Yandex Built a Search Box. Then It Put the City Inside It.

The first Yandex index fit on a single server. The company had 25 employees, a name that meant “yet another indexer,” and a thorny language problem. Russian words bend, mutate and carry meaning in their endings, which made literal keyword matching a blunt instrument. Yandex built search around morphology and context. That early decision became a template: start with something locally difficult, apply mathematics, watch what people do, then turn the answer into a service.

A quarter-century later, the answer may arrive as a blue route line, a bag of groceries, a movie suggestion or a car at the curb. Yandex operates more than 100 services. Its search engine remains the front door and its advertising marketplace remains a major profit center, but the company now looks less like a website than a control system for daily life. Search understands the request. Maps understands where it can happen. Go and Eats move people or goods. Market closes a purchase. Plus bundles entertainment. Cloud rents the machinery to businesses. Alice AI is becoming the conversational layer across all of it.

The search box grew legs

Most search companies are paid for helping a user decide. Yandex kept following the user after the decision. A query for dinner can lead to a restaurant listing, a route, a reservation, a delivery order and a payment. A product search can end on Market, with a merchant buying promotion and Yandex logistics handling the parcel. That path from curiosity to completion is the strategic center of the company.

This is why “the Russian Google” is both useful shorthand and an incomplete description. Google explains the search and advertising heritage. It does not capture the dark stores, couriers, ride marketplace, streaming bundle, bank-like products, televisions or delivery robots. A closer comparison is Google, Uber, Amazon, Spotify, DoorDash and a slice of AWS sharing one account and one map.

One spine, many storefronts

The connective tissue is machine learning. Ranking, routing, demand forecasting, ad selection, speech recognition, translation, recommendations and weather prediction are variations on the same organizational skill: make a prediction from messy local signals, test it at scale and improve it quickly. Yandex built MatrixNet, an early proprietary machine-learning method, in 2009. It introduced machine translation in 2011, Alice in 2017, and the YandexGPT and YandexART model families in 2023.

The latest Alice AI is not merely a chat window. Yandex is putting it in Search, Browser, Go, smart speakers and a wearable called Yandex Drops. The ambition is practical: ask the agent to find information, remember a thought, book something or trigger an action elsewhere in the ecosystem. The difference from a standalone assistant is access to services that already know places, inventory, routes and transactions.

“The catalog is large. The feedback loop is the product.”YesPress analysis

Maps offers the cleanest example. In May 2026, Yandex said the service reached 95 million monthly users. Its newest navigation prompts use more than 10,000 visible landmarks, telling a driver to turn after a grocery store instead of relying only on street names and exit numbers. It sounds like a small interface improvement. Underneath sits computer vision, street panoramas, routing logic and an understanding of which object a driver can actually see. Local context becomes software.

Advertising pays. Transactions compound.

Yandex has several businesses because it has several ways to charge. Direct sells targeted advertising and promotion across search and partner inventory. Market, Eats and other platforms collect commissions and service fees from transactions. Go earns from mobility and delivery activity. Plus collects recurring subscriptions for music, film and ecosystem benefits. Cloud and Yandex 360 sell infrastructure, data, AI and collaboration tools. Financial services earn from lending, payments and related products. Smart devices add hardware revenue and give Alice a place in the home.

The mix is moving beyond the old search model. For full-year 2025, Yandex reported revenue of RUB 1.441 trillion, up 32 percent, with adjusted EBITDA of RUB 280.8 billion. In the first half of 2026, group revenue reached RUB 759 billion. City Services produced RUB 426.1 billion before internal eliminations, compared with RUB 260 billion from Search and AI. Search and ads still deliver much stronger margins, but the street-level businesses now carry more revenue.

Each customer enters through a different door. A neighborhood cafe may buy a Direct campaign, update its Maps listing, accept an Eats order and use delivery without ever thinking of itself as a Yandex customer four times over. A retailer can promote inventory, sell it on Market and outsource the difficult last mile. A developer can rent compute, call a speech model or embed MapKit. Large companies buy cloud capacity and workplace software. Consumers pay with attention, transaction fees or a Plus subscription, depending on the moment. The problems are similarly concrete: finding demand, predicting arrival times, matching supply to a street, translating speech, screening spam and keeping a high-load service alive. Competitors can offer any one of these tools. Yandex's pitch is the reduced friction between them.

That creates a useful flywheel. Advertising turns attention into cash. Cash funds services that generate transactions. Transactions make maps, logistics and recommendations smarter. Better services create more attention. The model is difficult to copy because a rival needs not only an app, but also merchants, drivers, couriers, maps, infrastructure and enough repeated usage to train the system.

The unglamorous moat

Yandex's expertise is often most visible in the parts users never see. The company operates tens of thousands of servers and builds for loads that make ordinary software buckle. Some of the resulting tools become public: YDB is a distributed SQL database, YTsaurus handles large-scale data, userver supports high-load C++ services, and YaFSDP helps train large language models. In April 2026, userver 3.0 arrived with more than 2,500 improvements. In June, Yandex released YaFF, a serialization library designed to cut CPU work in busy systems.

Open source is partly recruitment, partly reputation and partly evidence. It shows that the consumer empire rests on deep infrastructure rather than a collection of thin storefronts. The same instinct appears in education. Yandex runs the School of Data Analysis, works with universities, helped establish a computer science faculty with the Higher School of Economics and said more than 25,000 technology professionals completed its programs in 2024. That pipeline matters in a market where specialized engineers are scarce.

A Yandex product often begins where a generic global product becomes awkward: an inflected sentence, a snowy side street, a crowded interchange, a merchant with local inventory.The local-context advantage

A company redrawn in 2024

Any modern profile needs one corporate footnote. In 2024, the former Dutch parent completed a restructuring that separated the Russia-centered Yandex businesses from a group of international ventures. The Yandex brand and the core regional ecosystem moved under International Public Joint-Stock Company Yandex, headquartered in Moscow and traded on the Moscow Exchange as YDEX. The former parent became Nebius Group and retained businesses including an AI cloud operation, Toloka, autonomous-driving company Avride and education company TripleTen.

The split narrowed the present company's geographic center of gravity, even as its product surface kept expanding. Its closest competitors now change by hour: Google or VK for discovery, Ozon and Wildberries for shopping, local platforms for rides, banks for payments, streaming services for attention, cloud providers for corporate workloads and global model makers for AI. Few competitors meet Yandex across all those surfaces. That breadth is an advantage, but it also creates the permanent management problem of every ecosystem: coordination must be worth the complexity.

What people can actually do with it

For a consumer, Yandex reduces the number of handoffs in an ordinary day. Ask Alice a question, plan a route, call a car, order lunch, buy a household item, listen to music and watch a film without repeatedly establishing identity, location or payment. For a small business, the system can supply discovery, advertising, analytics, delivery and marketplace demand. For a large enterprise, Yandex offers cloud infrastructure, workplace software and AI APIs. For developers, there are map kits, speech tools, model access and open-source infrastructure.

The wager is that convenience compounds when the pieces recognize one another. A maps pin is more valuable when it can become a booking. A search result is more valuable when the product can arrive. An assistant is more useful when it can act. Yandex began by understanding how words change. Its present business is built around understanding what a request becomes next.