Published on

Has Google won the AI race?

Authors
  • avatar
    Name
    Edwin Popham
    Twitter
ThreadsShare on Threads

For most of the last two years, the story of artificial intelligence has been written as a two-horse race between OpenAI and Anthropic, with Google cast as the lumbering incumbent that got caught napping. That story was never quite right, and lately it looks plainly wrong. Google has not won by building the flashiest model. It has won — or is well on the way to winning — by owning the places where ordinary people actually meet AI.

A trophy on a podium

The shift became obvious at WWDC in June 2026, when Apple rebuilt Siri on top of a custom Google Gemini model — a roughly 1.2-trillion-parameter system licensed under a multi-year deal reported at about a billion dollars a year. The new Siri, now in public beta with iOS 27, holds multi-turn conversations, reads what is on your screen, and acts inside third-party apps, drawing on Gemini for the world-knowledge questions Apple's own models can't answer. The Apple logo sits on the front; Google's model does the thinking.

That is the detail that matters. Apple, the company that spent a decade insisting its intelligence would be built in Cupertino, quietly admitted it could not build a competitive foundation model at scale and paid its closest rival to do it instead. When one of the most valuable companies on earth decides your model is good enough to be the brain inside two billion devices, you are not losing the AI race.

Distribution is the moat

The Apple deal is the headline. The quieter story is Chrome. Over the first half of 2026, Google baked Gemini directly into the browser most of the planet already uses — an "Ask Gemini" button in the toolbar, page-aware summarising, tab comparison, and hooks into Gmail, Calendar, Maps and YouTube, all without leaving the page you're on. By mid-July it had reached the UK after rolling out across more than 50 countries.

Stack that on top of Gemini already living inside Google Search's AI Mode, inside Android, inside Workspace, and you see the shape of the advantage. OpenAI built a destination app that people have to choose to visit. Google built AI into the surfaces people were never going to leave. Google doesn't need every user to wake up and pick Gemini as a destination — it can route AI into Search, Android, Workspace and the browser. That is a different growth engine, and it is working: Gemini's share of the consumer AI audience roughly tripled over the past year to around 27%, while ChatGPT, still dominant, slipped from near-monopoly levels down toward the high-40s.

Convenience is an underrated force in technology. The model that is one click away, that already knows your calendar and your inbox, usually beats the marginally smarter model that lives in a separate tab.

The rivals are looking elsewhere

While Google quietly stitched AI into everyday life, OpenAI and Anthropic turned their attention toward the boardroom. Both are now racing to make their models unavoidable inside enterprise technology — cloud partnerships, consulting-style deployment teams, private marketplaces, native software integrations — turning foundation models into operating infrastructure for business. Anthropic has leaned hardest into coding and governed enterprise adoption, powered deeper by Amazon's billions and tight integration into AWS Bedrock.

None of that is a sign of defeat. Enterprise AI is where the serious money is, and the strategy makes sense for two challenger labs that need paying customers to fund their compute bills. But it is a sign of where the two see their clearest path. They are building "everywhere ecosystems" inside companies. Google is building them everywhere, full stop — in the office and out of it.

This is the structural point. Google does not need to court enterprise money the way the independent labs do, because it already has the search-advertising cash machine, the cloud, the device partnerships and the distribution rails. The real divide is simple: Google DeepMind holds the strongest structural position because Gemini is supported by Search, Workspace, Android, Google Cloud and custom infrastructure, while OpenAI has the strongest standalone AI brand and Anthropic the clearest specialist strategy. When you own the pipes, you don't have to win the marketing war to win the market.

Has it actually won, though?

"Won" is always a dangerous word in technology. Google's position is real but not unassailable. Regulators are circling — the new Siri won't even ship in the EU at launch, thanks to the Digital Markets Act. The same antitrust scrutiny that has dogged Google's search-default deals now looms over its AI-default deals. And there is an honesty problem baked in: every Siri query routed to Google's cloud deepens Apple's strategic dependence on a direct competitor, even if Apple's privacy architecture anonymises the requests and bars Google from training on them.

There is also the matter of trust. Convenience wins until it doesn't. A browser assistant that can read the tab you're on and an AI that digests your email raise fair questions about how much of our digital life we are willing to hand a single company in exchange for helpfulness. That bargain has generally held for Google so far. Whether it holds as the stakes move from search results to autonomous agents acting on our behalf is an open question.

So has Google won the AI race? Not the whole thing — there is no single race to win. There is a race for the best model, which is still genuinely contested. There is a race for the enterprise, which OpenAI and Anthropic are fighting hard and well. And there is a race for the consumer, for the everyday moments where ordinary people reach for AI — and on that front, Google has quietly pulled into the lead, not because it built something everyone talks about, but because it built something everyone already uses without having to decide to.

In the end, that may matter more than the benchmarks. The company that owns the defaults rarely loses.

Sources

ThreadsShare on Threads