FOV Venture Partner, David Ripert, breaks down how spatial computing became the foundation for universal robotics, and why this creates unprecedented opportunities for early-stage investors.

David has a diverse background in technology, including roles at Google/YouTube, where he led EMEA production, programming, and creator initiatives, including with XR experimentation. He also co-founded Poplar Studio, a renowned, VC backed AR start-up. And is the President of the UK chapter of the VR/AR Association, among other industry roles.

If you are a founder building in this space we’d love to speak to you! Get in touch here.

Most ‘venture scale’ thinking is still constrained by 16 inch-screens and 105 keys.

At FOV Ventures, we believe that one of the next big, non-consensus opportunities is beyond screens.

Mainframes gave us processing power, desktop internet connected us globally, mobile put the digital world in our pockets. The next era of computing will seamlessly blend the digital world with our physical one.

This is only made possible by the convergence of enabling technologies such as artificial intelligence, computer vision, edge and XR.

For us, with over a decade investing into spatial computing, it makes perfect sense that this extends into robotics. The technologies we've refined for understanding human logic and mapping human spaces are becoming the foundation for machines that can perceive, adapt, and interact with the real world.

While many still view spatial computing through the narrow lens of virtual reality, we believe its potential also lies in the practical application of these “mechanical brains” integrated into physical objects.

This shift is a huge deal - and it’s already happening.

Waymo, Google’s self-driving car division, uses several spatial computing technologies as part of its tech stack, including Light Detection and Ranging (Lidar) for distance measurement, cameras for object recognition, Simultaneous Localization and Mapping (SLAM), AI and ML in the form of computer vision for sensing and edge computing for processing, among others. They have increased paid weekly trips tenfold year-over-year, now reaching 100,000 trips per week.

That’s a staggering milestone. Cars are literally driving themselves, yet most of us are barely paying attention. During a recent trip to San Francisco, I found myself in the backseat of one of these vehicles, living out my childhood science fiction fantasy.

And soon, this won't be limited to Silicon Valley test beds. The UK's Automated Vehicles Act will put self-driving cars on British roads by 2026, marking another moment when life imitates art, as these once-fantastical ideas slip seamlessly into our everyday lives.

But autonomous vehicles are just the beginning. Cars, with over 100 years of using essentially the same shape and function, are a logical first mechanical outlet for this technology. What’s increasingly coming, though, is robots of all shapes and sizes — from humanoids to specialised machines built for every conceivable task for households, the workplace and in industry.

This is where spatial computing intersects seamlessly with robotics. Decades of progress in virtual and augmented reality and artificial intelligence have laid the groundwork for these systems to “see,” interpret, and interact with the world around them. In many ways, the sensory and processing advancements from spatial computing are the connective tissue that will empower robots to function in dynamic, unstructured environments.

The scale of this opportunity dwarfs previous computing paradigms. The global labour market alone represents $15 trillion, and the transformation extends far beyond automation into logistics, maintenance, design, and the very fabric of human society. This convergence will create multiple hundred-billion-dollar companies in spatial computing alone, as these technologies become the foundation for how machines interact with our world.

Having made our first two investments in robotics we’d love to hear from other investors and founders going deep in this space, and we hope the below encapsulates our thinking around why FOV is well placed to pick the winners in this sector.

The Problem: Breaking Free From Structure

Every decade since the 1960s has promised a robotics revolution, yet robots (from factory arms to your dishwasher!) have remained confined to highly structured environments. Despite advanced mechanics and processing power, the core limitation was their inability to understand and adapt to the human world.

Traditional robots relied on what's essentially a sophisticated measuring system - they could detect walls, edges, and distances, but had no deeper understanding of their environment. Think of it like having a detailed 3D blueprint but no understanding of what the rooms are for.

Our world is built for human bodies, human-scale movement, and human vision. Advanced mapping, real-time object recognition, and understanding of how humans use spaces are the foundation for machines that can work alongside us in our homes, offices, and public spaces.

The scale of this transformation dwarfs even the $15 trillion global labor market. When robots become ubiquitous, they'll create entirely new industries: robot logistics networks to transport and deploy them, maintenance hubs, insurance, compliance, security… robot therapists, robo-chef food trucks, and maybe even a franchise of Mos Eisley Cantinas.

Since spatial computing is what makes all of this possible - helping machines understand and operate in our world - we're going to see dozens of billion-dollar companies emerge just in the spatial computing layer alone.

Four Spatial Technologies Helping Robots Understand Our World

You could say that robotics and virtual reality have occupied parallel but distinct technological tracks.

Robotics focused on physical manipulation in structured environments (mostly in the form of robotic arms in factories for the past few decades), while VR pursued perfect digital representation of space and presence (mostly in the form of entertainment for the past few decades).

Yet the fundamental technologies enabling immersive virtual worlds are precisely those needed for robots to understand and interact with our physical reality.

1. VR Teleoperations

While the robotics industry has long used narrow AI for specific tasks like path planning or object recognition, the emergence of foundation models - particularly in computer vision and language understanding - has fundamentally altered how robots learn and adapt. Modern humanoid robots are beginning to walk, talk, and understand their environment with increasing sophistication, they haven't yet achieved full autonomy. Instead, the industry is making progress through a logical stepping stone: teleoperation through VR.

Tesla's Cybercab event showcased impressive demonstrations of their Optimus robots, but what wasn't immediately apparent was that these systems were teleoperated through VR headsets. This approach combines two powerful innovations.

First, VR systems can capture precise human movements using infrared cameras and neural networks, tracking positions with sub-millimeter accuracy while using minimal power. All while using about as much power as an LED bulb.

Second, this same spatial tracking architecture can be adapted for robotic control at a fraction of the traditional cost - what once required $100,000 in custom hardware can now be achieved for a tenth of that amount - or by using off-the-shelf headsets like Meta’s Quest or AVP.

But the real breakthrough isn't just cost reduction - but data collection. VR-driven teleoperation offers a win-win solution: companies can generate revenue today through remote operation while simultaneously collecting rich training data for future AI-modeled autonomy. Every operator session captures nuanced human behaviours, from movement intent to task sequencing, creating datasets in days that would traditionally take months to compile.

Humanoid companies like Tesla (Optimus), Figure, 1X, UBTech, and Skild AI are using this approach to create a systematic path toward autonomy. Their robots start with pure teleoperation, then gradually learn to handle routine subtasks autonomously while maintaining human oversight for complex decisions. Each operation session feeds into machine learning models that learn to replicate human-like decision making.

For startups like Dopl (healthcare) or Teleo (construction), teleoperation solves immediate market needs - enabling remote control over manual processes and providing safety benefits. But the real value lies in their accumulating data. This proprietary dataset becomes a competitive moat that will likely translate into significant enterprise value as larger players race to build autonomous capabilities.

For venture investors, this presents a compelling opportunity. Robotics companies now have the ability to demonstrate clear revenue paths while strengthening their technological advantages with proprietary training data. The most successful companies will combine near-term utility through teleoperation with a systematic approach to achieving full autonomy. This strategy positions them to capture significant value in the trillion-dollar labor automation market, along the way likely to increase M&A activity in the spatial computing space.

2. Simulation: Virtual Training Grounds

Before we see millions of robots in the physical world, we'll have billions of AI agents learning in virtual ones. Game engines like Unreal and Unity - originally built for entertainment - have become sophisticated platforms for robotic training. These engines excel at precisely what robots need, accurate physics simulation, real-time 3D rendering, and complex environmental interactions.

The applications are far-reaching. We’ve written before how digital twins of factories enable robots to learn complex assembly tasks without risking physical equipment and autonomous vehicle systems train in simulated cities, encountering years worth of traffic scenarios in days. Even architectural firms use these platforms to test how robots might navigate buildings before construction begins.

New companies are pushing this technology further. Nunu.ai demonstrates how gaming and robotics naturally converge. Their multimodal AI agents, originally designed to test video games, are now controlling physical robots with minimal adaptation.

Their hypothesis is elegant - if an AI can master navigation and interaction in any virtual environment, those same skills should transfer to physical robots. Early results are exceeding expectations, suggesting that game-playing AI and robotic control are fundamentally linked.

Elsewhere, this approach is attracting major investment. Fei-Fei Li's World Labs, backed by $230 million from a16z and others, is developing "large world models" - AI systems that understand and interact with 3D environments. While their initial applications target game companies and movie studios, the technology has clear implications for robotics. Their models, trained entirely in simulation, are learning to understand physical spaces and interactions in ways that directly translate to robotic control.

The key insight here is that this convergence is creating a virtuous cycle.

Better simulations = smarter robots, which in turn inform more accurate simulations, allowing companies to test and deploy smarter systems faster and at lower cost.

3. Vision Language Action Models (VLAM): Learning from Observation

The latest breakthrough in robot training comes from systems that can learn by watching. DeepMind's RT-2 and RoboCat models demonstrate how robots can learn complex tasks simply by observing videos or photos of humans performing them. As Brett Adcock, who runs Figure, a company building general-purpose humanoid robots, puts it best - “What would have taken us 30 days to code was learned by our neural net in just seven hours” just by watching humans make coffee:

These systems don't just copy - they understand the intent and context behind actions. Open-source VLAM models (e.g. Mistral 7B) with 7 billion parameters are now available, democratising this technology for robotics startups.

‘Democratising’ is buzzword’s buzzword in venture capital but the significance of this being open source is that startups can customise these models for niche applications without the high costs of proprietary AI. This opens the door to innovation in areas previously out of reach - like picking grapes or bricklaying… both billion dollar markets by the way.

4. Simultaneous Localisation and Mapping (SLAM): Spatial Understanding

To deliver compelling VR experiences, devices need to understand the environment instantly and continuously - tracking position, mapping spaces, and identifying surfaces in real-time. Meta's massive investment in solving this problem for VR has effectively subsidised the development of advanced SLAM technology.

The economics of spatial understanding have been transformed by VR's mass-market scale. Depth sensing units that cost $75,000 in 2015 now sell for under $500, driven largely by Meta's demand for precise spatial tracking for Quest devices. Combining SLAM with advances in battery technology (now exceeding 265Wh/kg in production cells), robots can now operate untethered for full 8-hour shifts while maintaining sophisticated environmental awareness.

Through projects like PyTorch3D and Habitat, Meta has open-sourced these capabilities (significance in VLAM section above), effectively subsidising fundamental robotics research.

Meta’s planned Project Orion aims even higher. The Orion glasses demonstrate how SLAM can evolve beyond simple geometric mapping. They can recognise functional spaces, identifying surfaces as workspaces, walkways, or storage areas. This advanced spatial understanding allows for more intuitive and context-aware AR experiences, such as placing virtual objects on appropriate surfaces or providing relevant information based on the user's environment.

But Meta isn’t the only one advancing SLAM; NVIDIA’s Isaac SDK, Google’s Cartographer, and open-source tools like ORB-SLAM3 contribute to a thriving ecosystem. More importantly, for investors, the impact is quantifiable. Robotics deployment costs in warehouses have dropped 60% as robots can now adapt to new spaces without extensive reprogramming. What once required custom hardware costing $100,000s can now be built with commodity sensors and open-source software.

For venture investors, again, the technology is mature enough for rapid deployment, while still early enough for significant innovation in how it's applied.

Market Timing & Catalysts

Much like VR/AR market timing in robotics has always been treacherous - promising technologies repeatedly arrived before their practical enablers. But today, a unique investment window is opening for VC investors.

Because of this, the global robotics market is expected to grow to $200B by 2030 and $1.3T by 2040, with humanoid robotics alone projected between $38B (Goldman Sachs) to a exponentially more bullish $24T (Ark Invest).

While these projections vary widely, they all point to an industry poised for explosive growth and transformative impact across multiple sectors of the global economy.

Historical Parallels Point to Rapid Adoption

The adoption curve for humanoid robots is tracking remarkably similar to early automobiles. Just as it took 17 years (1893-1910) for annual car sales to reach one million units, analysts project a similar 17-year trajectory (2013-2030) for humanoid robots. This pattern suggests we're at the inflection point of exponential growth.

Industry leaders paint an even more ambitious picture. While traditional financial institutions project conservative numbers (Goldman Sachs: 1.4 million units by 2035; Morgan Stanley: 63 million U.S. units by 2050), industry veterans like NVIDIA's Jensen Huang predict humanoid robots will become "as common as cars," and Elon Musk suggests they'll be "10X more common." Figure AI's Brett Adcock envisions a need for about 10 billion humanoid robots from various manufacturers by 2040.

These predictions might seem extraordinary, but so too would today's global automobile numbers have seemed to observers in 1910. Just as cars transformed from luxury items to essential tools for modern life, humanoid robots appear poised for a similar trajectory - driven by demographic necessities rather than just technological possibility.

Demographic Pressure Creates Inelastic Demand

Major economies face an unprecedented demographic crisis. By 2030, the U.S. will have a 25% "dependency ratio" - 25 people over 70 for every 100 working-age individuals. Japan's ratio will be twice as severe at 50%, while Western Europe approaches 35%. China's ratio, currently 20%, is expected to double by 2050.

This demographic pressure manifests in immediate market dynamics. Manufacturing wages have risen 23% since 2020, while warehouse worker turnover exceeds 40% annually. Unlike previous automation waves driven by cost reduction, today's robotics adoption is driven by basic operational necessity. Companies must automate because human workers simply aren't available at any price point.

McKinsey projects that 40% of physical work could be automated through robotics within 20 years, and this transformation is already underway across industries. In manufacturing, Figure's humanoid robots are tackling complex, dexterity-demanding assembly tasks at BMW's Spartanburg plant. PaintJet is automating industrial painting, Avid - commercial cleaning, and Built Robotics is transforming solar panel installation. Some companies are pursuing even bolder strategies - Electric Sheep, for instance, acquires entire landscaping businesses to deploy their robots, capturing the full value chain.

Most significantly, humanoid robot costs are plummeting faster than expected - dropping 40% in just one year versus the predicted 15-20% annual decline. This cost reduction is enabling new business models across industries. Contoro Robotics now offers container unloading at just $250 per container, while Dexory's warehouse systems can scan 10,000 pallets per hour across a million square feet of space. By 2034, Figure projects their robot leasing costs will match a Honda Civic lease at $250/month for a 24/7 worker, making automation accessible to businesses of all sizes.

The evidence from large-scale robotics deployments validates this economic transformation.

Amazon's deployment of over 750,000 robots has achieved a 40% reduction in fulfilment costs compared to non-automated facilities. These robots, leveraging spatial computing advances from VR/AR development, can adapt to different warehouse layouts and product types while maintaining consistent performance.

In self-driving vehicles, Waymo's operations demonstrate the scalability of spatial computing tech in consumer applications. Their rides show 70% gross margins at current scale, with clear paths to 85% as utilisation improves. These metrics prove that robotics companies can build sustainable, high-margin businesses while improving through accumulated operational data.

This unique combination of demographic necessity, tech breakthroughs, and proven economics creates a perfect storm for venture investors looking to back projects with 100x potential that also generate revenue and demonstrate traction today.

The Spatial Computing Advantage

The mass-market adoption of VR and AR technologies has created powerful feedback loops in robotics development. Companies developing better spatial understanding for VR can monetise the same technology in robotics, while improvements in robot perception feed back into better AR experiences. This cross-pollination accelerates development while reducing risk - technologies can be validated in consumer VR before being hardened for robotics applications.

Once again for investors, this creates a unique opportunity. Companies can now deploy robots profitably in unstructured environments while building defensible advantages through accumulated training data.

This means particularly compelling opportunities in:

• Training data platforms that leverage VR/simulation to accelerate robot learning.

• Spatial understanding systems that help robots navigate unstructured environments.

• Tools that enable efficient teleoperation while collecting valuable training data.

• Companies using game engine technology to create sophisticated robot training environments.

• Platforms that help manage and make sense of the massive data streams robots generate.

FOV's decade of experience in spatial computing positions us uniquely to evaluate and support these opportunities. We've seen how technologies developed for VR/AR can transform robotics, and we understand the technical challenges and opportunities at this intersection.

Whether you're building a robotics company leveraging spatial computing, or developing spatial technologies that could accelerate robotics development, we want to hear from you. The robotics revolution is happening - let's build it together.