Jeff Cardenas, CEO and Co-Founder of Apptronik

Apptronik stated it is pursuing a broad vision of humanoid robotics in which versatility, and not optimization around one fixed task, is the defining advantage. The company believes a single robot eventually needs to perform hundreds or thousands of tasks, allowing customers to shift the same asset between workflows as demand changes. The company argues this versatility is particularly compelling in the logistics vertical, where inbound and outbound activity can peak at different times of day and dedicated automation often produces unattractive utilization. Apptronik believes its modular approach separates mobility from manipulation, and that emphasizing two-armed manipulation is the most versatile architecture. In effect, the company explained that when it is attempting to put all these pieces together to build the “PC of robots”: a mass-manufacturable platform whose functionality expands through software and interchangeable configurations, rather than through consistently needed to re-deploy new hardware versions.

An immediate priority, according to Apptronik, is data and deployment. The company shared that Apptronik’s Robot Park is intended to function as a data factory, combining robot-generated data, egocentric human data and synthetic data to train and validate future systems. Customer programs with Mercedes-Benz (MBGYY, Not Rated) and GXO (GXO, Not Rated) were highlighted in the presentation as iterative learning processes, putting robots into real operating environments that should strengthen the deployment-data flywheel. The company believes collaboration with Google DeepMind is central to this strategy, linking foundation-model development with purpose-built hardware. Management also argued that its “show, not tell” posture reflects the depth of its engineering work (including more than 20 robot designs and roughly 80 actuator iterations) but the company acknowledges that Apollo 3 must ultimately demonstrate that these advances translate into manufacturability, reliability and useful out-of-the-box intelligence.

On deployment, management explained that they expect it to proceed in stages over time: warehouses and logistics first, followed by customer-facing settings such as healthcare and retail, and eventually homes and assistive care. This is because this sequencing appropriately prioritizes structured environments where the economics and workflows are easiest to prove. However, management argued longer-term urgency is due to demographic challenges: declining birth rates and shortages of caregivers could make assistive robotics a necessity rather than a convenience. Apptronik acknowledges the competition, and notes it is fighting to scale manufacturing and compress the transition from pilots to recurring commercial fleets before competing platforms establish their own data advantages.

David Golden and Leif Jentoft, Co-Founder and Head of AI Solutions at Standard Bots

In the conversation with Standard Bots, the company shared it is targeting what it sees as the largest practical barrier to industrial automation: it believes adoption has been limited because robots remain too difficult and expensive for most manufacturers to deploy. The company contends that adoption is concentrated among large automotive companies that can spend millions of dollars and months integrating a robot around a single repetitive process (such as the automotive industry). Outside those environments, Standard Bots believes penetration remains low because the integration burden overwhelms the potential return. The company argues its primary business proposition is to remove the “PhD” from deployment by allowing operators to configure robots through intuitive software (including conversational AI and an iPad-style interface) rather than specialized programming and custom engineering.

This focus makes deployment particularly urgent, according to the company. It has approximately 150 robots in service and is scaling operations in Ohio, while also operating a production facility in Glen Cove, New York. They believe the use of off-the-shelf hardware should reduce supply-chain risk and allow the company to concentrate investment on software, usability and application integration. Reliability nevertheless remains non-negotiable for the management team: they state an easy-to-program robot has limited value if downtime stops a customer’s entire production line. Standard Bots therefore stated its balancing rapid fleet growth with the need to establish industrial-grade dependability.

This led the company to explain its strategic advantage is less the robotic arm itself than the integration and data layers connecting the machine to the end application; deployments across small and mid-sized businesses may be especially valuable, according to the company, because they expose the platform to a broader variety of workflows than highly standardized automotive plants. It believes each customer installation contributes data that can improve models and product design, making the installed base increasingly difficult to replicate. The company's conclusion is that this creates a strong incentive to deploy now: the vendor that makes robotics accessible to bakers, machine shops and other non-expert users can build a differentiated data asset before humanoids or lower-cost foreign competitors broaden their reach.

Tao Chen, Co-Founder and CEO of Dexmate

Dexmate’s core argument is straightforward: a robot that can operate close to 24/7 can produce compelling economics even before fully general-purpose autonomy arrives. The company explained how it believes that seemingly small design choices (including camera positioning, actuator geometry and data infrastructure) can materially affect whether models can interpret the environment and transfer successfully into production, and that hardware with AI requirements incorporated from the outset rather than adding perception and intelligence to a legacy machine make for more efficient hardware. The company also explained that demand from frontier AI laboratories also provides technical validation and exposes Dexmate to advanced model-development workflows.

The discussion made it clear that it is not intent on winning the foundation-model race itself. Instead, the company wants to make existing AI models accessible and useful through integrated hardware and software. They see this as a pragmatic allocation of capital because model capabilities are advancing rapidly across the ecosystem, while installation, training, workflow adaptation and user experience remain persistent customer bottlenecks. Dexmate’s argued that its differentiation must therefore come from faster integration, easier operation and lower total cost rather than from claiming the most advanced proprietary model.

Manufacturing and logistics represent the most immediate deployment opportunities, according to Dexmate, because customers can quantify labor savings and utilization easily, while the company contends that hospitals and hotels may follow more slowly given their less structured environments and higher sensitivity to service interruptions. The company also noted that requirements also differ materially among large customers such as electronics manufacturers, and robot training is not yet universal across platforms or applications. Dexmate’s near-term imperative is to reduce the time between installation and productive utility; if that friction remains high, the company believes lower-priced Asian hardware could capture the market despite weaker software integration.

Hamza Derbas, Founder and CEO of Maven Robotics

We understand that Maven Robotics is building a vertically integrated industrial robotics system spanning hardware, software and AI. The company notes its initial focus is work that remains predominantly manual despite decades of factory automation: moving containers, handling totes, mixed-case palletizing and supplying production lines. It believes these workflows require more than a repetitive pick-and-place motion; robots must manage loose items, change tools, recover from anomalies and coordinate across long sequences. Maven shared its thinking on productization is that wheeled, dual-arm architecture and quick-change end effectors attempt to combine versatility with higher throughput than a fully generalized humanoid. The platform is reportedly capable of carrying roughly 30 kilograms and changing tools to open, manipulate and precisely stack different items.

Importantly, management does not claim that one model can perform every industrial task today. Maven instead explained it occupies a middle ground: progressing toward generalization while delivering economic value in defined workflows now. Its software reportedly reduces implementation from months of custom programming to a much shorter configuration process, and learning from one robot can propagate across the fleet. Acccording to the company this is critical because customer demand appears to exceed current supply. Maven reported it has five customers, including several that could eventually support fleets of thousands, and expects deployments across roughly 10 sites with multiple robots per sit, and management emphasized that these are intended as production contracts (not pilots or proofs of concept) which raises the urgency around manufacturing capacity and field reliability.

The company noted its capacity is currently modest at roughly 200–300 robots annually, but management outlined a modular factory model capable of producing approximately 2,500 robots per line, with an ambition to reach 25,000–30,000 deployed units in 2028. It contended that public market estimates broadly support an approximately $80–$100 billion opportunity for tote handling and palletizing, with broader assembly, fabrication and industrial-material-handling applications potentially adding several hundred billion dollars. The immediate constraint, in Maven's view, is therefore supply and execution, not market imagination: the company highlighted that customers facing labor shortages or existential pressure to automate want systems now, and any delay gives competitors time to capture deployments, operational data and recurring revenue. Elaborating on this shortage, Derbas noted that the Manufacturing Institute (MI) anticipates 4M job openings for manufacturing anticipated by 2030, with an anticipated 2.1M shortfall, and Deloitte worked with MI collaboratively to note an anticipated $1Tr in economic cost for shortfalls that need to be solved.

Scott Smith, VP of Finance at Path Robotics

Path Robotics notes it selected welding as its initial beachhead because it combines high labor cost, scarce skills and direct consequences for manufacturing throughput. The company found that welding is frequently a critical production bottleneck, and workforce availability is deteriorating; indeed, the American Welding Society estimates that roughly 320,500 new welders will be needed by 2029. Path believes its installed base of approximately 50 customers across multiple industries suggests that the value proposition is transferable, while successful economics in a relatively lower-cost manufacturing market such as Mexico would indicate even stronger potential in higher-wage regions.

The company believes its Robotics-as-a-Service model is central to accelerating adoption. Rather than requiring customers to justify a large upfront capital purchase and wait for a conventional payback period, Path aims to deliver operating-expense savings from the start and contract around annual productivity. This reduces the friction associated with automation budgets and shifts some performance risk from the customer to Path. The company deploys its own integrated hardware, with the company highlighting this is rather than relying exclusively on installed equipment, which it believes gives it greater control over performance but also creates capital, manufacturing and field-service requirements.

With welding economics established, Path postulated on expanding into the next logical steps: upstream assembly and fit-up (the work required to prepare components for welding). They believe that expansion could increase wallet share and allow Path to automate a larger portion of the production flow, while mobile platforms such as quadrupeds may extend welding into outdoor or difficult-to-access environments. Shipbuilding is a particularly compelling prospective market, according to the company, because welding intensity and constrained domestic capacity create an urgent need for productivity. The key execution risk that Path sees is if it can broaden applications without losing the high-return, repeatable deployment model that made welding an attractive initial use case.

Lindon Gao, CEO and Co-Founder of Dyna Robotics

Dyna Robotics is focused on dexterous, commercially useful workflows in laundromats, hotels, restaurants and gyms. Dyna stated these markets are attractive because textile and food-service tasks remain highly manual, repetitive and difficult for traditional automation. Dyna’s deployments include napkin folding for Din Tai Fung, where the customer is reportedly requesting additional robots, providing an important signal that the system can generate operational value rather than simply perform a controlled demonstration. The broader opportunity, according to Dyna, is to expand from an individual task (such as folding) to an end-to-end workflow covering handling, loading, unloading and sorting.

Dyna sees the principal gating factor for adoption being reliability. Gao believes customer acceptance in the 80%–90% range may be encouraging for an emerging system, but it is inadequate for unattended, scaled operation. Dyna’s goal of approximately 99.9% reliability is therefore not an incremental engineering target; it is the threshold between an assisted product and scalable automation. Extending demonstrations from one continuous hour (which Dyna has a posted video of) to 24 hours should expose the edge cases that matter in production, including dropped items, variable fabrics and the cumulative effect of small errors throughout a workflow. As management succinctly frames it, deployment itself is the most meaningful evaluation.

Dyna explained that its Taku hardware was developed internally to align the body, mobility and end effectors with these workflows. It believes a human-compatible form factor offers access to environments designed around people and benefits from abundant human task data, although that does not necessarily require a fully bipedal humanoid for every application. The urgency is clear, per Dyna's commentary: existing customers are asking for more systems, but scaling ahead of 99.9% end-to-end performance risks increasing support costs and damaging trust, in the company's view. Dyna states it must therefore expand deployments quickly enough to collect data while maintaining the balancing act of enough operational control to close the final reliability gap and maintain customer trust.