Le 3DGS s'impose comme la nouvelle norme en matière de production audiovisuelle à l'IBC 2026 : XGRIDS ouvre la voie, de l'innovation à la mise en œuvre sur le terrain

XGRIDS showcased its 3D Gaussian Splatting (3DGS) workflow at IBC 2026, enabling photorealistic real-world locations captured with Lixel K2 and PortalCam to be integrated into Unreal Engine, XR and live-production pipelines. The workflow supports platforms including Pixotope, Aximmetry, Disguise and nDisplay, while South Korea's SBS has already deployed the technology for live news broadcasts. PortalCam also won four NAB Show 2026 Product of the Year awards, and XGRIDS introduced LCC Scan for on-device 3DGS model creation on iPhone.

Analysis

The investable implication is not XGRIDS itself but a potential reduction in the cost and elapsed time of creating reusable virtual sets. If 3D Gaussian Splatting workflows move from specialist VFX use into recurring live production, the economic value accrues first to rendering/GPU infrastructure (NVDA), real-time content engines and production software vendors rather than to capture-device vendors, where hardware commoditization is likely. Near-term revenue impact for listed platforms is immaterial; the relevant signal is whether broadcasters standardize 3DGS assets across multiple programs rather than treating them as isolated demos.

The competitive risk is highest for labor-intensive photogrammetry, traditional 3D modeling and bespoke virtual-set services: faster capture lowers billable reconstruction hours, even as it expands total production volume. Unity (U) could benefit from broader real-time 3D adoption, but its exposure is less direct than Unreal Engine's ecosystem and its execution risk remains company-specific. Autodesk (ADSK) has more defensive exposure through production pipelines, although automated scene capture could pressure parts of modeling demand before it creates incremental asset-management and finishing workloads.

Consensus is likely to overread trade-show adoption as a near-term earnings catalyst. Broadcaster procurement cycles, rights-management requirements, on-air reliability standards and GPU/rendering costs should make adoption a 6-18 month workflow transition, not a 1-3 month revenue event. The thesis is falsified if public broadcasters and production houses do not disclose repeat deployments, or if captured scenes remain too difficult to edit, relight and integrate into established pipelines.

Key Decisions for Investors

• No standalone position on this announcement: impact is too small and the primary beneficiary, Epic Games, is private. Treat subsequent disclosed enterprise deployments as a thematic confirmation rather than an earnings catalyst.
• Add NVDA to a 6-18 month watchlist for spatial-computing production demand, but require evidence of recurring broadcast/production workloads or GPU-cloud spend before initiating; this use case is currently far too small to alter data-center estimates.
• Monitor U versus ADSK as a relative-value research pair over the next 2-3 quarters. A long ADSK / short U structure is only actionable if Unity fails to show real-time production adoption while Autodesk demonstrates resilient media-and-entertainment subscription growth; invalidate on a material Unity enterprise-design win or Autodesk guidance reduction.
• Track CSGP, owner of Matterport, for competitive read-through rather than immediate upside. Low-cost, mobile-first spatial capture could compress premium digital-twin capture pricing; reassess if Matterport reports slower capture growth, lower subscription attach, or increased customer churn over the next two earnings cycles.