Google launches Gemini Spark, a cloud AI agent, for $100 a month

Google has unveiled Gemini Spark, a cloud AI agent available in the premium Google AI Ultra plan. Unlike OpenAI Operator, Spark runs in the cloud and can carry out tasks around the clock without a human involved.

How Spark works

Gemini Spark is a cloud agent built on Gemini 2.0, Google's latest generation of models. It sees the screen, clicks buttons, fills in forms, and can even stay active in the background while waiting for results. For example, the agent can track a ticket price and buy the ticket when it drops, or monitor stock levels at a warehouse.

Spark's key feature is cloud execution. Browser-based agents work only while a tab is open; Spark is not tied to the user's browser. Even if you close the tab or switch off your computer, the agent keeps running on Google's servers. That suits long-running tasks that take hours or days.

The agent can remember context between runs: if you asked it to find flights cheaper than $200, it will remember that condition and come back with results once it finds a suitable option.

Use cases

Google gave several examples of how Spark can help in everyday life. The agent can sign you up for a webinar, fill in the application form and remind you an hour before it starts. It can compare the price of a product across five marketplaces, compile a summary and recommend the best option.

For business, this could mean competitor monitoring: the agent tracks competitors' prices, checks product availability, scrapes reviews and generates a report every morning.

Pricing and availability

Google AI Ultra starts at $100 a month and includes limited access to Spark. At launch it is a beta with a cap on the number of operations per day. Google did not specify the exact limits, but hinted that serious automation will require extra payment.

Key features of the base plan:

• Cloud execution 24/7 without interruptions
• Browser automation support (shopping, bookings, research)
• Context retention between sessions
• Integration with major web services (Gmail, Drive, etc.)
• A daily cap on the number of operations

For comparison: OpenAI Operator is currently available only in ChatGPT Pro ($20/month) and works in the browser, manually, under the user's visual supervision. It can also perform tasks on the web, but not around the clock.

Strategic position

Cloud AI agents are becoming a premium-tier product. Google is betting on ecosystem integration (Gmail, Drive, Docs) and cloud execution; OpenAI is betting on browser simplicity and explicit control. Both strategies make sense.

The cloud is more reliable for critical operations: if an agent must check a price every hour, it is better for that to happen on Google's servers than on your computer. The browser, in turn, is more transparent and easier for the user to follow: you see every click and every action.

What this means

For end users, it means automation is moving into the premium layer. With basic ChatGPT or Gemini you get a chatbot. For $20–100 a month you get an AI agent that does things for you.

For businesses, it opens the possibility of integrating cloud agents into their processes, through an API or built-in tools. But this will remain the preserve of expensive subscriptions and enterprise solutions for a long time.