Google is restructuring how it sells AI tools to everyday users. The company has announced three subscription tiers — AI Plus at $4.99 a month, AI Pro at $19.99 a month, and AI Ultra at $99.99 a month — consolidating what used to be scattered under different product names into one clear family.

This move comes alongside several big announcements: Google is planning to spend between $180 billion and $190 billion per year on AI infrastructure, launching a new AI assistant called Gemini Spark that works more on its own, and signing a partnership with SpaceX for cloud computing services from October 2026 through June 2029.

Three Price Tiers for Three Types of Users

AI Plus at $4.99 a month is the entry point. This is impulse-purchase pricing — cheap enough that most people won't overthink whether it's worth it. Google is offering promotional discounts for new subscribers after its January 2026 launch.

AI Pro at $19.99 a month is where most current paying Google AI users will likely land. It's also priced to compete directly with ChatGPT Plus and Claude Pro from competitors OpenAI and Anthropic, which charge similar amounts. This is the tier aimed at regular users who want consistent access to good AI tools.

AI Ultra at $99.99 a month is for power users — researchers, developers, software engineers, and professionals who need the absolute best performance and the highest limits on how much they can use the service. It signals access to Google's most capable AI models and priority support. The big price jump between Pro and Ultra suggests real differences in capability, not just minor extras.

Gemini Spark: An AI That Acts Without Being Asked

The more interesting technical announcement from Google I/O in May 2026 is called Gemini Spark. It's designed to do things for you before you ask, rather than sitting around waiting for a question.

Think of the difference this way: today's AI chatbots are like a helpful person you call when you need something done. Gemini Spark is trying to be more like a personal assistant who knows your schedule, your inbox, and your routines well enough to take action on their own — without you prompting them every time.

To work that way, the AI needs to remember a lot about you — your calendar, your emails, your preferences, patterns of work. It also needs you to trust it. Getting that balance right — being helpful without being annoying — is one of the hardest problems in AI right now. Google hasn't fully explained which subscription tier Gemini Spark will come with, but it's likely most useful at the Pro or Ultra level.

The Big Money Behind the Scenes

The subscription plans are really just the public face of something much larger: Google is betting enormous amounts of money on AI infrastructure. The company raised its yearly spending forecast to $180 billion to $190 billion in April 2026. That's an enormous sum — far bigger than the investment Google made in data centres in the early 2010s.

We have seen tech companies make bets like this before. In the early 2010s, the major cloud providers like Amazon, Microsoft, and Google raced to build data centres. Those companies spent huge amounts, and over time the competition compressed profits until things stabilised. The AI spending cycle is moving faster and costing more, because AI systems require special chips — particularly graphics processors — that are very expensive. The big open question is whether Google can sell enough AI subscriptions to justify all this spending.

A Cloud Deal With SpaceX

Google recently signed a cloud computing partnership with SpaceX running through June 2029. Why does this matter. SpaceX isn't a typical office software customer. Its work includes satellite tracking, launch simulations, and managing the Starlink internet network — all computationally demanding and requiring fast response times. Winning that deal against Amazon and Microsoft suggests Google Cloud can handle demanding technical workloads and likely offers competitive pricing for companies doing heavy computing.

It also shows Google Cloud's strategy: pursue customers doing sophisticated technical work that stress-test the platform's real capabilities, rather than just signing up companies migrating their office software to the cloud.

The Strategy: Price, Features, and Stickiness

Looking at all three subscription tiers together with Gemini Spark and the massive infrastructure spending, Google's plan becomes clear.

AI Plus captures people who might otherwise stick with the free version or try a competitor's cheap offering. AI Pro holds the majority of people willing to pay, competing head-to-head with OpenAI and Anthropic. AI Ultra goes after the small number of users — often developers and power users — willing to pay a premium, because those users tend to influence others.

The bigger bet is on Gemini Spark. If an AI assistant can reliably help you with repeating tasks — like scheduling meetings, organizing research, or automating workflows — you'll depend on it more over time. The assistant learns more about you the longer you use it, which makes it harder to switch to a competitor. This kind of increasing usefulness over time is what every major AI company is chasing right now.

Google has an advantage here. It already knows a lot about you through Search, Gmail, Calendar, Maps, and YouTube. That gives it more raw material to build an AI assistant that understands your life and work. The real test, though, is whether people will actually trust an AI to act on its own. That's still an open question.

The subscription tiers themselves are designed to generate steady monthly revenue — money the company needs to pay for all that AI infrastructure spending. At $180 billion to $190 billion per year, Google needs a lot of people paying $19.99 a month. That's why making Gemini Spark genuinely useful matters so much: it's the product that could make those subscriptions feel essential rather than optional.