Vantora Raises $100M to Mass-Produce Physical AI Startups

Vantora has raised $100 million in funding, sharpening its focus on physical AI. Formerly known as UP.Labs, the company builds startups for corporate clients. According to reporting by TechCrunch AI’s Kirsten Korosec, the investor is Silversmith Capital Partners. The announcement was made on September 18, 2026. In addition to the rebrand and the investment, the company also announced a shift in its business model.

The company was founded in 2022. Its first corporate partner was Porsche. It later expanded partnerships to include Alaska Airlines, J.B. Hunt, Wabash, and TDG. TDG is the parent company of Ashley Furniture. It is also working with new clients in industrial manufacturing and oil and gas. Some client names have not been disclosed. Its approach of designing startups around corporate clients’ problems remains unchanged. However, its thinking on target markets and ownership has changed significantly. It has introduced a model that does not assume expansion into external markets.

Strategic Shift Toward Proprietary Acquisitions

Vantora is shifting to what it calls a “proprietary M&A pipeline.” Corporate partners invest and adopt the product as the first customer. The partner then has the option to integrate the startup into its core business. Founder and CEO John Kuolt explained the strategy. It is a mechanism that allows startups to be retained in-house rather than sold externally. Previously, startups were designed with expansion beyond the corporate client in mind. Now the focus is on building businesses exclusively for corporate clients. The aim is to make it easier to handle projects directly tied to a partner’s competitiveness. Development can proceed even in highly confidential areas. It also offers the advantage of clarifying ownership of intellectual property. For corporate clients, adoption decisions become easier. For Vantora, it opens a path into large-scale projects. The model sits between contract development and investment. It differs from typical venture-building models that aim for gains from external sales. Its defining feature is deep integration with the client company’s business strategy.

This shift also affects relationships with existing clients. Projects previously conceived but set aside are now subject to reconsideration. The more a project concerns a corporate client’s core operations, the greater the value of exclusivity. Partners can control the development direction through investment and adoption. An option to acquire the startup upon completion is secured in advance. Startups can validate in real-world settings from an early stage. Securing a proving ground is especially important in physical AI. Trials in factories and logistics networks determine development speed. They can leverage partners’ equipment and operational data. There is no need to rush to generalize for external markets. They can focus on the requirements of a specific client. That allows for deeper technical refinement. Partners can also lead decisions on scaled deployment. This framework is seen as the premise for the large investment.

Background and Rationale for the Focus on

Physical AI

The tilt toward physical AI is closely linked to the proprietary model. Kuolt spoke candidly about past challenges.

“We were missing on the biggest value problems, which had the biggest upside because of that”

The point is that the most valuable problems could not be externalized, and therefore could not be tackled. Autonomy for industrial machinery is a prime example. It requires retrofitting existing fleets of equipment for autonomous operation. There is a strong need to own the intelligence layer in-house. Reliance on third parties is difficult to accept. Partners will not allow sales to competitors. These constraints had been the limit of the previous model. The proprietary model removes this barrier. It makes it possible to take on highly confidential autonomy projects. Kuolt also said:

“Imagine you’re a Fortune 100 industrial company and you need to retrofit all of your hardware and machines for autonomy. You need to own that, it needs to be sovereign, and you can’t rely on a third party to go do that for you.”

The key point is securing a sovereign intelligence layer. Designs are required that do not expose device control or operational data externally. On-site processing and the use of dedicated circuits are also at issue. As shown by Tenstorrent Publishes GCC Patch for RISC-V Core “Ascalon XG”, development of on-device computing infrastructure is also progressing. Autonomous mobility and inspection automation are likely early use cases. Predictive maintenance and process optimization are also in scope. All are areas tied to machines in the field. Language models alone are not sufficient. Measurement, control, and learning must be integrated. This integrated domain is the core of physical AI. Vantora plans to concentrate resources there.

An idea related to J.B. Hunt was cited as a specific example. It was reportedly an AI application to advance the company’s business. It was set aside because the partner would not allow external deployment. Under the proprietary model, such projects can be pursued. Development aligned with real-world logistics operations becomes possible. On-site data from vehicles and warehouses can be utilized. It is freed from the constraints of building a general-purpose product. Adjustments can be made to fit the partner’s operating procedures. Results accumulate within the partner. Spillover to competitors can be contained. For autonomy retrofits, phased introduction is realistic. The proprietary model is well suited to phased introduction. That is because iterative field validation can be completed internally.

Partnership Track Record with Industrial

Clients and Current Expansion

Vantora’s partnership network spans automotive and aviation to logistics and furniture manufacturing. It has launched multiple startups for Porsche. With Alaska Airlines, it partnered in operations and customer service. With J.B. Hunt and Wabash, it collaborated at the intersection of logistics and equipment. With TDG, it collaborated in manufacturing and supply networks. In each case, the company was designed starting from a problem. The common thread is that the corporate client becomes the initial customer. Solving on-the-ground problems becomes the product requirement. Feedback from real-world operations guides development. This cycle is the company’s strength.

Industrial manufacturing and oil and gas have newly been added. The company names have not been disclosed. In asset-intensive industries, demand for autonomy is large. Aging equipment retrofits and labor shortages are behind this. Maintaining safety and uptime is a management challenge. There are many situations where AI-based monitoring and control are effective. In oil and gas, remote monitoring is particularly significant. Automated inspection of wide-area facilities is required. Improving anomaly detection and response decisions is a challenge. In manufacturing, automating transport between processes is a focus. Automating quality inspection is also important. Such field environments are difficult to address with general-purpose products. Proprietary development is well suited. That is because it requires tailoring to the partner’s equipment specifications. For autonomous processing frameworks, the options organized in LangGraph vs AutoGen vs CrewAI: AI Agent Comparison 2026 are a useful reference. Long-term operation and field integration are important considerations. Regarding device-embedded intelligence, the development in Huawei Announces HarmonyOS 7 with Integrated AI Agent Xiaoyi is also suggestive. Agentic execution on devices could also extend to industrial equipment.

Major Fundraise and Establishment of

Independent Operations

This $100 million is the company’s first outside investment. It was provided by Silversmith Capital Partners. It marks a move from self-funded operation since founding to the use of outside capital. The funds will be used to create and nurture startups. Physical AI projects tend to incur high validation costs. Securing real machines and field environments costs money. On the talent side, both mechanical and computing expertise are needed. A stronger funding base directly enables expansion of the development organization. It creates room for hiring and capital investment. It also increases capacity to develop new partnerships.

The independence of its operating structure was also clarified. In its early days, UP.Labs had a relationship with Up.Partners. There was reportedly no financial tie. It still shares part of the same premises. But as a corporate entity, it is separate. Kuolt explained that point. The relationships among investors, ventures created, and clients have been sorted out. The entry of outside capital also clarifies the governance structure. Corporate clients can more easily structure investment and acquisition terms. Vantora can focus on creating and nurturing ventures. The progress in separating interests is a positive. Going forward, its ability to manage proprietary projects will be tested. It will need to handle confidential information from multiple clients in parallel. Designing information barriers and reuse boundaries will be important. The challenge will be balancing common technology infrastructure with boundaries for client exclusivity.

Editorial Opinion

In the short term, we see proprietary venture-building changing the competition for industrial autonomy contracts. Corporate clients can commission confidential projects without concern over external leakage. Vantora can restart large projects it had previously set aside. In 3-6 months, we expect to see more demonstrations in manufacturing and logistics.

In the long term, we assess that in-house ownership of the intelligence layer will become the dividing line in industrial competition. Equipment retrofits and accumulated operational data will become capitalized as assets. Separating common infrastructure from proprietary parts in the design will become important. Within 1-3 years, we expect to see competing venture builders emerge.

The remaining issues are the boundaries of technology reuse and maintaining independence. How far will horizontal deployment of common components be permitted. Can client exclusivity and business efficiency coexist. The question is how the discipline of outside capital will affect transparency in development.

References

• “A startup that builds other startups raised $100M, and is all-in on physical AI”, by Kirsten Korosec — TechCrunch AI, 2026-09-18T23:25:01.000Z (ARR)
• Source URL: https://techcrunch.com/2026/09/18/a-startup-that-builds-other-startups-raised-100m-and-is-all-in-on-physical-ai/