Yes, the next Nvidia in robotics could very well be a robotics systems software firm. Just as Nvidia controls the computational substrate that powers modern AI—every major language model, data center, and GPU-dependent application runs on Nvidia hardware—a dominant robotics software platform could become equally indispensable. The difference is that in robotics, the software layer controls not just processing but the entire decision-making architecture of machines that need to perceive, learn, and act in the physical world. When Figure AI raised $1 billion at a $39 billion valuation in September 2024, it wasn’t primarily a hardware play; it was a bet on the company’s ability to build the software stack that would eventually power millions of autonomous machines.
That pattern—betting on the software layer as the true moat—is repeating across the industry. The robotics software opportunity mirrors Nvidia’s dominance, but with a critical twist. Nvidia became valuable because it owned the platform layer that everyone else built on top of. A robotics systems software firm could achieve similar dominance by owning the real-world learning framework, the world models, the robot control systems, and the data pipelines that allow machines to improve over time. The venture capital markets are already signaling this: global robotics funding reached nearly $14 billion in 2025, up 70% from 2024, and specialized robotics startups captured over 70% of the capital flowing into the sector in Q1 2025 alone.
Table of Contents
- How Software Platforms Become More Valuable Than the Hardware They Run On
- The Funding Evidence Points to Software-First Startups, Not Hardware-Centric Ones
- Nvidia’s Current Robotics Strategy and Where It Has Gaps
- Why Vertical Robotics Creates Software Moats That Hardware Alone Cannot
- The Data and Learning Problem That Software Companies Must Solve
- Real Examples of Software Firms Reaching Nvidia-Scale Valuations in Real Time
- The Path Forward: What Would Make a Robotics Software Firm the Next Nvidia
- Conclusion
How Software Platforms Become More Valuable Than the Hardware They Run On
The pattern is well-established in technology: the platform layer eventually becomes more valuable than the devices it powers. Microsoft didn’t build the most powerful computers; it built the operating system everyone needed. Apple didn’t invent the smartphone; it created the ecosystem. nvidia accelerated this pattern by controlling the computational layer that AI systems depend on. In robotics, the equivalent would be a software firm that provides the fundamental stack—the vision systems, the learning algorithms, the control frameworks, and the operational systems—that robotics companies license or build upon. What makes robotics different is that the software firm doesn’t just control computation; it controls adaptation.
A robot operating in the real world needs to constantly adjust to variations in lighting, surfaces, obstacles, and task variations that training data never fully captured. The software that handles this adaptation—learning from real-world failures, updating models based on live telemetry, orchestrating fleets—becomes as essential as the motors and sensors. Companies like Physical Intelligence, which hit a $2.8 billion valuation in its first year of operation, are explicitly building this layer: general-purpose robot learning systems that any hardware manufacturer could theoretically license and integrate. The limitation to keep in mind: software-only dominance assumes hardware commoditization, which hasn’t fully happened in robotics yet. Unlike GPUs, robot hardware remains highly specialized by use case. An industrial arm requires different hardware than a humanoid that requires different hardware than a mobile manipulator. This fragmentation means software firms still need partnerships or deep hardware integration to prove their value, making them more vulnerable to being captured by vertically integrated hardware companies than Nvidia ever was.
The Funding Evidence Points to Software-First Startups, Not Hardware-Centric Ones
The venture capital allocation tells a clear story. Skild AI, which builds reinforcement learning software for robots, saw its valuation triple to over $14 billion in a single funding round announced in January 2025. Apptronik raised $935 million for its Apollo humanoid robot, but the real money went into companies solving the software problem: how to make robots reliably perform tasks without being hand-programmed. Even Figure AI’s $39 billion valuation came from investors betting that its control software and learning systems would be more valuable than its robot hardware over time. In Q1 2025 alone, global robotics funding totaled $2.26 billion, with specialized robotics startups accounting for over 70% of that capital. This shift toward vertical robotics—companies that integrate software, hardware, and AI for specific use cases—represents investors moving away from the idea that a single platform (like ROS or Nvidia’s Isaac) will dominate.
Instead, they’re betting on specialized software stacks that solve specific problems: warehouse automation, surgical robotics, autonomous mobile manipulation, and so on. Each vertical has the potential to develop its own “Nvidia equivalent”—a software platform that becomes so central to that industry that manufacturers have to integrate with it. The warning here is about ecosystem lock-in. A software firm could become dominant in one robotics vertical while remaining irrelevant in another. Nvidia avoided this by controlling the fundamental layer—computation—that all verticals need. A robotics software firm might need to dominate multiple verticals to achieve Nvidia-scale value, which is a harder problem. A company that wins in autonomous warehouses might find itself irrelevant to surgical robots or humanoid development, creating a fragmented landscape rather than a concentrated one.
Nvidia’s Current Robotics Strategy and Where It Has Gaps
Nvidia isn’t passive in robotics. The company has released Isaac, its robotics simulation and development platform, and GR00T, a foundation model for robot learning. Major players—Agility Robotics, Boston Dynamics, Figure AI, and Skild AI—are actively adopting these tools. Nvidia CEO Jensen Huang stated at GTC 2026 that “every industrial company will become a robotics company,” explicitly positioning Nvidia as the foundational layer. This is the same playbook that won Nvidia the AI compute market: own the infrastructure, and collect rents from everyone building on top. But Nvidia’s robotics strategy has a fundamental gap. A software platform that specializes in one robot type or one industry vertical can optimize far more deeply than a general-purpose compute platform.
Nvidia provides the foundation; a specialized software firm provides the domain expertise. If you’re building a humanoid robot optimized for warehouse work, you might use Nvidia’s hardware and Isaac simulation, but you also need software that understands the specific control problems, learning curves, and failure modes of humanoid movement in that environment. That domain-specific layer is where value concentrates, and it’s where competitors to Nvidia can build moats that GPUs alone don’t provide. The catch is that Nvidia’s size and resources are staggering. The company could acquire a successful robotics software startup or build its own domain expertise faster than the startup could scale. Nvidia’s bet is that it owns the bottom layer and can therefore own pieces of every layer above it. A software startup’s counter-bet has to be that specialization and focus create advantages that scale can’t overcome—at least not quickly enough.
Why Vertical Robotics Creates Software Moats That Hardware Alone Cannot
The robotics industry is shifting toward vertical robotics—specialized companies that own the entire stack for a specific use case. A company optimizing for surgical robots doesn’t need general-purpose hardware; it needs hardware designed for precision, with software designed to interpret surgical images, predict tissue response, and handle the liability and regulatory requirements of medical devices. The software here becomes so entwined with the hardware that they’re almost inseparable. This creates a moat that a general-purpose compute company like Nvidia struggles to own. Consider Figure AI’s strategy: the company is building humanoid robots specifically for industrial labor. To win, Figure needs software that understands factory environments, that learns from human demonstrations, that optimizes for safety and efficiency in that context. Investors valued Figure at $39 billion because they believed the software—the learning systems, the task planning, the continuous improvement—would be more valuable than the hardware over time.
The hardware becomes commoditized; the software becomes the differentiator. This is exactly how Nvidia became dominant in AI compute: the chips were table stakes, but the software ecosystem and the compute performance became inseparable advantages. The tradeoff is that vertical robotics is inherently smaller in addressable market than horizontal infrastructure. Nvidia’s GPU platform covers all of AI, all of gaming, all of scientific computing. A surgical robotics software company covers only the surgical robotics market. A company could become the “Nvidia of surgical robotics” and still be significantly smaller than Nvidia. To become the “next Nvidia” in robotics broadly, a software firm would need to either dominate multiple verticals or own a horizontal layer—like learning systems or data pipelines—that all verticals depend on.
The Data and Learning Problem That Software Companies Must Solve
The most valuable robotics software firm won’t be the one with the most sophisticated control algorithms; it will be the one that builds the best learning systems from real-world robot experience. This is Nvidia’s core competency in AI: the company doesn’t just make GPUs; it makes the ecosystem that collects, processes, and learns from data at scale. A robotics software firm would need to solve the equivalent problem: how to aggregate and learn from terabytes of robot telemetry, failures, and successes happening in thousands of real-world environments. This is why companies like Physical Intelligence, Skild AI, and Figure are investing heavily in data collection and foundation models. They’re trying to become the platform that robots learn from collectively rather than in isolation. The robot learning that happens at one facility improves all robots on the platform.
This creates a compounding advantage: as more robots join the platform, the data gets richer, the models get better, and robots become more capable. Eventually, no individual competitor can match the performance because no competitor has access to the same volume and diversity of learning data. The warning is that this advantage can be disrupted by scale. If a large robotics manufacturer—say, ABB or FANUC—collects billions of robot-hours of data from their existing installed base, they suddenly have more learning data than any software startup. Alternatively, if Nvidia can incentivize robotics companies to run all their learning on Nvidia hardware, Nvidia could collect the same cross-company data that a specialized software firm hopes to monopolize. The winner in this space will likely be the company that convinces the most diverse robotics manufacturers to share data through its platform.
Real Examples of Software Firms Reaching Nvidia-Scale Valuations in Real Time
We’re watching this happen in real time. Physical Intelligence achieved a $2.8 billion valuation in its first year of operation, while remaining primarily a software company. The bet on Physical Intelligence isn’t about robot hardware; it’s about the company’s approach to robot learning—specifically, its approach to general-purpose robot control models that work across different hardware platforms.
If Physical Intelligence can make its learning system reliable and versatile enough, it could become to robotics what OpenAI became to large language models: a software platform that everyone else builds on top of. Skild AI’s valuation surge to over $14 billion in January 2025 signals that investors see robotics software as a potentially trillion-dollar market. The company’s focus on learning and control systems for various robot types indicates a strategy similar to Physical Intelligence’s: own the learning layer, remain hardware-agnostic, and license the platform to manufacturers. Both companies are explicitly trying to replicate Nvidia’s playbook—own the foundational layer that everyone else depends on—but in the software stack rather than the hardware stack.
The Path Forward: What Would Make a Robotics Software Firm the Next Nvidia
For a robotics software firm to achieve Nvidia-scale dominance, three conditions need to align. First, the firm needs to own a layer of the stack that every robotics company—across all verticals—depends on. This could be learning systems, simulation environments, data pipelines, or control frameworks. Second, the firm needs to maintain hardware neutrality while being deeply integrated with hardware partners, just as Nvidia remains agnostic to AI frameworks while being essential to all of them. Third, the firm needs to create network effects where the value of the platform increases as more robotics companies join it.
The timeline matters too. We’re likely 5-10 years away from robotics software consolidation around dominant platforms. The market is still fragmented, with many startups pursuing different approaches. But once one software platform achieves enough adoption, the compounding advantage of better learning data and broader hardware integration could create a Nvidia-like moat that’s extremely difficult to dislodge. The question isn’t whether such a firm will emerge, but whether it will be a specialized startup like Physical Intelligence or Skild AI, or whether it will be a large technology company like Nvidia, Google, or Tesla that already has resources and relationships in place.
Conclusion
The next Nvidia in robotics is likely to be a robotics systems software firm because software—not hardware—controls the learning, adaptation, and continuous improvement that makes robots valuable over time. The venture capital market is already pricing in this possibility, with specialized robotics software startups reaching billion-dollar valuations. Companies like Physical Intelligence and Skild AI are explicitly building the platforms that hardware manufacturers will need to license, much as every AI company needs Nvidia’s GPUs. The software firm that owns the learning layer, maintains hardware neutrality, and builds network effects across diverse robot types will have the scale and leverage to command the market the way Nvidia does today.
The path forward is uncertain, but the destination is clear. In a world where “every industrial company will become a robotics company,” as Nvidia’s CEO put it, the infrastructure that enables this transition will be worth trillions. That infrastructure is increasingly software, not silicon. The question is which software firm will own it.
