GDC 2025 Executive Reception Recap: The Future of Geospatial Intelligence with AI and AR

April 2, 2025 by Carmen Goldstein

At this year’s GDC, Niantic’s Executive Reception brought together a powerhouse mix of developers, innovators, and thought leaders across gaming and enterprise tech—all under one roof to explore the future of spatial computing. With Scopely’s recent acquisition of Niantic’s games business, the event marked a pivotal moment—not just for beloved real-world titles like Pokémon GO, Monster Hunter Now, and Pikmin Bloom, but also for the accelerating momentum behind the Niantic Spatial Platform, as it steps into a bold new chapter.

The Power Behind Niantic Spatial Platform

Hugh Hayden, Senior Director and GTM lead for Niantic Spatial, kicked off the evening with a warm welcome and a big-picture look at Niantic’s evolution—from game development to building a spatial services platform that’s redefining how we interact with the physical world through AR and AI. Hayden emphasized that the underlying technology—battle-tested and proven at scale through iconic games like Pokémon GO—is now being extended across consumer and enterprise applications, from logistics to remote assistance.

From Google Earth to Real-World Spatial Intelligence

Brian McClendon, Niantic’s SVP of Engineering and a pioneer behind Google Earth and Google Maps, took the stage for a personal and technical deep dive into the evolution of geospatial computing. Drawing on decades of mapping experience, McClendon connected the dots between early satellite imagery and Niantic’s latest innovations in spatial intelligence. “One of the hardest parts of a great AR experience is aligning yourself with the real world,” McClendon told a packed room.

Traditional maps weren't built for real-time alignment

Niantic’s solution: turn the world into the map by capturing it with high enough fidelity for devices to localize themselves.

This breakthrough led to the launch of the Niantic Visual Positioning System (VPS)—a key component of the Niantic Spatial Platform that enables devices to precisely determine their position in the real world. This system now powers experiences in Pokémon GO, Peridot Beyond, Pikmin Bloom, and thousands of third-party apps, all running on the Niantic Map.

Today, the platform operates at global scale:

  • 1 million scans are submitted each week

  • 10 million locations contain usable spatial data

  • 1 million+ VPS locations are fully activated for real-world localization

By fusing AI, AR, and rich geospatial data, Niantic is not just mapping the world—it’s building the infrastructure for the real-world web.

Spatial Computing Building Blocks

As companies race to integrate AI, Niantic holds a unique advantage—its vast geospatial dataset and years of AR experience. “We have some advantages with this data and some of the things we’ve created,” said Brian McClendon, SVP of Engineering.

At the intersection of XR, GIS, and AI lies spatial computing—and Niantic has been quietly assembling its foundation. These core building blocks—Capture, Localize, and Augment—make it possible to recreate, understand, and interact with the real world through digital layers. They’re flexible, scalable, and designed for both enterprise and consumer applications.

Capture: Reconstruct the Real World in Minutes

Niantic offers two powerful ways to capture spatial environments:

  • Scaniverse (iOS + Android): The Scaniverse app lets anyone capture and reconstruct 3D scenes in under a minute. Now with Gaussian Splats, it delivers photorealistic, real-time reconstructions. “It’s the biggest deal since texture mapping in 1991,” said McClendon, showing a striking comparison between traditional 3D tiles and Niantic’s ground-level splats.

  • On-Demand Mapping: Niantic’s on-demand, in-the-field scanning service supports geospatial data capture for areas of any size. Capturing the San Francisco Ferry Building took just 25 minutes. Once scanned, the data can be used to build meshes, generate VPS anchors, and render rich Gaussian Splats.

Localize: Position with Centimeter-Level Accuracy

“Orienting yourself in the real world with respect to real objects is very important,” McClendon noted. That’s where Niantic’s Visual Positioning System (VPS) comes in. Once a location is captured, any compatible device can be localized with centimeter-level precision. This is what enables persistent, shared AR experiences—whether public or private, indoor or outdoor. “The ability to collect large-scale environments and fine-grained detail is a powerful enabler,” he said. VPS currently powers over a million locations and continues to grow through crowd-sourced scans and partner integrations.

Augment: Anchor Digital Content to the Real World

With capture and localization in place, AR content can be placed and persist across users and devices. “This ability to interact—and to keep data anchored consistently for the next person—is very important,” McClendon emphasized. “At this level of resolution, nothing else works.” From whimsical Peridot creatures that remember where they’ve been, to wayfinding overlays and remote assistance markers, Niantic’s platform enables contextual, shared digital experiences layered on top of the physical world.

The Rise of the Large Geospatial Model

McClendon introduced Niantic’s next frontier: the Large Geospatial Model—a spatial AI framework designed to both localize and interpret the world. It’s like a GPT for geospatial understanding, trained not on text and images, but on scans, splats, LiDAR, drone imagery, and VPS anchors. “We are solving problems where a computer that ‘walks’ down a hallway one way can figure out how to make its way back—even without seeing it from the right angle,” said McClendon.

The LGM leverages:

  • User-generated scans from Pokémon GO, Ingress, and Scaniverse

  • Advanced VPS data, including 50 million neural networks

  • 150 trillion parameters powering location intelligence

  • Input from LiDAR, drone data, and GIS systems

Unlike LLMs (optimized for language) or synthetic world generators (used in robotics), Niantic’s model extends real-world maps beyond what the data directly shows. It can infer what’s around the corner, even when no picture exists—making it invaluable for everything from AR navigation to industrial robotics. “It can see not just what the map is seeing, but more of the world around it,” McClendon explained.

Live Demos and Real-World Use Cases

Throughout the evening, guests explored a range of live demos showcasing how Niantic’s spatial technology is bridging consumer and enterprise use cases. Highlights included Niantic’s delivery verification prototype, Sphere’s remote collaboration app featuring a robot arm, Niantic’s own multi-user remote collaboration app, and Peridot running on both Snap Spectacles and Meta Quest 3.

On the Niantic Spatial for Games panel, leaders from Pokémon GO, Peridot, and Reality Crisis shared how they’re integrating spatial services to enhance gameplay with real-world context and precision. The conversation also looked ahead to how these same technologies are unlocking new frontiers in AR commerce, logistics, and industrial training.

What’s Next for Niantic Spatial

Niantic isn’t just using AI to enhance AR—it’s rethinking how machines perceive and interact with the physical world. “Early on, we saw the ability to not just localize yourself, but recognize what’s near you,” said Brian McClendon.

That insight led to the integration of Large Language Models (LLMs) into gameplay, like in Peridot, where AI-driven characters don’t just react—they understand their surroundings and generate actions based on real-world context. “It made us even more excited about the future when combined with maps,” McClendon continued. “Imagine a world where we have the meshes, the splats, and the semantics of the real world—then a model can understand things far beyond what’s visible.”

As Niantic prepares to spin out its geospatial AI business as a standalone company—Niantic Spatial, Inc.—the message from GDC 2025 was clear: the next era of computing won’t be confined to screens. It will be rooted in place, powered by maps, and brought to life by AI.

With its Large Geospatial Model, advanced VPS, and a rapidly growing pool of 3D spatial data, Niantic Spatial is unlocking a new kind of map—one that thinks, learns, and adapts. This isn’t just about helping devices navigate the world. It’s about building a real-world intelligence layer that empowers developers, creators, and enterprises to build context-aware, persistent, and immersive experiences.

Whether you’re building games, enterprise tools, or next-gen AR applications—Niantic Spatial wants to be your partner in anchoring those experiences to the real world.