Close Menu
  • AI
  • Content Creation
  • Tech
  • Robotics
AI-trends.todayAI-trends.today
  • AI
  • Content Creation
  • Tech
  • Robotics
Trending
  • A Coding Implementation on Qwen 3.6-35B-A3B Masking Multimodal Inference, Considering Management, Device Calling, MoE Routing, RAG, and Session Persistence
  • A Coding Implementation on Microsoft’s Phi-4-Mini for Quantized Inference Reasoning Instrument Use RAG and LoRA High-quality-Tuning
  • Moonshot AI Releases Kimi K2.6 with Lengthy-Horizon Coding, Agent Swarm Scaling to 300 Sub-Brokers and 4,000 Coordinated Steps
  • In China, a humanoid robot set a record for the half-marathon.
  • Prego Has a Dinner-Conversation-Recording Device, Capisce?
  • AI CEOs think they can be everywhere at once
  • OpenAI’s GPT-5.4 Cyber: A Finely Tuned Model for Verified Security Defenders
  • Code Implementation for an AI-Powered Pipeline to Detect File Types and Perform Security Analysis with OpenAI and Magika
AI-trends.todayAI-trends.today
Home»AI»Extropic is aiming to disrupt the Data Center Bonanza

Extropic is aiming to disrupt the Data Center Bonanza

AI By Gavin Wallace29/10/20253 Mins Read
Facebook Twitter LinkedIn Email
A United Arab Emirates Lab Announces Frontier AI Projects—and a
A United Arab Emirates Lab Announces Frontier AI Projects—and a
Share
Facebook Twitter LinkedIn Email

Extropic, a startup developing an exotic new kind The following are some examples of how to use computer chip It has also demonstrated that advanced hardware can be used to tackle more useful tasks. artificial intelligence Scientific research is important.

These chips are designed to work differently than those from Nvidia or AMD. When scaled-up, they will be up to thousands of time more energy efficient. With AI companies pouring billions of dollars into building data centersAn entirely new approach to the production of chips could be a much less expensive alternative.

Extropic refers to its processors as thermodynamic sampling unit (TSU) instead of central processing units or graphics processing devices (GPUs). TSUs are silicon-based devices that harness thermodynamic electronic fluctuations and shape them in order to create probabilities models for complex systems like weather forecasts or AI algorithms capable of creating images, texts or videos.

Extropic has shared the first chip with several partners. These include representatives of various governments, as well as frontier AI laboratories, companies working in weather modelling, and startups. Extropic declined to name the partners.

“This allows all sorts of developers to kick the tires,” Guillaume Verdon is the CEO of Extropic. He gained fame in the tech industry as an entrepreneur. colorful and sometimes controversial online persona called Based Beff Jezos A new technology philosophy is known as effective accelerationism Or e/acc, before founding the start-up. Verdon, along with Trever McCourt (Extropic’s Chief Technology Officer), previously worked together on quantum computing Google, before they pursued their innovative computing approach.

Johan Mathe – CEO of Atmo – is among those testing out the new hardware. Atmo uses AI models which can provide a higher level of resolution in forecasting than would otherwise be possible. Customers include the Department of Defense. Mathe claims that Extropic chips will make it easier to predict weather.

Extropic also releases software, TRHML. This allows users to emulate the Extropic chip behavior using a GPU. Mathe’s used the TRHML software, as well as a real Extropic chip. “I was able to run a few p-bits and see that they behave the way they are supposed to,” Mathe Says

The hardware of the company is called XTR-0. This consists of a FPGA (field-programmable-gate-array) chip which can be configured for various tasks. Two of its probabilist chips X-0 each contain a small number of qubits.

The XTR0-.

Extropic

Single daughterboard.

A single daughterboard.

Extropic

The new chip uses p-bits to model uncertainty instead of the conventional 1 and 0 bits. Though limited in scope, this chip shows off the capabilities of the company’s new approach.

“We have a machine-learning primitive that is far more efficient than matrix multiplication,” McCourt: “The question is, how do you build something on the scale of ChatGPT or Midjourney.”

artificial intelligence data machine learning nvidia
Share. Facebook Twitter LinkedIn Email
Avatar
Gavin Wallace

Related Posts

In China, a humanoid robot set a record for the half-marathon.

20/04/2026

Prego Has a Dinner-Conversation-Recording Device, Capisce?

20/04/2026

AI CEOs think they can be everywhere at once

20/04/2026

Schematik Is ‘Cursor for Hardware.’ The Anthropics Want In

18/04/2026
Top News

OpenAI had banned military use. Microsoft was still used to test the Pentagon’s models.

Daniela Amodei of Anthropic believes that the market rewards safe investments

Can AI Kill Venture Capitalists?

Elon Musk’s xAI Sues Apple & OpenAI for App Store Rankings

Why not check out WIRED?Seriously, What Is ‘Superintelligence’? | WIRED

Load More
AI-Trends.Today

Your daily source of AI news and trends. Stay up to date with everything AI and automation!

X (Twitter) Instagram
Top Insights

Google AI introduces DS STAR, a multi-agent data science system that plans, codes and verifies end to end analytics.

06/11/2025

The engagement of these platforms is increased by 5 to 42% when you reply to comments

12/11/2025
Latest News

A Coding Implementation on Qwen 3.6-35B-A3B Masking Multimodal Inference, Considering Management, Device Calling, MoE Routing, RAG, and Session Persistence

21/04/2026

A Coding Implementation on Microsoft’s Phi-4-Mini for Quantized Inference Reasoning Instrument Use RAG and LoRA High-quality-Tuning

21/04/2026
X (Twitter) Instagram
  • Privacy Policy
  • Contact Us
  • Terms and Conditions
© 2026 AI-Trends.Today

Type above and press Enter to search. Press Esc to cancel.