2025 MOU to embed machining and inspection cells inside LMT’s Missiles and Fire Control facility in Grand Prairie, Texas.
- Key near-term milestones: 1) Progress on additional factories and munitions/ autonomous-system production lines targeting August 2027; 2) Fourth Alabama Factory phase-one targeting full-rate production within 24 months of contract award.
Genesis AI
| Field | Genesis AI |
|---|---|
| Description | General-purpose robots and robot AI |
| Founded | 2024 |
| Last raise / valuation | ~$105m seed (July 2025) |
| Key investors | Eclipse, Khosla Ventures (co-leads) |
| Operating scale | Dozens of robots built; ~60 employees (June 2026) |
The valuation information provided here is included for illustrative purposes only. The information is based on publicly available information (funding rounds, company disclosure, third-party vendors, etc.) and was not produced or endorsed by Morgan Stanley Research.
- Company description: Genesis AI develops general-purpose robots and robotics foundation models for manufacturing, logistics and laboratories. Its integrated hardware and learned control target variable, dexterous tasks that fixed automation cannot handle.
- Product/service verticals: Eno, unveiled June 2026, pairs a wheeled, height-adjustable body and human-like dexterous hands with GENE, its AI model for planning and controlling multistep tasks. Its data engine uses glove-recorded human actions and physics simulation to train and test models, aiming to reduce costly robot demonstrations.
- Listed company relationships: LG CNS (KRX: 064400) – Partner evaluating Eno for manufacturing and logistics at LG’s US operations (June 2026); NVIDIA (NASDAQ: NVDA) – Competing robot-training stack via Isaac GR00T models and Isaac Sim; Alphabet (NASDAQ: GOOGL) – Competitor in robot foundation models via Google DeepMind’s Gemini Robotics.
- Key near-term milestones: 1) Eno production start and initial customer deployments targeting end-2026.
Shield AI
| Field | Shield AI |
|---|---|
| Description | AI pilots and autonomous aircraft |
| Founded | 2015 |
| Last raise / valuation | ~$1.5b Series G (June 2026); ~$12.7b post-money |
| Key investors | Advent (lead), JPMorgan Chase (NYSE: JPM) (co-lead); Blackstone (NYSE: BX) funds (preferred equity) |
| Operating scale | Not disclosed |
The valuation information provided here is included for illustrative purposes only. The information is based on publicly available information (funding rounds, company disclosure, third-party vendors, etc.) and was not produced or endorsed by Morgan Stanley Research.
- Company description: Shield AI develops AI pilot software and unmanned aircraft systems (UAS) for US and allied militaries to operate aircraft with limited human intervention.
- Product/ service verticals: Hivemind: Flagship AI-powered pilot software that allows for centralized control of a wide range of unmanned systems to execute complex missions in disrupted and GPS/ comms-denied environments. V-BAT: Combat-proven unmanned aircraft system (UAS) designed for contested GPS/ comms-denied environments. It offers unassisted vertical take-off (VTOL) technology and swarming capabilities. X-BAT: Development-stage AI-piloted, vertical take-off and landing (VTOL) stealth fighter jet designed to operate without a traditional runway. It can carry air-to-air, air-to-surface weapons, and large strike weapons. Vision Systems: Shield AI’s computer-vision hardware and software portfolio where its Visual Detection and Ranging (ViDAR) electro-optical and infrared (EO/IR) sensors and Tracker software automate detection and tracking for surveillance and counter-drone missions. Aechelon: Simulation software for pilot training and autonomous-system testing; acquired June 22, 2026.
- Listed company relationships: Competes with AeroVironment (NASDAQ: AVAV) and Textron (NYSE: TXT) in tactical unmanned aircraft/ISR, and with Boeing (NYSE: BA)/Insitu, RTX (NYSE: RTX)/Collins, Northrop Grumman (NYSE: NOC) and Lockheed Martin (NYSE: LMT) in mission-autonomy software. L3Harris (NYSE: LHX) licenses Shield's Tracker counter-drone software and partners on electronic-warfare autonomy; RTX (NYSE: RTX), Airbus (Euronext Paris: AIR) and Northrop Grumman (NYSE: NOC) integrate Hivemind into weapons, sensors or aircraft, although RTX (NYSE: RTX) and Northrop Grumman (NYSE: NOC) also field competing autonomy software. Palantir (NASDAQ: PLTR) partners on command-and-control integration and supplies manufacturing software, while GE Aerospace (NYSE: GE) supplies X-BAT's F110 propulsion system. Shield also works with Mitsubishi Heavy Industries (TSE: 7011) and Korea Aerospace Industries (KRX: 047810) to develop sovereign AI-pilot capabilities using Hivemind.
- Key near-term milestones: 1) X-BAT’s first vertical take-off and landing (VTOL) flights targeted for 2026, production for 2029; 2) Polish Navy V-BAT deliveries targeted for later in 2026; 3) Cooperative multi-aircraft operations with Republic of Singapore Air Force using Hivemind targeted for 2027 following manned/ unmanned teaming in 2026.
MoonPay
| Field | MoonPay |
|---|---|
| Description | Crypto payments and on/off-ramp infrastructure |
| Founded | 2019 |
| Last raise / valuation | ~$555m Series A (November 2021); ~$3.4b; reported talks at ~$5b with ICE (December 2025) |
| Key investors | Coatue, Tiger Global; Intercontinental Exchange (NYSE: ICE) (reported talks); Galaxy (March 2025 debt) |
| Operating scale | ~$108m revenue (2021); 160+ countries; ~440 employees (May 2026) |
The valuation information provided here is included for illustrative purposes only. The information is based on publicly available information (funding rounds, company disclosure, third-party vendors, etc.) and was not produced or endorsed by Morgan Stanley Research.
- Company description: MoonPay supplies crypto-payment infrastructure to wallets, fintechs and merchants. It connects bank/card payments with blockchains, bundling identity checks, liquidity and settlement so partners need not build these capabilities separately.
- Product/service verticals: Ramps embeds cryptocurrency purchases and cash-outs within apps; Trade routes token exchanges across blockchains. Commerce provides merchant checkout with crypto or traditional-currency settlement. Enterprise, launched August 2026, converts bank deposits into stablecoins and supports cross-border payouts and branded stablecoin issuance. Shared infrastructure reduces vendor integration; availability remains jurisdiction-dependent. Institutional supplies wallet-security and trading infrastructure to financial firms.
- Listed company relationships: PayPal (NASDAQ: PYPL) – Payment partner enabling purchases and cash-outs in eligible markets; Paysafe (NYSE: PSFE) – Customer using Commerce and Virtual Accounts for US online-gaming payments, launched April 2026, with upgraded checkout live September 2026; Coinbase (NASDAQ: COIN) – Competitor serving wallet/app developers through Onramp and Offramp.
- Key near-term milestones: 1) Rollout of Ether.fi partnership in next few months with Ether.fi moving core payments infrastructure onto MoonPay’s integrated stack
Mercury
| Field | Mercury |
|---|---|
| Description | Banking and finance software for startups |
| Founded | 2017 |
| Last raise / valuation | ~$200m Series D (May 2026); ~$5.2b |
| Key investors | TCV (lead); Sequoia Capital, Andreessen Horowitz, Coatue |
| Operating scale | 300k+ customers (September 2026); ~$650m annualized revenue (September 2025) |
The valuation information provided here is included for illustrative purposes only. The information is based on publicly available information (funding rounds, company disclosure, third-party vendors, etc.) and was not produced or endorsed by Morgan Stanley Research.
- Company description: Mercury provides banking, payments and financial software to startups and small businesses. It combines accounts, cards, bill pay and accounting in one platform, reducing reconciliation across separate systems.
- Product/service verticals: Free business accounts connect IO corporate cards, bill payments, invoicing and expense controls; Treasury invests surplus cash in short-term funds. Venture debt and working-capital loans finance startups and ecommerce businesses, respectively. Books, launched September 2026, categorizes and reconciles transactions; Command uses artificial intelligence (AI) for customer-approved financial tasks.
- Listed company relationships: State Street (NYSE: STT) – Partner supplying an exclusive government money-market fund share class; Intuit (NASDAQ: INTU) – Competitor via QuickBooks, also integrated with Mercury; Capital One (NYSE: COF) – Competitor in corporate cards and spend management through Brex.
- Key near-term milestones: 1) Embedded payroll via acquired Central targeting 2026; 2) Treasury Ladders (individual US Treasury portfolios) launch targeting 2026.
EliseAI
| Field | EliseAI |
|---|---|
| Description | Vertical AI agents for housing and healthcare operations |
| Founded | 2017 |
| Last raise / valuation | ~$350m (September 2026); ~$4b |
| Key investors | Andreessen Horowitz, Bessemer Venture Partners, Ontario Teachers' Pension Plan (co-leads); Sapphire Ventures, Navitas Capital |
| Operating scale | $200mm+ ARR (2026); 1 in 6 apartments in the U.S. (2026); 30M unique renters reached since launch (Sep 2026) |
The valuation information provided here is included for illustrative purposes only. The information is based on publicly available information (funding rounds, company disclosure, third-party vendors, etc.) and was not produced or endorsed by Morgan Stanley Research.
- Company description: EliseAI is a New York company founded in 2017 by CEO Minna Song and CTO Tony Stoyanov. Its vertical AI agents automate leasing, resident, maintenance and patient-access work for rental-housing operators and specialty physician groups through integrations with customers’ workflow systems.
- Product/service verticals: Housing combines LeasingAI, ResidentAI, VoiceAI and EliseCRM, a customer relationship management (CRM) hub: it answers renters across phone, text, email and webchat, books tours, routes work orders, manages renewals and collections, and centralizes records. Apollo, launched in September 2026, is a prompt-based agent that executes permissioned tasks across the platform. Healthcare automates calls, scheduling, intake, referrals, insurance verification, chart preparation and follow-up.
- Listed company relationships: Zillow Group (NASDAQ: Z, ZG) – partner on AI Assist embedded in Zillow Rentals listings; Vivmark Residential (NYSE: VMRK) – successor to Equity Residential and AvalonBay, which EliseAI named as customers in August 2024; post-merger contract status is undisclosed; AppFolio (NASDAQ: APPF) – direct competitor in AI leasing, maintenance and resident operations through Realm-X; Phreesia (NYSE: PHR) – competitor in healthcare voice, scheduling, intake and payment workflows.
- Key near-term milestones: N/A
Genspark
| Field | Genspark |
|---|---|
| Description | Agentic AI workspace for knowledge workers |
| Founded | 2023 |
| Last raise / valuation | ~$100m Series B extension (June 2026); ~$2.6b post-money |
| Key investors | Emergence Capital (led March 2026 extension); Sozo Ventures, Mirae Asset, UpHonest Capital (extension); LG Technology Ventures, SBI, Pavilion Capital |
| Operating scale | ~$250m ARR (June 2026) |
The valuation information provided here is included for illustrative purposes only. The information is based on publicly available information (funding rounds, company disclosure, third-party vendors, etc.) and was not produced or endorsed by Morgan Stanley Research.
- Company description: Genspark, a wholly owned subsidiary of MainFunc, is an agentic AI workspace for knowledge workers and enterprises. Users state an objective; the service gathers context, routes work across models and tools, then returns editable deliverables or executes actions, reducing manual switching among research, office creation and business applications.
- Product/service verticals: SecondBrain retains work context; Super Agent’s Mixture-of-Agents routes subtasks across models; Office, Build and Content suites create documents, spreadsheets, presentations, code, databases and media; GenTeam and Claw add human-agent collaboration and cloud/desktop execution. September 2026’s Gen-1 Slides is Genspark’s first model further trained for presentation work.
- Listed company relationships: Microsoft (NASDAQ: MSFT) – Agent 365 launch partner with Word/Excel/PowerPoint integration; Amazon (NASDAQ: AMZN) – preferred cloud and Bedrock supplier for agents; Twilio (NYSE: TWLO) – Programmable Voice supplier for Call for Me in 40+ countries; Alphabet (NASDAQ: GOOGL) – Genspark incorporates Gemini models and is listed on Google Cloud Marketplace; Gemini Enterprise is a potential workplace-agent substitute.
- Key near-term milestones: N/A
Cresta
| Field | Cresta |
|---|---|
| Description | AI platform for contact centers |
| Founded | 2017 |
| Last raise / valuation | ~$125m Series D (November 2024); ~$1.6b valuation in Series C 2022 raise |
| Key investors | WiL, QIA (leads); Accenture, EnvisionX, LG Technology Ventures, Qualcomm (NASDAQ: QCOM) via Qualcomm Ventures, Workday (NASDAQ: WDAY) via Workday Ventures; Andreessen Horowitz, Greylock, J.P. Morgan (NYSE: JPM), Sequoia Capital, Tiger Global |
| Operating scale | ~$100m ARR (April 2026, Sacra estimate) |
The valuation information provided here is included for illustrative purposes only. The information is based on publicly available information (funding rounds, company disclosure, third-party vendors, etc.) and was not produced or endorsed by Morgan Stanley Research.
- Company description: Founded at Stanford’s AI Lab, Cresta sells AI software to large travel, financial-services, telecommunications, healthcare and retail contact centers. Its platform automates suitable voice and digital contacts, guides human agents on complex interactions and analyzes conversation data, consolidating automation, agent assistance and quality management on one system.
- Product/service verticals: AI Agent handles voice and digital conversations; Agent Assist supplies live guidance, source-backed knowledge and summaries; Conversation Intelligence analyzes 100% of interactions for quality, coaching and automation. Opera lets customers configure workflows without code; Conductor helps developers build, test and optimize agents using customer-conversation data.
- Listed company relationships: United Airlines Holdings (NASDAQ: UAL) – customer using AI Agent, Agent Assist and Conversation Intelligence; TELUS Corporation (TSX: T; NYSE: TU) – its wholly owned TELUS Digital unit is a preferred implementation and managed-services partner; Amazon.com (NASDAQ: AMZN) – cloud infrastructure and marketplace provider for Cresta’s SaaS offer; NiCE (NASDAQ: NICE) – direct competitor through CXone’s AI agents, Copilot and interaction analytics.
- Key near-term milestones: N/A
Flock Freight
| Field | Flock Freight |
|---|---|
| Description | Shared-truckload freight brokerage |
| Founded | 2015 |
| Last raise / valuation | ~$60m Series E (May 2025), valuation undisclosed; >$1b at Series D (January 2022) |
| Key investors | O'Neil Strategic Capital (lead); SoftBank Group (TSE: 9984) via Vision Fund 2 (Series D lead) |
| Operating scale | Thousands of shippers and carrier partners (November 2025) |
The valuation information provided here is included for illustrative purposes only. The information is based on publicly available information (funding rounds, company disclosure, third-party vendors, etc.) and was not produced or endorsed by Morgan Stanley Research.
- Company description: Flock Freight is a US freight broker for retail, food-and-beverage and industrial shippers. Its patented shared truckload (STL) software pools compatible shipments onto one truck, so shippers pay only for space used and avoid terminal transfers.
- Product/service verticals: FlockDirect sells STL capacity at spot or contracted rates based on pallets or linear feet; conventional full-truckload and less-than-truckload (LTL) services share quoting, booking and tracking tools. STL AddOns fills carriers’ remaining truck space with compatible loads. Transport is procured from third-party carriers.
- Listed company relationships: Oracle (NYSE: ORCL) – Integration partner for FlockDirect quotes within Oracle Transportation Management; WiseTech Global (ASX: WTC) – Integration partner via e2open; Academy Sports + Outdoors (NASDAQ: ASO) – STL customer; C.H. Robinson (NASDAQ: CHRW) – Competitor in US truckload and LTL brokerage.
- Key near-term milestones: N/A
Pallet
| Field | Pallet |
|---|---|
| Description | AI agents for logistics operations |
| Founded | 2022 |
| Last raise / valuation | ~$27m Series B (May 2025); valuation undisclosed |
| Key investors | General Catalyst (lead); Bain Capital Ventures, Activant, Bessemer |
| Operating scale | ~2m annual agent runs at Lineage (single customer) |
The valuation information provided here is included for illustrative purposes only. The information is based on publicly available information (funding rounds, company disclosure, third-party vendors, etc.) and was not produced or endorsed by Morgan Stanley Research.
- Company description: Pallet provides AI agents for logistics providers, manufacturers, distributors and retailers. Its agents read emails and documents and execute tasks in existing operating systems, reducing manual coordination without requiring replacement of core software.
- Product/service verticals: CoPallet agents automate quoting, order entry, tracking and billing; Atlas combines customer-system data with agent activity, letting operators ask plain-English questions about routes, capacity and customer opportunities. Forge generates and tests agents; Memory supplies customer-specific operating rules.
- Listed company relationships: Lineage (NASDAQ: LINE) – Customer automating receiving, freight-forwarding and warehouse orders; WiseTech Global (ASX: WTC) – CargoWise integration used to create orders in Lineage’s system; Manhattan Associates (NASDAQ: MANH) – Competitor via embedded Manhattan Active agents.
- Key near-term milestones: 1) 3m+ Lineage shipments processed in 2026, with expansion across US and international sites.
AtoB
| Field | AtoB |
|---|---|
| Description | Fuel cards and trucking payments |
| Founded | 2019 |
| Last raise / valuation | ~$130m Series C, equity and debt (September 2024); valuation undisclosed |
| Key investors | General Catalyst, Bloomberg Beta (leads); Mastercard (NYSE: MA) |
| Operating scale | 35k+ fleets (undated) |
The valuation information provided here is included for illustrative purposes only. The information is based on publicly available information (funding rounds, company disclosure, third-party vendors, etc.) and was not produced or endorsed by Morgan Stanley Research.
- Company description: AtoB provides fuel cards, payments and working-capital tools to trucking companies and fleets. Its cards and payment accounts help control fuel spending and fund immediate operating costs before freight invoices are paid.
- Product/service verticals: Flex fuel card provides weekly credit; Unlimited is prefunded. Both offer fuel discounts, transaction reporting and dashboard spending controls. Business Expense cards finance broader operating expenses in weekly installments; the digital wallet receives freight payments and supports round-the-clock driver payouts. Vehicle-location and fuel-level checks block suspicious purchases.
- Listed company relationships: Mastercard (NYSE: MA) – Investor and exclusive network for AtoB’s commercial card program; Uber Technologies (NYSE: UBER) – Partner via Uber Freight’s AtoB-powered carrier fuel card; Murphy USA (NYSE: MUSA) – Partner via the AtoB-powered Murphy USA Business Fuel Card (2026); WEX (NYSE: WEX) – Competitor in fleet cards and SecureFuel fraud controls; Corpay (NYSE: CPAY) – Competitor in fuel and expense cards through Comdata and Corpay One.
- Key near-term milestones: N/A