Transcript
In March 2025, Niantic announced it was selling Pokémon Go and all its other games to a mobile gaming company called Scopely for $3.5 billion. That's a significant sum. But the more interesting part of the deal was what Niantic kept. The company retained its map — a proprietary geospatial database built from over 30 billion images — and immediately funded a separate entity called Niantic Spatial with $250 million. The games were valuable. The map was the point all along.
That distinction raises a question worth taking seriously: how did a gaming company end up with one of the largest street-level spatial datasets in the world? The answer runs through one person and spans roughly three decades of geospatial computing, starting with a small startup in the late 1990s, long before anyone had heard of PokéStops or Pokémon.
John Hanke grew up in Cross Plains, a small town in central Texas. By his own account, he wanted to see the world. That impulse turns out to be the literal engine of everything that follows. Hanke co-founded a company called Keyhole, which built something genuinely new at the time: a navigable, three-dimensional digital model of the earth assembled from satellite and aerial imagery. You could fly through it, zooming from orbit to street level. It was, in practice, an interactive globe.
Google acquired Keyhole in 2004. Hanke went with it, and Google relaunched the product as Google Earth, which has since been downloaded by more than one billion users. That number matters. A billion downloads means a billion people who experienced the world as a navigable, zoomable spatial object — many of them for the first time. After the acquisition, Hanke became Vice President of Google's Geo division, which meant he ran Google Maps, Street View, SketchUp, and a handful of related products. He was, for a period, arguably the single person most responsible for how the world was being digitally mapped.
In 2011, still inside Google, Hanke founded a small internal group called Niantic Labs. Niantic spun out as an independent company in 2015. The games it made — first Ingress, then Pokémon Go in 2016 — were understood publicly as augmented reality entertainment. People walked around, caught fictional creatures, captured fictional territories, met in parks. The health angle got significant coverage: Pokémon Go made people walk, and that was real. But there was a structural logic running beneath the entertainment layer that most players never considered.
Ingress, Niantic's first game, was built around locations called portals — real-world landmarks like public art, historic buildings, and notable architecture that players had to physically visit. Players could submit new portal candidates, effectively crowd-auditing the world for interesting, locatable features. When Pokémon Go launched and needed a database of game locations — PokéStops, gyms, places where things happened — Niantic seeded it from the Ingress portal network. Millions of Ingress players had already identified and verified real-world anchor points. That data became the skeleton of Pokémon Go's geography.
The second layer was more technically ambitious. Niantic built an AR scanning feature into its games that allowed players to contribute imagery of real-world locations using their phone cameras. Players would walk around a landmark and record it from multiple angles. The phone's motion sensors logged the precise position and orientation of each frame — what technicians call a posed image, meaning an image tagged with exactly where the camera was and which direction it was pointing. Those posed images, contributed voluntarily by players, fed directly into Niantic's geospatial database. Over time, that database reached 30 billion posed images. Google Street View stitches together imagery from fleets of specially equipped cars. Niantic built something comparable — and in some respects denser — using people who thought they were playing a game.
Before accepting that framing uncritically, it's worth stating the strongest counterargument. Players shared imagery in exchange for in-game rewards. The incentive was transparent enough: scan a PokéStop, receive game items. Crowdsourcing has a long and legitimate history in mapping — OpenStreetMap has been built by volunteers for twenty years, and Wikipedia runs on the same principle. The fact that Niantic extracted commercial value from player contributions doesn't automatically make it exploitative. Players got a free game; Niantic got data. That's a trade.
That argument holds real weight. But there is a specific tension it doesn't resolve, and the 2025 announcement makes that tension harder to ignore. The 30 billion image figure wasn't buried in a technical paper. Niantic put it in its public announcement as evidence of strategic value, framing the entire player base — 100 million annual players, one billion friend connections made, 30 billion miles walked — as inputs into a geospatial intelligence project. The question this raises is not whether the trade was fair in general terms, but whether players understood, at the moment they pointed their phone at a statue and pressed record, that they were contributing training data for a commercial AI model. The Niantic announcements don't address this. There's no discussion of what disclosures were made at the point of collection, or what proportion of players actively contributed scan data versus simply playing the game. That gap sits at the centre of any honest account of what Niantic built.
What is not in dispute is the technical ambition of what the data is being used for. Niantic Spatial describes its Large Geospatial Model as infrastructure for AI more broadly. The argument is that current large language models have a significant weakness in spatial reasoning: they can answer questions, but they don't have a coherent model of how physical space is structured. Niantic Spatial is positioning its dataset as a way to fill that gap, enabling AI systems to understand buildings, navigate interiors, and operate at centimetre-level precision in real environments. The intended applications span manufacturing, logistics, construction, and tourism. The $250 million in capitalisation — $200 million from Niantic's own balance sheet and $50 million contributed by Scopely as part of the deal — reflects that repositioning.
The 2025 split also answers, retrospectively, a question about what kind of company Niantic always was. The games generated revenue, built the user base, and collected the data. When it came time to separate the strategic asset from the revenue-generating product, the map went one way and the games went the other. Scopely got Pokémon Go. Niantic kept the 30 billion images. That ordering tells you what the board considered more valuable in the long run.
Whether that value holds is genuinely open. Niantic Spatial faces serious competition. Google has been building Street View for nearly two decades and has its own substantial aerial and satellite datasets. Apple has been investing heavily in mapping since it broke from Google in 2012. What Niantic has that these competitors lack is specifically interior and pedestrian-level density — imagery contributed by people on foot, in parks, inside plazas, at angles no car-mounted camera would capture. Whether that specific advantage compounds into a durable competitive position, or whether competitors with larger engineering budgets eventually replicate it, is not something the current evidence resolves.
There are also open questions about data flows going forward. Now that the games sit with a different company, it's unclear whether Niantic Spatial retains ongoing access to imagery and location data generated by players. The 30 billion images exist as a historical asset. Whether the database keeps growing at scale depends on contractual arrangements that haven't been made public.
Here is where the evidence lands. Hanke built his career around a single conviction: that a precise, navigable model of the physical world is one of the most valuable things technology can produce. He pursued that through Keyhole, through Google Earth, through Google's geo division, and then through games that made tens of millions of people walk to real places and point cameras at them. The incentive design was effective — probably the most successful large-scale crowdsourced mapping effort ever constructed, precisely because participants experienced it as entertainment rather than labour. The 2025 restructuring confirms that the map was always the core asset.
Whether the players who built that map understood their role in building it is a question the available sources don't answer. Voluntariness depends on understanding what you're agreeing to, and that understanding requires disclosure. As the Large Geospatial Model becomes the foundation for commercial AI applications across multiple industries, that question will carry more weight, not less. Brilliant incentive design and a legitimate privacy concern are not mutually exclusive. Based on what Niantic has disclosed and what it hasn't, the honest conclusion is that both are true at once.