How Much Power Does AI Really Use? Find Out Here!

Lionel Ringenbach just exposed something at Vancouver AI that’ll make you rethink every ChatGPT query you’ve ever made: 55% of AI requests in code editors are completely wasted. Not inefficient. Not suboptimal. Literally thrown away before they’re even used.

The Hidden Energy Crisis

Every time you type in VS Code with GitHub Copilot, here’s what actually happens:

Type one character → AI request sent
Type another character → Previous request cancelled, new one sent
Type faster → More cancelled requests
Result: Massive energy burn for absolutely nothing

Meet Wattsup: The Tool That Changes Everything – Lionel built what the industry won’t: an actual energy counter for AI usage. Not abstract metrics.

His VS Code extension tracks:

– Energy per request in real-time
– Daily/weekly/monthly analytics
– Efficiency scores
– The true cost of your “helpful” coding assistant

The Nuclear Reality Check

Eric Schmidt (ex-Google CEO) just admitted the US needs 92 gigawatts of new power just for AI. That’s 92 nuclear plants. They’ve built 2 in the last 30 years.

Sam Altman’s response? Invest in small modular reactors so tech companies can build private nuclear plants.

Trump’s response? Remove all environmental reviews for AI data centers.

The industry’s response? Build more nuclear instead of asking why autocomplete needs a power plant.

The Local Computing Revolution: Here’s what Kush pointed out that nobody’s discussing: Your M3 MacBook can run a 3-billion parameter model locally. So why is every keystroke going to a cloud API?

The scam:

– Government subsidizes mega data centers
– Zero incentives for personal compute
– Default settings force cloud usage
– You pay for API calls while your powerful hardware sits idle

What Lionel Actually Built
Wattsup isn’t just another monitoring tool. It’s accountability infrastructure. When you see your IDE burned 20 hours of microwave energy to help you code, you start asking different questions:

– Why can’t this run on my machine?
– Who profits from this inefficiency?
– How much AI is just AI talking to other AI?

The Transparency They Fear
Big Tech hides energy usage behind elegant interfaces. They know that the moment you see the real numbers, the whole narrative falls apart.

Lionel’s making those numbers visible:

– Every request tracked
– Every wasted token counted
– Every invisible cost exposed

This isn’t about using less AI. It’s about knowing what you’re burning and why.

Why This Matters Right Now
We’re at a crossroads:
Path 1: Build 92 nuclear plants, strip environmental protections, pretend efficiency doesn’t matter
Path 2: Make energy usage visible, run models locally, question every wasteful default
Lionel’s building the tools for Path 2. The question is whether we’ll use them before Path 1 becomes irreversible.

The Call to Action

For Developers:

– Fork Lionel’s code (GitHub link in description)
– Port it to your IDE
– Track your actual usage
– Share your numbers publicly

For Everyone:

– Question every AI interaction
– Demand energy transparency
– Support projects exposing real costs
– Choose local models when possible

The Uncomfortable Truth: The industry wants you thinking about features, not footprints. They want you amazed by capabilities, not concerned about costs. They want nuclear reactors, not efficiency metrics.

Wattsup breaks that narrative. It shows that maybe we don’t need 92 nuclear plants. Maybe we just need to stop wasting 55% of our compute on cancelled requests.

What Happens Next

Lionel’s presenting at MUTEK Montreal August 20th. The code is open source. The movement is just beginning.

This isn’t about AI skepticism. It’s about AI accountability. The difference between building tools that serve us versus systems that consume everything.

The future doesn’t need more nuclear reactors. It needs honest accounting.

Subscribe to Lionel’s newsletter. Install the extension. Make the invisible visible.

Because once we see the real cost, we can’t unsee it. And that’s exactly what Big Tech fears most.

00:00 The Truth: 55% of AI Requests Are Wasted
02:20 Meet Lionel: The Rebel Counting Every Watt
03:05 WhatsApp Project: Shining Light on Hidden Energy
04:01 The Invisible Problem: AI Everywhere, Transparency Nowhere
05:26 MUTEK Selection: From Vancouver Idea to Montreal Stage
06:13 The Discovery: Someone Already Built Part of This
07:00 GitHub Copilot’s Secret: Most Requests Never Get Used
08:23 Microsoft Goes Open Source (But Hides the Energy Data)
09:02 The Dashboard That Changes Everything
10:14 Microwave Metrics: Making Energy Real
11:04 Trump’s Nuclear Gift to Big Tech
12:40 What’s Next: Building the Movement
14:57 Q&A: The Hard Questions Begin
17:07 Kris Krüg: “Your Shit’s Foundational”
19:24 Kush’s Challenge: Local vs Cloud Reality Check
23:20 Eric Schmidt’s 92 Gigawatt Admission
25:30 The Nuclear Question: Do We Really Need This?