Fred ThielChairman and CEO, MARA Holdings

On our Q2 earnings call I said the market was getting a little frothy. We had huge tailwinds from institutional buy in, big inflows to ETFs, and a lot of expectations around Fed rate cuts. When that many positives stack up, people crowd into the same trade, and it gets fragile fast.

It was a couple things back to back. The Fed rate cut suddenly looked like a low likelihood, which created risk off. Then there was a scare around AI assets, debt, and the AI circular economy. Add Japan rate hikes and the carry trade unwind, and it became a compounding effect.

Since late summer we saw older whales with wallets 10 to 13 years old moving. When an old wallet that has not moved in many years moves, people freak out. Some were selling, some were transferring into ETFs, which looks like a sale. In risk off, that fear spreads quickly.

In a risk off environment, Bitcoin operates a lot like equity in portfolios today. Then you see layered strategies and highly leveraged positions unwind. People pile into short trades on perpetual markets, and when the move goes against them, forced unwinds add more selling pressure on top of everything.

I think quantitative tightening is officially over and you will start seeing some quantitative easing. The repo market is going to need help, and there are areas where money gets injected into the economy. The upside is liquidity, but the problem is inflation and how you keep it under control.

A period of uncertainty makes people hesitate. The government shutdown did not help and the lack of data did not help. You focus on what is priced in, what you can control, and avoid panic decisions. Then you look for stable footing and plan what matters between now and next quarters.

Borrowing against Bitcoin to invest in AI can be a challenging situation. If Bitcoin price is depressed, those profiles can go negative and it causes issues. Miners should stay focused on mining Bitcoin, then leverage the assets they already have and convert them into higher value options carefully.

There are multiple ways. You can lease a site to an HPC developer and get a long term lease. Some build a building, buy GPUs, run them, and rent that out. Others do land and power deals with developers. There is also a hybrid model where you mine Bitcoin and operate AI at the same site.

We are moving toward inference AI in the same containerized modular format as our Bitcoin mining. We do not have to build big buildings, and we do not need water cooling, it is air cooled. We think training should be best left to hyperscalers, but inference is where we can play.

If you run inference AI, you run a model and you provision it with data, and that data is like gold. You want to protect it, so private cloud becomes very important. That is part of why we invested in Exaion in France, with strong data sovereignty and data protection capabilities.

We announced in our Q3 call we would start selling some of our production Bitcoin, and we have done that. But we still net increase our balances every month anyway. The goal is to fund operations responsibly while keeping a strong balance sheet and growing the asset base over time.

We are certainly not cutting back our hash rate. Global hash rate has not slowed down much, except recently when older equipment shut down after price drops. Many peers are converting capacity they have not brought online yet, or building green field sites, not necessarily ripping out running sites.

Two things matter: the reward per block and how many blocks you win, which depends on your share of global hash rate. Global hash rate grew all year, so difficulty got harder. If you did not add capacity, you went backwards. Cost structure decides who survives when price dips.

If you own and operate sites, your cost per kilowatt hour is lower than hosted models. Hosted operators pay more and feel it first when Bitcoin comes down into the 80s or 90s. At our owned and operated sites, we have some of the lowest costs in the industry, which protects margins.

We operate behind the meter at a wind farm we own, so marginal energy cost can be near zero when the wind blows. We move fully depreciated miners there, so there is no depreciation expense, and they run until they expire. We also do similar low cost work using flare gas in oil fields.

Many investors value us close to the value of our Bitcoin holdings, and mining can get little attribution in market cap. But savvy investors look at our sites and power pipeline. The biggest constraint in AI is power, and the fastest way to get power is converting a mining site, so optionality matters.

Closing the Exaion investment after regulatory approval and finalizing the MPLX deal are key, and we expect both in Q1. We also have international steps like Saudi and France footprints. The focus is treating land and power as a portfolio of assets and maximizing value across Bitcoin and inference AI.