A former OpenAI researcher is making a massive bet on Bitcoin mining — and the reasoning behind it may surprise many in the crypto world.

Leopold Aschenbrenner, a 24-year-old researcher who left OpenAI in 2024 after an alleged information leak, has invested roughly $1 billion into Bitcoin mining companies through his hedge fund Situational Awareness LP.

The fund is now worth about $5.5 billion, and it just shared its positions in a new filing with the US Securities and Exchange Commission.

A Billion-Dollar Vote of Confidence

Aschenbrenner has made a huge investment in the Bitcoin mining sector, one of the biggest we've seen from a major institution in a while.

He has a lot of money invested in some big mining companies that are listed on the stock market.

* Core Scientific

* Iris Energy

* Cipher Mining

* Riot Platforms

* Hut 8

Many of these companies are currently shifting part of their business toward AI infrastructure and high-performance computing.

That shift is exactly what makes them attractive.

The Real Asset: Electricity

It seems that the value of companies that mine Bitcoin might not be about the Bitcoin they produce, but about something entirely different - the power infrastructure they use to do it.

Mining companies already operate large-scale facilities connected to massive electricity supplies. In an era where artificial intelligence is struggling to secure enough energy to train models, those assets are becoming extremely valuable.

“A miner’s true value has always resided in its energy infrastructure and grid access,” said Nishant Sharma, founder of the mining and computing consulting firm Blocksbridge.

Miners have something that AI companies really need: they have access to a lot of power that's ready to use.

Why Miners Are Pivoting to AI

Bitcoin miners have been under heavy pressure since the 2024 halving, which cut block rewards in half and reduced mining revenue.

Right now, the cost of doing transactions is still pretty low because not as many people are using the network.

As a result, many mining companies have started exploring new revenue streams.

AI infrastructure has become the most attractive option.

Several miners have already signed major deals:

* Core Scientific secured a 12-year contract with AI cloud provider CoreWeave, expected to generate around $10 billion in revenue.

* Iris Energy is aiming to make over $500 million each year from its AI cloud services by the start of 2026.

* Riot Platforms signed a 10-year data-center lease with AMD to expand into high-performance computing.

This new approach gives investors like Aschenbrenner something that miners have always needed but rarely had: a steady and reliable income stream.

AI’s Biggest Problem: Power

The AI industry faces a major bottleneck — electricity.

Creating advanced AI systems like GPT-4 uses a huge amount of energy. To give you an idea, training GPT-4 needed over 12 megawatts of power. That's about the same amount of electricity that 12,000 homes use. This shows just how much power is required to develop these complex AI models.

Future models are expected to require even more.

Building new data centers and connecting them to power grids can take three to five years in the United States due to regulatory approvals and infrastructure upgrades.

Many Bitcoin miners have set up their operations in places with access to a lot of electricity, so they can power their equipment.

This makes them really appealing to companies that are trying to quickly increase their AI capabilities.

A Strategic Bet on Infrastructure

For Aschenbrenner, the investment isn’t just about Bitcoin.

It’s about energy infrastructure, data centres, and computing power — resources that are becoming increasingly scarce in the age of AI.

If demand for AI computing continues to surge, the power-ready facilities owned by Bitcoin miners could become some of the most valuable assets in the technology industry.

Why This Fired OpenAI Researcher Is Betting $1 Billion on Bitcoin Miners

Related Articles

John Crypto

Cryptocurrency — finally simple