Gemini Spark Reaches AI Pro Subscribers — But Free Users Are Still Locked Out

Google announced Gemini Spark at I/O 2026 as its answer to the agentic AI wave. Now it’s becoming real for more users — though the price of entry says a lot about who Google thinks this product is actually for.

As of this week, Gemini Spark is rolling out to Google AI Pro subscribers in the US — people paying $20 a month — alongside a global expansion to AI Ultra subscribers, who pay between $100 and $200 per month. Free-tier users aren’t in the picture yet. The expansion excludes users in the European Economic Area, Switzerland, the UK, and Nigeria, where availability remains limited. Those gaining access are also advised to check local language support before diving in.

What Gemini Spark Actually Does

Spark isn’t a chatbot upgrade. It’s a background agent: it runs on dedicated virtual machines on Google Cloud, meaning it works around the clock without needing your phone or laptop to stay open. You assign it tasks — via the Spark page in the Gemini sidebar on desktop, or through a menu option on mobile — and it executes them independently.

The feature that gives Google a structural edge over rivals is native Workspace integration. Spark connects directly to Gmail, Google Calendar, Drive, Docs, Sheets, Slides, YouTube, and Maps — turned off by default, but activatable through settings. Once connected, it can draft email replies, pull context from calendar events, extract action items from long threads, or build an outreach list from your inbox without any manual setup. Competitors relying on third-party MCP integrations to access the same services are starting from a significant disadvantage.

Spark is built on Gemini 3.5 and powered by Google’s Antigravity agent harness, the same platform underlying its agentic development tools. Users can also email Spark tasks directly through a dedicated Gmail address — a design choice that lowers the friction for anyone already spending most of their working day in Gmail. It also supports MCP integrations for third-party services, with more connections expected in coming months.

The Pricing Signals the Real Strategy

Framing Spark as a productivity tool for everyone would be misleading. Google AI Pro at $20/month has a reasonably broad audience; AI Ultra at $100–$200/month is a narrower segment by design. The tiered rollout — starting with Ultra at launch, then expanding to Pro now — follows a pattern Google and its competitors have used to test agentic workloads before opening them more widely.

The parallel with what Anthropic has done with Claude Cowork, or what OpenAI has built with ChatGPT’s agentic capabilities, is clear. All three are converging on the same product category: a persistent, context-aware digital worker that handles multi-step tasks across apps without constant prompting. Google’s differentiation isn’t the agent itself — it’s the ecosystem. Having Gmail, Drive, and Calendar natively accessible without integration work is a meaningful moat that neither Anthropic nor OpenAI can replicate.

For a deeper look at how Google’s engineering infrastructure connects to Spark’s agent architecture, Gemini’s managed agents and budget controls reveal the same underlying design principles at the developer tier. And for context on how token efficiency affects the economics of running agents at scale, Gemini 3.6 Flash’s efficiency gains matter here too — Spark’s background execution is only cost-viable if the underlying model is lean.

What the Expansion Doesn’t Include

Free users remain excluded entirely. Google hasn’t announced a timeline for broader access, and the geographic carveout for the EEA, UK, Switzerland, and Nigeria suggests regulatory caution rather than a technical limitation — likely tied to ongoing EU AI Act compliance work and data residency questions around an agent that reads personal email and calendar data.

That combination — a premium price floor plus regulatory exclusion in major markets — means Spark’s actual user base right now is a narrow slice of Google’s total audience. The product exists; the scale doesn’t yet.

For the agentic AI market, the more interesting question is whether $20/month is the right long-term price point for a persistent personal agent, or whether competition from OpenAI and Anthropic forces that floor down. Google has the ecosystem advantage. Whether it can monetize it without ceding the mass market remains the open bet.

Frequently Asked Questions

What is Gemini Spark?

Gemini Spark is Google’s 24/7 personal AI agent, built on Gemini 3.5 and the Antigravity agent platform. It runs on Google Cloud VMs and can execute tasks in the background — even when your devices are off — with native integration across Gmail, Calendar, Drive, and other Workspace apps.

Who can access Gemini Spark now?

As of July 2026, Spark is available to Google AI Pro subscribers ($20/month) in the US and to Google AI Ultra subscribers ($100–$200/month) globally, excluding the EEA, Switzerland, UK, and Nigeria. Free-tier users do not have access.

How do you assign tasks to Gemini Spark?

You can access Spark via the sidebar on desktop or through the menu option on mobile within the Gemini app. You can also email tasks directly to Spark through a dedicated Gmail address.

How does Gemini Spark compare to ChatGPT’s agent or Claude Cowork?

All three are persistent background agents capable of multi-step task execution. Spark’s key differentiator is native, out-of-the-box integration with Google Workspace — no third-party connectors required. Rivals must rely on integrations to access the same Gmail or Calendar data that Spark reads natively.