Enterprise

The Factory Floor Won the Spatial Computing Debate

While consumer headsets piled up in closets, industrial companies quietly built the most consequential deployments in spatial computing history. Digital twins, AR-guided assembly, and VR training have moved past proof of concept and into production.

Mainstay Communications Team · September 4, 2026 · 7 min read

In brief

Industrial spatial computing has crossed from pilot programs into production infrastructure, driven by concrete deployments at companies like PepsiCo, Foxconn, and Boeing. This article examines the technology stack enabling that shift, the exit of Microsoft from enterprise AR hardware, the rise of high-fidelity headsets from Varjo, and the training and digital twin ROI data that is persuading operations leaders to commit real capital.

The Gap Between Demo and Deployment

For most of the past decade, spatial computing demos in industry followed a predictable script. A vendor brought a headset to a trade show floor. Engineers marveled at a digital overlay on a turbine blade. An executive approved a pilot. Then the pilot sat on a shelf.

That pattern has broken. In the past 18 months, a set of credible enterprise deployments has moved the conversation from "can this work in a factory?" to "what's the operational baseline without it?"

The evidence isn't from headset vendors selling optimism. It's from the companies actually running production lines.

The PepsiCo Inflection Point

The clearest signal came at CES 2026 in January, when PepsiCo announced a three-way collaboration with Siemens and NVIDIA to deploy physics-based digital twins across U.S. manufacturing and warehouse facilities. The headline numbers, disclosed by all three companies in a joint press release, were specific: a 20% increase in throughput on initial deployment, 10-15% reductions in capital expenditure, and the ability to identify up to 90% of potential issues before any physical modification takes place.

PepsiCo's Global Chief Strategy and Transformation Officer Athina Kanioura described it as "the first digital blueprint that reimagines how the supply chain is designed, built, and scaled."

The technology stack matters here. PepsiCo is using Siemens' Digital Twin Composer, a software product Siemens CEO Roland Busch introduced at CES 2026, built on NVIDIA Omniverse simulation libraries and fed with real-world engineering data. The Composer is scheduled to reach the Siemens Xcelerator Marketplace in mid-2026. PepsiCo used it to model machines, conveyor systems, pallet flow, and operator paths in 3D before committing to physical changes.

The collaboration represents something structurally different from prior pilots: three companies with actual operating budgets treating digital twins as production infrastructure, not research.

NVIDIA's Industrial Bet

NVIDIA's Omniverse platform has become the substrate for much of this industrial shift. At CES 2025, the company unveiled Mega, an Omniverse Blueprint for building digital twins of robot fleets and factory environments before physical deployment. The target use case: enterprises that want to run hundreds of operational scenarios in simulation before a robot ever touches a shopfloor.

KION Group, which makes supply chain solutions, became an early adopter. Foxconn is using the Mega blueprint to design and optimize a 242,287-square-foot facility in Houston, Texas, where it will manufacture NVIDIA AI infrastructure systems. The facility is being optimized in digital twin form before production starts.

Siemens was named as the first company to develop digital twin software compatible with the Mega blueprint, integrating the capability into its Xcelerator platform.

What's notable is the ecosystem: Accenture has bolted the Mega blueprint into its AI Refinery for Simulation and Robotics. The platform is no longer a point product from a graphics company; it's becoming the simulation layer that industrial software vendors build on top of.

Microsoft Steps Back, the Market Fills In

One of the structural shifts in enterprise XR is the exit of Microsoft from the hardware business. In late 2024, Microsoft ended production of HoloLens 2 and cancelled HoloLens 3. Support for existing devices runs through December 2027. The company didn't exit enterprise spatial computing entirely; its military IVAS contract, built on HoloLens technology, continues. But the ambition of building a general-purpose enterprise AR headset is gone.

The gap hasn't gone unfilled. Varjo, the Finnish hardware maker focused on defense and high-precision industrial applications, launched updated XR-4 Series headsets in October 2025 and followed with integrated, ready-to-deploy XR systems in February 2026. The systems ship with pre-installed software, bundled licenses, and long-term hardware support committed through 2030. Varjo's approach is deliberately narrow: high-fidelity optics for defense simulation, aerospace training, and industrial engineering environments where visual precision isn't optional.

Varjo's December 2025 State of XR in Simulation Training Report, compiled from 106 defense and simulation professionals, found that most organizations surveyed have moved past the experimentation phase and treat XR as a core component of their training ecosystem. Over 56% of respondents were using mixed reality specifically, combining real and virtual elements rather than pure VR immersion.

Training: Where the ROI Is Clearest

The clearest enterprise ROI case remains training, and the strongest independent data comes from a PwC study conducted across 12 U.S. office locations with 1,600 participants. VR learners completed the same training 4 times faster than classroom participants. At 3,000 learners, VR training costs 52% less than classroom equivalents.

The study involved soft skills training, not industrial assembly, so extrapolating directly to factory floors requires care. But the efficiency mechanism, replacing physical props and instructor time with repeatable simulation, applies across both contexts.

Boeing has publicly stated that it can cut training time by 75% per person for complex assembly tasks using VR, citing procedures like cargo door seal installation that involve 50 steps and only a handful of qualified technicians company-wide. That figure comes from Boeing's own communications and has been reported by Light Reading and trade outlets, though Boeing hasn't published a peer-reviewed methodology. The directional logic holds even if the exact number is taken as Boeing's best case rather than a controlled baseline.

PTC's Vuforia platform, which has been in enterprise AR deployment for field service since the early 2020s, remains one of the more widely deployed AR toolsets in industrial maintenance. Vuforia ranked at the top of Quadrant's analysis of AR solution providers for four consecutive years through its most recent rating cycle. PTC positions the platform around technician access to real-time guidance on live equipment, though the company's published ROI figures rely primarily on customer-specific case studies rather than independently audited data.

The Platform Convergence

What's changed structurally is that the enabling platforms are converging. Siemens Xcelerator, NVIDIA Omniverse, Unity and Unreal Engine (the two most cited software environments in Varjo's simulation survey) are becoming common infrastructure rather than competing islands. A manufacturer choosing a digital twin vendor in 2026 can reasonably expect that simulation outputs will move between these environments, that real sensor data will feed back into them, and that the same 3D model used in design review will appear in the AR-guided assembly step and the VR training module.

That interoperability wasn't practical three years ago. The convergence is why companies like PepsiCo can commit to digital twin infrastructure without betting on a single vendor surviving the decade.

What the Enterprise Adoption Wave Looks Like

The broader XR market context: Mordor Intelligence estimates the extended reality market at $10.64 billion in 2026, growing from $7.55 billion in 2025, with enterprise applications driving a significant share of that growth. Those figures come from market research modeling rather than direct audited sales data, and market sizing estimates in the XR category vary significantly by methodology. The directional pattern, enterprise leading consumer, is consistent across research firms even when specific numbers differ.

The companies driving that enterprise share aren't buying XR for inspiration. They're buying it because digital twins catch equipment failures before they become $340,000 shutdowns, because VR training scales complex procedures faster than any classroom alternative, and because simulating a factory upgrade costs a fraction of redesigning it after construction starts.

The demos are done. The operational deployments have started publishing results.

Sources

Companies mentioned

Mainstay Digital · Spatial and 3D visualization for industrial companies

Back to Newsroom