NEW YORK – Sundar Pichai had a number to offer and a product name to go with it. Speaking at Google’s second-quarter earnings call, the chief executive confirmed that Gemini now counts nearly 950 million monthly active users – a figure that has tripled year-on-year – and announced that Gemini Spark, the company’s personalized AI agent, is now available in the United States and internationally. The milestone puts Gemini in the company of Google Search, Maps, and YouTube as one of the small group of products the company has built to near-billion-user scale.
Gemini Spark is not a new model or a new interface. It is a mode: a version of Gemini that learns from a user’s calendar, preferences, and existing Google product history and acts on that information without waiting to be asked. Pichai described it as a “personalized agent” that surfaces relevant information and takes action in context – anticipating a travel update before departure, suggesting a follow-up email after a meeting ends, or organizing a morning briefing around the week ahead. That scope, if delivered consistently, puts Spark in different territory than a search bar with a text field.
Getting to this point took longer than Google’s early Gemini messaging suggested. When the product launched it was positioned primarily as a better answer to ChatGPT – a conversational AI from a company with more proprietary data and greater compute capacity. The premise was true in parts, but the usage numbers did not follow immediately. What eventually moved the needle was distribution. Alphabet committed a record $205 billion in capital expenditure for 2026 to sustain the AI infrastructure build-out, and integrated Gemini into Android, Chrome, Google Docs, and Gmail across the existing product fleet. The monthly active user count tracked the integrations.
That integration strategy extended to hardware this week. Samsung unveiled Galaxy Glasses at Galaxy Unpacked in London, built on Android XR with Gemini AI powering live translation and contextual overlay functions. The product illustrates how Google has shifted from presenting Gemini as a standalone destination to embedding it as a layer that surfaces wherever users already are – phone, browser, document editor, or wearable. At 950 million monthly users, that distribution model has worked.
The competitive motivation for the Spark rollout is legible. OpenAI has been developing and testing agentic functionality for ChatGPT, and Apple has announced a more capable on-device Siri with assistant-grade features – though neither has shipped at the scale Google has now claimed. Pichai’s decision to name and globally deploy Gemini Spark this week, ahead of those releases reaching critical mass, is a deliberate move to establish a baseline expectation for what an AI agent should look like in practice. TechCrunch reported that Gemini is now closing in on becoming Google’s twelfth product to reach a billion users, a list that has historically been built over years rather than months.
What Spark costs – or whether it requires a Google One AI Premium subscription – is not fully resolved in Google’s public disclosures. The company has maintained a tiered Gemini model, with the more capable functions reserved for paying subscribers, but the specific tier at which Spark activates has not been stated clearly. At 950 million monthly active users, the question of how many are paying anything at all is the number that matters most to the company’s AI revenue picture.
Gemini’s value proposition in Spark mode depends substantially on access to Google’s existing data infrastructure. The agent’s ability to anticipate needs rather than respond to queries rests on what it knows about a user’s patterns – calendar data, email history, location signals – which is precisely the data Google has accumulated over two decades across its core products. That is a structural advantage over AI-first competitors that lack that depth of behavioral data. It is also the reason European and US regulators have been monitoring Google’s AI product strategy closely, as the same data infrastructure that enables personalization is the one that regulators have repeatedly scrutinized for competitive foreclosure.
Google Cloud, the enterprise arm, has been the cleaner revenue story. The unit grew 82 percent year-on-year in Q2, driven by AI infrastructure spending from enterprise customers building on Google’s Vertex AI platform. Consumer AI has been harder to price: the core Gemini app is free at entry level, the platform is bundled into products people already use, and the conversion rate from free engagement to paid subscription has not been publicly disclosed. Google’s investor narrative depends on the argument that scale now produces subscription revenue and advertising inventory later.
Whether 950 million monthly active users is the right way to count what Google has built is a reasonable question. Not all of those users chose Gemini as a product – many encountered it because it appeared inside Gmail or Chrome or their Android keyboard. A user who dismissed a Gemini suggestion in a Google Doc is statistically an MAU. That does not invalidate the number, but it does complicate the claim that nearly a billion people have embraced a Google AI agent in any meaningful sense.
What the scale does establish is a distribution lead that AI-first competitors cannot easily replicate. OpenAI’s path to hundreds of millions of users required original downloads and paid conversions. Google’s path ran through the default apps people already had. Spark is Google’s attempt to convert that passive reach into active utility – an agent that earns its place through usefulness rather than just appearing because it was installed. Whether those 950 million encounters become 950 million habits is what the next year of Gemini’s story will measure.
