One Prompt, Five Hours Gone: Why Google’s New Gemini Limits Are Sparking a User Revolt

 

 One Prompt, Five Hours Gone: Why Google’s New Gemini Limits Are Sparking a User Revolt

**Subheading:** *Google just switched from counting prompts to counting "compute." For one AI Pro subscriber, a single failed video generation drained his entire 5-hour allowance in minutes. Now, the backlash is forcing the company into emergency damage control.*

**Estimated Read Time:** 6 minutes

**Target Keywords:** *Gemini usage cap, compute-based limits, Google AI Pro limit 5 hours, Gemini video generation failed, Antigravity limits tripled, Google AI Ultra price.*

---

## Part 1: The Human Touch – The 4-Minute Nightmare

Let me tell you about the fastest $20 subscription burnout in tech history.

Ashutosh Shrivastava, a Google AI Pro subscriber, sat down to use Gemini. He had a specific task in mind: generate a short video using the avatar-based video generation feature. He typed his prompt. He hit enter. The AI whirred for about four minutes.

Then it stopped.

The video generation failed. But the damage was already done.

Shrivastava opened his usage dashboard and saw something that made his blood run cold. His five-hour usage cap was at **100%**. A single prompt. A few minutes of processing time. And his entire allowance for the next five hours was gone.

"I started with 0% usage on my five-hour limit, then gave one simple prompt for video generation using the avatar feature," Shrivastava wrote on X, posting video proof of the debacle . "It ran for around three to four minutes, hit 100% of the rate limit, and the video generation failed as well."

The reaction from Google’s Gemini lead, Josh Woodward, was about what you'd expect from someone who just realized their product had eaten a paying customer's entire day. "Yikes, let us take a look!" Woodward responded .

This is the story of how Google’s well-intentioned "compute-based" usage limits became a PR disaster, why a single failed task can now nuke your entire afternoon, and whether the company's emergency 3x quota increase is enough to stop the revolt.

## Part 2: The Professional – How the New "Compute" System Works

To understand why Shrivastava’s nightmare happened, you have to understand the fundamental shift in how Google measures your usage.

### The Old Way: Counting Prompts

Until recently, AI limits were simple: X prompts per day. Did you ask a simple question? One prompt. Did you ask the AI to write a novel? One prompt. Easy.

### The New Way: Counting Compute

On May 17, 2026, Google flipped the switch on a completely new system . Instead of counting prompts, Gemini now tallies up the **actual computational resources** your request consumes .

That calculation includes :

- **Prompt complexity:** A simple "Hello" uses very little compute. A request to generate a video uses a massive amount.

- **Chat length (Context window):** The longer you talk to the AI, the more "memory" it has to process. Every new message gets more expensive than the last.

- **Features used:** Are you using Deep Research? Image generation? Video generation? These are compute hogs.

### The Five-Hour Refresh Cycle

Here's where the user pain really kicks in. Under the new rules, you have a rolling five-hour limit. Once you burn through it, you are locked out of premium features until the five-hour window expires . You can still use Gemini, but likely with a slower, less powerful model.

| Subscription Plan | Monthly Price | Usage Limit (vs. Free) | Context Window |

| :--- | :--- | :--- | :--- |

| **Free** | $0 | Standard | 32K tokens (~50 pages) |

| **AI Plus** | $7.99 | 2x Standard | 128K tokens |

| **AI Pro** | $19.99 | 4x Standard | 1 million tokens |

| **AI Ultra 5x** | $99.99 | 5x AI Pro | Very high |

| **AI Ultra 20x** | $199.99 | 20x AI Pro | Highest |

Source: 

In Shrivastava’s case, his $19.99/month AI Pro plan (which offers roughly 4x the limits of the free tier) was devoured in one go. The system calculated that the computational cost of generating that single video was equivalent to *five hours* of typical use.

## Part 3: The Creative – The "Context Window" Trap

The biggest hidden danger in the new system isn't video generation. It's something far more common: **long chats**.

### The Creeping Cost

Imagine you are a software developer using Gemini to debug a large codebase. You upload a file, it reads it. You ask a question, it answers. You go back and forth for an hour. The AI is storing all of that conversation as "context" to understand your follow-ups.

By the time you ask your tenth question, the AI is processing not just your new question, but the entire history of the chat. Under the new compute-based system, that final prompt might cost 10x or 20x more than your first prompt . **The longer you talk to Gemini, the more expensive each new word becomes.**

This is the silent killer of productivity. Developers are finding that their "5-hour limit" is getting exhausted after just one or two hours of genuine, in-depth work . The system is actively punishing you for staying in the same conversation.

### The Flawed Logic

The logic of the new system is that you should break up complex tasks into dozens of smaller, fresh chats. This is theoretically "more efficient" for Google’s servers.

But it is terrible for human beings. We think in threads. We build context. We iterate. Asking a developer to "start a new chat" every 20 minutes to save compute is like asking a novelist to start a new Word document every chapter to save ink.

## Part 4: Viral Spread – The Backlash and the Google 'Firefighting'

The backlash hit almost immediately. Within days of the rollout, the Gemini subreddit and X were flooded with complaints.

### The "Calculator" Complaints

One user on Reddit told Android Authority that they’re canceling their Pro subscription after exhausting 50% of their usage limit with a simple five-prompt back-and-forth with the AI . Another pointed out that using the "personalization" feature—where Gemini remembers details about your life—drastically impacts limits because it constantly loads your personal information into the context window.

### Google’s Emergency Response

The backlash was so severe that Google was forced into an emergency "firefighting" mode . The Director of Antigravity (Google’s coding assistant platform), Varun Mohan, announced two major patches:

1.  **Full Reset:** All Gemini Pro and Ultra users had their weekly usage limits reset immediately.

2.  **Limits Tripled (For Some):** For the Antigravity coding platform, usage limits were permanently increased by **3x** .

Additionally, existing Ultra subscribers were offered a $100 credit for "Antigravity rewards" as an apology .

But for the everyday AI Pro user generating videos or images in the main Gemini app? Those core limits remain largely unchanged.

## Part 5: Pattern Recognition – What This Means for You

### The "Ultra" Escape Hatch

Google’s strategy appears to be segmentation. If you’re a power user, they want you to shell out for the $99.99 or $199.99 AI Ultra plans, which offer "5x to 20x higher limits" .

The problem is that the *pricing* for these tiers just changed. Google slashed the price of the top-tier Ultra 20x plan from $250 to $199.99 at I/O 2026 . They are aware of the "compute crunch" and are trying to offer a viable path. But for the average Pro user paying $20/month, the new limits feel like a bait-and-switch .

### The Transparency Problem

Google added a new "usage meter" to the settings menu . You can now see exactly how much of your five-hour window you have left. But as Shrivastava’s case proves, knowing you are about to hit the limit doesn't help if you hit it immediately.

### What You Can Do

| Problem | Workaround |

| :--- | :--- |

| **High compute costs** | Break long conversations into separate, shorter chats. Avoid letting the context window get too heavy. |

| **Unpredictable video/image costs** | Assume any media generation could wipe out a significant chunk of your quota, and be ready to upgrade or wait it out. |

| **Check your usage** | Go to `gemini.google.com/usage` to see exactly where your limits stand before you start a heavy task . |

## Conclusion: The "Compute" Tax on Creativity

Let me give you the bottom line.

Google’s shift to compute-based limits was inevitable. As AI models get smarter and generate more complex media (video, audio, code), the old "prompt counting" system became economically unsustainable. You wouldn't expect a 4K video to cost the same to stream as a text file.

However, Google’s rollout has been a disaster. The limits are too strict, the transparency is lacking, and the educational effort has failed. Paying users should not be finding out that a single task has locked them out for five hours by reading about it on X.

**Here’s what I believe, friendly and straight:**

Ashutosh Shrivastava paid for access, not for anxiety. Watching a usage meter climb to 100% while your task fails is not a "premium experience." It is a broken user interface that happens to be charging you by the millisecond of compute.

Google has a choice: dramatically increase the base limits for Pro users to account for the true cost of video/AI agents, or watch their subscriber base churn to a competitor that offers a simpler, more transparent pricing model.

**What you should do right now:**

| Step | Action |

| :--- | :--- |

| **Step 1** | **Check your usage meter** before running any major tasks. You can find it in the settings menu or at gemini.google.com/usage. |

| **Step 2** | **Shorten your chats.** If you have a large coding or writing project, break it into smaller, isolated conversations rather than one massive thread. |

| **Step 3** | **Consider the tier.** If you are a heavy video or image creator, the $19.99 Pro plan may not be enough. You might need to budget for the $100 Ultra plan. |

| **Step 4** | **Vote with your wallet.** If the limits are too restrictive, cancel your subscription. Competing models from Anthropic and OpenAI have different (and sometimes simpler) pricing structures. |

**The final word:**

The age of "unlimited AI" is over. We are entering the age of "metered AI." The technology is incredible, but the billing is brutal. Until Google sorts out its algorithm, treat every click like it costs a nickel—because, in compute terms, it just might.

---

## FREQUENTLY ASKING QUESTIONS (FAQ)

**Q1: What happened to the Gemini usage limits?**

**A:** Google replaced its old "per-prompt" limits with a new "compute-based" system on May 17, 2026 . This new system charges usage based on the complexity of the prompt, the length of the chat, and the features used (like video or audio generation). Limits refresh every 5 hours until a weekly cap is hit .

**Q2: Did Google respond to the "one prompt, five-hour limit" complaint?**

**A:** Yes. Josh Woodward, Google’s Gemini lead, responded publicly to the user, Ashutosh Shrivastava, saying "Yikes, let us take a look!" . The company has since reset usage for some users and temporarily increased Antigravity limits, though the core Gemini limits remain in effect .

**Q3: Does a "long chat" use more of my quota?**

**A:** Yes. This is the most important hidden detail. Because the system has to process the entire conversation history (the context window) every time you send a new message, a 20th message in a long chat costs significantly more compute than the first message . Start fresh chats often to conserve quota.

**Q4: Is there a way to check my usage before I run out?**

**A:** Yes. Google recently added a usage meter to the Gemini settings menu. Users can see their remaining limits for the current 5-hour window and the weekly limit at `gemini.google.com/usage` .

**Q5: Did Google increase the limits after the backlash?**

**A:** For users of **Antigravity**, Google permanently **tripled the usage limits** . For regular Gemini App users, the AI Pro limits are technically unchanged, though Google reset the weekly usage for many users.

**Q6: What is the price of the Google AI Ultra plan?**

**A:** Google AI Ultra costs **$99.99 per month** for the "5x" plan (5x AI Pro limits) and **$199.99 per month** for the "20x" plan (20x AI Pro limits) . Google recently reduced the price of the 20x plan from $250/month .

**Q7: What is the "5-hour window" everyone is talking about?**

**A:** Instead of a simple "daily" limit, Gemini now operates on a **5-hour rolling window**. You have a certain amount of "compute" you can use in any 5-hour period. Once you use it up, you are throttled until the next 5-hour window begins .

**Q8: Will Google walk back these changes?**

**A:** Unlikely entirely. The shift to "compute-based" billing is strategic. However, given the massive backlash, it is almost certain they will raise the **size** of the limits in the coming weeks.

---

**Disclaimer:** This article is based on user reports and Google announcements as of May 26, 2026. Usage limits, pricing, and features are subject to change by Google.