🟪 Bitcoin mining in the time of AI

Natural ice — considered a necessity before electric lights, natural gas or running water — was a big business in the 19th century.

Brought to you by:

"Done the right way, it is a very big benefit to the grid. Done the wrong way, it's the complete opposite."

Patrick Fleury on bitcoin mining

Bitcoin mining in the time of AI

Natural ice — considered a necessity before electric lights, natural gas or running water — was a big business in the 19th century.

The industry is thought to have begun in 1805 when the "ice king," Frederic Tudor of Boston, began cutting blocks of ice from northern lakes and shipping them by boat and rail around the US.

By mid-century, approximately 52,000 tons of natural ice were traveling annually to 28 cities across the country. Nearly every home, grocery store and bar came to have a sign in their window instructing the iceman on how many pounds they wanted left in their icebox that day.

When the Civil War began in 1861, however, the South was cut off from the natural ice harvested in the North — this is why mass production of artificial ice was pioneered in sweltering Confederate states like Texas.

To enable all that production, utility companies built power plants throughout Texas, much to the benefit of the nearby homes and businesses who were offered any excess power not consumed by the ice makers. It was the first access to electricity for Texans. 

Once connected to the grid, people gradually replaced their ice boxes with refrigerators — by 1944, 85% of American households had one.

All of those refrigerators were powered by an electrical grid that was financed largely by demand for ice.

Sound familiar yet?

The environmentally friendly argument for bitcoin is that bitcoin miners, by using stranded, intermittent power, will promote the buildout of a cheaper, cleaner and more stable electrical grid for everyone.

The details have always been a little fuzzy to me, though — could power-guzzling proof-of-work Bitcoin possibly be a net positive to society?

To find out, I spoke to Patrick Fleury of TeraWulf, one of the world’s lowest-cost and greenest producers of bitcoin (95% of the power they consume is zero carbon).

I still have a lot of investigating to do, but my initial takeaway is that bitcoin mining — when done right — already seems like a net benefit to society.

But it also seems like quite a lot of it is not done right  — and that AI’s enormous demand for power may be complicating the picture.

Could bitcoin miners bridge us to the time of AI in the same way ice-makers bridged us to the time of refrigerators?

It’s too early to say, but speaking with Patrick Fleury was a great way to start thinking about it.

Here are some highlights (edited for brevity) from my talk with the CFO of TeraWulf:

On whether bitcoin mining is a good business:

If you have low priced power, yes. The most important part of this business is being vertically integrated and having a very low cost of power. Our average price of power last year realized across two sites was 3.2 cents. Some of my biggest peers — they’re at least double that. If you don't have low price power, you're never going to make money. 

On using stranded power:

For us, that is Niagara Falls. There, we can be a resource to the grid by consuming a lot of power that would otherwise go unused. 

What bitcoin mining is doing, if done correctly, is consuming a lot of excess stranded power and converting that power from electrons to bitcoin. 

On how bitcoin mining lowers utility costs for retail customers:

By paying transmission charges. If you have more industrial users using the system, it lowers  system charges [for retail], because now that fixed cost is spread out over a much larger number of users.

On stabilizing the grid:

If I agree to shut down my site during the 20 to 100 hours a year where the grid is in crisis mode and prices are very high, then I am an asset to the grid because I'm helping stabilize it by consuming a lot of excess capacity. 

If the grid needs us, they call us a day ahead or the day of and tell us to shut down. At our upstate New York site, I can turn the whole site off in a couple of minutes. That is called demand response. And we participate in that in New York state. We get paid a nominal fee for providing that flexibility to the grid because that is very valuable.

On why TeraWulf stops mining for a nominal fee:

Because it's the right thing to do. I mean — if we don't do it, then we have a bullseye on our back from politicians and from you [the media]. 

And we should. The whole point is we're sitting there and soaking up power that's not being used most of the time. It's relatively insignificant to our operations, but it's a huge service back to the grid. 

When people need us to power down, we power down because that's a benefit to the grid and a benefit to society.

On whether the industry can be a net benefit to the grid:

As long as they are run the right way, [bitcoin miners] are a net positive to the grid because [mining] lowers fixed costs for everybody. That's a big benefit. 

And it's a flexible load, meaning in times of high power demand, we come offline. So, done the right way, it is a very big benefit to the grid. Done the wrong way, it's the complete opposite. Because if you're not curtailing your site when energy demand is very high, then you're just adding onto the problem of peak pricing.

On approaching zero carbon:

Ninety-five percent of the power that we consume is zero carbon, meaning it came from a nuke, a hydroelectric plant, or wind or solar. And our goal is to be 100% in the future.

On Q2 results for the industry:

A bunch of our big peers are going to lose money mining bitcoin this quarter and that should be an alarm bell for anybody that's invested in those companies. Go buy the ETF and at least you won’t get diluted by a management team that’s running the company not for profit, but effectively for their own use as a lifestyle company. 

On future profitability:

Now, [miners are] in a world where [they’re] competing with the likes of Meta, Amazon, Microsoft, Google and Oracle for power. Then you have a problem and you're never going to be able to fix that. [For higher-cost miners], the only way you get bailed out is if [the price of] bitcoin goes to a couple hundred thousand. 

But then, again, as an investor, you’re better off just owning the ETF. That's a better investment strategy than owning a non-profitable bitcoin miner that sells equity, stacks bitcoin on their balance sheet, and just constantly dilutes you. And pays their executives like NBA players. That's not a business. That's not a viable business model.

On AI competing for stranded and intermittent power sources:

I think you'll see more of the hyperscalers [Amazon, Meta, Google, Microsoft] enter that business as well because they're less focused now on latency. Those companies are already political hot potatoes for antitrust and other reasons. And now [they’re] going to double their consumption of power and that's not going to have any political ramifications? Of course it is. They have to change the way they have traditionally consumed power because power for data centers was typically an afterthought. They didn't care about the power price because the economics are such that it just [didn’t] matter. They're going to have to care about the power price for political reasons because it's going to be untenable.

On the overlap between bitcoin mining and AI compute:

The high-power compute AI business is going to have to act more like bitcoin miners in the future to permit that business to grow at the growth rate it wants to grow at — so that it’s politically and economically achievable. 

[Bitcoin miners and AI providers] are both significant users of power, so that's the commonality. Beyond that, there's really none. Bitcoin mining is a totally different machine. High-power computing AI [uses] a totally different chip. And then the infrastructure associated with both is completely different. It's a totally different business. But the commonality is you're converting electrons to either bitcoin mining or to high-power computing AI capacity.

On TeraWulf offering AI compute:

We previously [dedicated our] infrastructure to bitcoin mining because that was the best use of those electrons. Now it seems like the best, most value-accretive use for those electrons is AI. So it's not that big of a pivot [for us]. Now we have a different customer, and a more expensive architecture of the building, and the rack and the things that go around the GPU. But that’s pretty much it. 

On why providing AI compute gets a higher valuation from investors:

The reason why business gets valued at a much higher multiple is because the customers are higher-credit quality and the contracts are 15- to 25-year contracts. So it's a much more predictable business [compared] to bitcoin mining.

On higher-cost bitcoin miners switching to AI:

Not all sites are created equal, right? [TeraWulf has] access to low cost, zero-carbon power. We have a lot of land. We've got roughly 1,800 acres at [the New York] site. And we have access to water, which is necessary for cooling. 

I've heard some other companies talk about doing high-power AI computing in West Texas. I don't know how that would work. I guess you could try to do air-cooled if you don't have access to water, but I think that would be much more difficult. But I don't know, it may be possible.

On whether TeraWulf will be mining bitcoin in five years:

I don't know. 

I think it will depend on the price of bitcoin because we have a lot of operating leverage in our business where we've got relatively fixed power price and cost. So if bitcoin moves to $100 or $200,000, we can make a lot of money. 

That being said, if the high-power compute and AI business continues to grow the way we think it's going to grow, and those companies stay valued the way that they are now…as a management team, and board and insiders that own 30% to 40% of the equity of the company, I can tell you it's hard not to want to put the whole business into [AI compute] because it's just valued at a much higher multiple. 

It's a great question.

Brought to you by:

The PayPal stablecoin PYUSD is live on Solana. PayPal has over two decades of payments experience, and is now making low-cost, high-throughput web3 payments possible through the PYUSD on Solana launch. Learn about PYUSD on Solana and start building the future of payments.

PayPal, Inc. is licensed to engage in virtual currency business activity by the New York State Department of Financial Services.

  1. Does the ETH ETF ‘fee war’ even matter to investors? — Read

  2. Don’t forget about merged mining [Opinion] — Read

  3. New security council debuts with Coinbase, Anchorage as founding members — Read

  4. Pi Squared raises $12.5 million to build universal ZK Circuit powered by Proof of Proof [Sponsored] — Read

  5. TRON founder Justin Sun wins landmark case in the People’s Court of China [Sponsored] — Read

Are New Tokens Always Better?

In this episode, Blockworks Research analysts discuss recent market movements and the problem with narrative trading. Tune in for insights about Germany’s bitcoin sales, Mt Gox distributions and the implications of FTX repayments.

Watch or listen to 0xResearch on YouTube, Spotify, or Apple.

DePIN is ripe to disrupt a range of traditional infrastructure networks.

Our latest FREE report, made free thanks to AIOZ Network and POKT Network unpacks how investors can best gain long-term exposure to the sector, and where the opportunity lies.

recent research

Research Report Templates.png


ZKPs enable efficient offchain transaction processing and validation, resulting in increased throughput and reduced fees. Solana's ZK Compression leverages ZKPs to minimize onchain storage costs, while Sui's zkLogin streamlines user onboarding by replacing complex key management with familiar OAuth credentials.