Developers using Google's Antigravity agentic AI coding tool are complaining about higher prices following an announcement yesterday that the company is evolving its AI plans. The shift reflects a common pattern in the AI sector where services that launch with apparent generosity pivot toward paid consumption models once adoption takes hold.
AI credits can now be used for Antigravity, with subscriptions providing some built-in credits while further credits are available for purchase as needed, at a cost of $25 for 2,500. Exactly what a credit is worth when used with Antigravity is not described in the documentation. This opacity creates immediate friction. Developers cannot calculate their true costs or plan budgets effectively when the company does not specify what a credit actually purchases.
Google's AI for Developers forum is filled with complaints, especially from those with AI Pro ($20.00 per month) subscriptions. The central grievance is not the pricing itself but the gap between what was promised and what is delivered. The company soon pointed users towards its Google AI Pro and Ultra subscriptions, using vague terms such as "high", "generous" and "meaningful" to describe quota limits. When preview launched in November 2025, developers who subscribed at $20 per month believed they had locked in a specific level of access.
That belief has collided with reality. One developer tracking token usage reported a dramatic collapse in available capacity: before January, they could consume over 300 million input and 1-2 million output tokens per week, but this week the same Pro subscription exhausted its limits after just 9 million input and 200,000 output tokens. The effective reduction in usable quota approaches 95 percent.
What has changed is not the subscription price but the underlying quota structure. According to sources describing the updated system, people who have been on the AI Pro plan since Antigravity was released feel that this is basically a bait and switch. The company had advertised quota refreshes every five hours for Pro subscribers, but those now refresh weekly instead, during which time developers cannot continue work without purchasing credits or upgrading to the Ultra tier.
Google's positioning of the tiers reveals the strategy. Google states that the Pro plan is suitable for "hobbyists, students and developers who live in the IDE," as opposed to relying on agents, and that these users can use Flash as a "taste test" for more advanced models. This reclassification matters. A tool originally marketed to all developers is now officially designated for learning only. Professional use demands the $249.99 per month Ultra plan.
The economics here are instructive. Antigravity still runs heavyweight models behind the scenes, in particular Gemini 3 Pro, which is already priced when accessed via the API. Internal references show Gemini 3 Pro token pricing in the order of $2 per million input tokens and $12 per million output tokens. Free IDE usage likely absorbs some of that cost today but may limit compute or impose charges for high volume. Google faces real compute constraints. The question is how fairly those constraints are distributed among users who paid in good faith for Pro subscriptions.
The broader context matters too. Pricing models shifted from flat-rate subscriptions to hybrid usage-based structures starting in late 2023, and that shift accelerated through 2025. This trend spans the AI industry, from OpenAI to Anthropic to Microsoft. As demand surges and infrastructure costs remain high, companies are moving from predictable subscription fees toward metered, consumption-based billing. It is a rational response to unpredictable usage patterns and genuine resource constraints.
Yet the timing of Google's announcement reveals a legitimate tension. Antigravity was launched in preview in November 2025, at which time pricing was not yet announced. This wording makes it hard to know the actual limits and how they may change. The company preserved flexibility by being intentionally vague. Developers who adopted early took a gamble on terms that could shift. Google held all the cards.
A fair question: should developers have anticipated this? The answer is complex. Those familiar with how other AI tools evolve might have seen it coming. But most developers do not track the pricing history of every platform they experiment with. They make reasonable assumptions based on what they read on the product page. When a service uses words like "generous" and "high," reasonable people infer a certain level of commitment.
For Australian developers and startups evaluating AI tools, the lesson is practical. Increasing pressure now stems from pricing mechanics, especially consumption-based pricing. Any tool marketed as "free" or based on opaque quota systems carries risk. The most reliable planning occurs when costs are predictable and transparent. When they are not, budgeting becomes guesswork and productivity planning becomes anxiety.
Google's move is not unusual in the AI market, but transparency lags behind it. The company should clarify what a credit buys, how quota calculations work, and whether Pro subscribers will see further changes. Developers deserve to know whether the $20 monthly investment represents a genuine tier or merely a gateway to higher spending. The company built Antigravity's early adoption on accessibility and ease. Sudden tightening of invisible quotas undermines both.