There is a wall around the world's best map data, and it has been there for twenty years.

Google Maps is the most complete record of physical places ever assembled, and in 2026 it is still not open. You can buy access to parts of it, on Google's terms, at Google's price, within Google's limits. Everybody building anything that touches the real world runs into that wall eventually.

VOYGR has walked up to the same wall, looked at it, and decided to build everything on the other side of it instead.

What they built

VOYGR describes itself as real-world place intelligence for AI apps and agents. The product is aimed at software that needs to know things about physical locations, rather than at a person holding a phone.

The founders make the choice legible. Vlad Baskakov, chief executive, came from Google Maps, where he led product strategy and go-to-market, bootstrapped new API products, shaped data sharing between Maps and Gemini, and worked on merchant experience and offline advertising. He has more than fifteen years across mapping and its main markets, ridesharing and travel, was previously at McKinsey, and holds an MBA from Harvard. Yarik Markov, chief technology officer, has twenty years in technology and more than ten leading infrastructure, machine learning and search at Google, Meta and Apple.

A Maps product lead who worked on the Maps-to-Gemini connection, paired with a search infrastructure engineer from three of the companies that built modern search, is about as precise as founder-market fit gets. These two know exactly where that wall is, because one of them helped build it.

Location data is not an empty field. Google Maps Platform is the default everybody benchmarks against. Mapbox and HERE built independent alternatives, TomTom and Esri serve cars and enterprises, and OpenStreetMap with the Overture Maps Foundation supply the open base layer much of the industry now sits on. Foursquare and SafeGraph specialise in places and footfall, Yelp and Tripadvisor hold the reviews, and OpenTable holds the tables. Every one of them was built for a human looking at a screen, and every one of them competes with the wall rather than complementing it.

The point: build everything the wall does not cover, and sell it to the agents

This is a very smart shape for a company. Here it is step by step.

Start with a service that is closed, valuable, and owned by a giant. Google Maps qualifies on all three. It is not going to open up, and nobody is going to out-build it at its own game, because twenty years of street imagery and billions of user corrections are not something a startup replicates.

Then, instead of attacking it, build everything it does not give you.

The gaps are real and specific. Is there a step at the entrance. Does this place actually take walk-ins after nine. Is the upstairs room genuinely bookable or only on the phone. Is the address the front door or the loading bay. None of that is in any map today, because no human ever needed it written down: a person turns up, looks, and works it out.

And here is the part that makes the whole thing work. Point that missing information specifically at AI agents, because an agent hits the wall exactly where a human does not. A person fills the gap by looking around. An agent cannot, so it needs the fact written down, and whoever writes it down becomes something the agent cannot finish the job without.

So VOYGR is standing at the same wall the agents are stuck at, and getting past it by collecting what nobody bothered to write down. The better the agents get, the more they need it. A cleverer agent still cannot see a step in a doorway.

Why that makes it something worth buying

This is my opinion, not something the company has said.

VOYGR is set up to be bought, and that is a good thing to be.

It will hold the exact data Google lacks, in a market Google cannot afford to lose. Buying VOYGR would be far cheaper and faster for Google than building it, so at some point Google should want to buy it.

There are three buyers here, not one.

Google. Its own AI products need this layer, and its advertising business makes serving machines awkward, so it is slow to build it itself.

The other map companies. For HERE, TomTom or Mapbox, buying VOYGR is the quickest way to become the map that AI agents actually use.

Google again, out of fear. If a rival buys VOYGR, that rival gets the thing that makes agents work on its map instead of Google's. That fear alone can produce an offer.

One buyer who wants you is good. A buyer who also cannot afford to let a competitor have you is much better, because then the price is set by what they lose if they say no.

The formula, which is the part to keep

Take the mapping out and you are left with a formula anyone can copy. That is the real reason this file exists.

Find a closed service owned by a giant that everybody has to work around. Build the complement rather than the competitor, which means everything the giant does not supply and shows no sign of supplying. Aim it at AI agents specifically, because agents hit the walls hardest and cannot improvise their way past them the way people do. Then hold a position the giant either has to buy or has to watch somebody else buy.

Once you look for that shape you find it everywhere a big platform owns the core data and ignores the awkward edges: app stores, payment networks, identity providers, calendars, ticketing, property records. Ask two questions. Does an AI agent get stuck at that wall? And can the missing piece be collected by hand, patiently, by a small team? If both answers are yes, the opening is real.

Why it is bigger than it sounds

The short description is location data for AI. The strategy underneath it is the interesting half.

  • It complements the giant instead of fighting it. Nobody out-builds twenty years of Google Maps, and nobody needs to. The value is in what the wall does not cover.
  • Agents need the facts that humans never wrote down. A person fills the gaps by looking. An agent cannot, so whoever records them becomes part of finishing the job.
  • It ends up wanted by more than one buyer. The giant, its rivals, and the giant again out of fear of the rivals. That is a much stronger position than being useful to one.

What to watch

The measure is completion. Not queries served, but how often an agent using VOYGR finishes the real-world task without a human checking something. That is the only number that distinguishes a reseller of existing place data from a genuinely machine-readable map.

The second is where the unusual facts come from. Anyone can license a places database. The value is the information nobody has written down, so the question is how it gets collected, kept current and verified. That is both the hard part and the thing that makes the position defensible.

The third is who starts depending on them. If the agent platforms and assistant makers begin routing real-world questions through VOYGR, the position described here is real and everything else follows from it.