Hi Synapticians!

Without kilowatts, your GPUs are nothing more than overpriced paperweights.

xAI has acquired the former site of a power plant in Southaven, Mississippi, just a few miles from its future mega–data center in Memphis. The real prize isn’t the concrete ; it’s the right to plug in hundreds of megawatts, whenever and wherever you want.

In Memphis, xAI’s Colossus supercomputer has already sparked controversy: roughly 200,000 GPUs, methane-fueled generators rushed into place, and growing local pushback. This project captures the new geopolitics of AI: promises of jobs, public health concerns, and communities demanding guardrails. AI doesn’t just run on silicon; it runs on politics.

Globally, the picture is just as clear: the International Energy Agency projects that data center power consumption will double by 2030, with AI driving much of that growth. The tech giants are already locking in firm gigawatts: Amazon has secured up to 1.92 GW of nuclear power; Google is tapping 115 MW of 24/7 geothermal. And some analysts estimate that by 2028, a single frontier model training run could draw 1–2 GW. To put that in perspective: 30 days of training at 1.5 GW consumes as much electricity as roughly 100,000 homes use in an entire year.

The takeaway: whoever controls dispatchable, clean energy will control AI.

With current architectures, the true barrier to entry in AI is no longer the model, nor even the GP; it’s guaranteed, predictable, decarbonized electricity. The companies that lock in 24/7 gigawatts will win on performance, costs, and social license to operate. The rest will be left improvising at best.

Here’s the rest of the news about AI today:

• OpenAI's AI System Achieves Gold Medal Level Score at IOI 2025

• Maximizing GPT-5 Performance: A Comprehensive Guide to Prompting for Agentic Tasks and Coding

• TRIBE: A Deep Neural Network for Multimodal Brain Response Prediction

Top AI news

1. Elon Musk's xAI Buys Former Gas Power Plant Site for AI
Elon Musk's xAI acquired a 640-acre former gas power plant in Southaven, Mississippi, for $10. The site will power its AI infrastructure, including the Grok AI model, ensuring energy independence. This strategic move addresses the massive energy demands of AI development, aiming for self-sufficiency and control over power supply to scale operations effectively. Read online 🕶️

2. OpenAI's AI System Achieves Gold Medal Level Score at IOI 2025
OpenAI's AI system earned a gold medal-level score at the International Olympiad in Informatics 2025. This prestigious competition for high school students showcases the AI's advanced problem-solving and programming skills. The achievement signifies a leap in AI capabilities, prompting discussions on its role in education and future skill development. Read online 🕶️

3. Maximizing GPT-5 Performance: A Comprehensive Guide to Prompting for Agentic Tasks and Coding
GPT-5 enhances agentic tasks and coding through advanced prompting. This guide covers controlling agentic eagerness, using tool preambles, and optimizing coding with specific frameworks and techniques. Key parameters like reasoning_effort and the Responses API are crucial for maximizing performance and efficiency. Experimentation is key to unlocking GPT-5's full potential. Read online 🕶️

4. TRIBE: A Deep Neural Network for Multimodal Brain Response Prediction
TRIBE, a novel deep learning pipeline, predicts whole-brain fMRI responses using text, audio, and video. It achieves state-of-the-art results by integrating multimodal, multisubject, and nonlinear approaches, outperforming previous methods. This work advances our understanding of brain activity and cognition by offering a unified, data-driven model. Read online 🕶️

Meme of the Day

Stay Connected

Feel free to contact us with any feedback or suggestions; we’d love to hear from you !