The following is a guest post by Shane Neagle, Editor-in-Chief of The Tokenist.
When the Federal Reserve tampered with the money supply by increasing it by about 40% in 2020, everyone has paid for the cost of that tampering through inflation. The result is that it has siphoned off people's living energy as their savings dwindle. When you have more money but can buy less, you have to expend extra energy to keep up the same pace.
The solution to this problem is obvious: make money tamper-proof through decentralization and a fixed supply – so no one can control it. This is what powers Bitcoin, but it requires a key component to function: a physical foundation.
If Bitcoin were just a digital asset, it would be easy to change the ledger on the network, known as the blockchain. A great solution to this is proof-of-work mining, which acts as an energy barrier that ties Bitcoin's digital code to real-world resources. If an attacker wanted to change a record on the ledger, they would need a huge amount of energy to harness the computing power.
At 733.41 EH/s (hashrate), such an energy barrier is virtually insurmountable. But this means that Bitcoin's energy requirements are the cost of immutable money. Similarly, energy is also the cost of data centers churning out text/images/video/code every time people tell their AI agents what to do.
In both cases, human productivity increases. But can we optimize energy demands in a symbiotic way?
The Energy Dynamics of Bitcoin Mining and AI
It is no exaggeration to say that being a developed country is strongly correlated with high energy usage, which is clearly seen when comparing electricity generation per capita in kilowatt-hours (kWh) with a country's Gross Domestic Product (GDP).
Image credit: ourworldindata.org
In other words, access to surplus energy is a requirement for any advancement in civilization to occur. After all, once multiple layers are added to a basic level of subsistence, such as agriculture, new layers must then be fed: manufacturing, transportation, public services, urbanization, computing, etc.
Beyond just data centers for internet browsing and online banking, generative AI and bitcoin mining represent the latest layer of civilization as high performance computing (HPC), which has very high energy demands.
According to the U.S. Department of Energy (DoE), data servers already consume 10 to 50 times more energy (per floor area) than commercial office buildings and account for 2% of total electricity use in the U.S. Given the trend for increased demand from data centers, the International Energy Agency (IEA) predicts that total electricity consumption for data centers could increase by more than 1,000 terawatt-hours (TWh) in 2026.
For comparison, such a surge in demand would be equivalent to Japan's current electricity consumption. In comparison, Bitcoin mining has generated electricity demand of 130 TWh, the EIA noted.
Image courtesy of IEA (data network center consumption excluded)
Given that a simple Google search query requires 0.3 watt-hours of power, and a single ChatGPT query requires 2.9 watt-hours of power, Goldman Sachs Research predicts that AI datacenter consumption will reach 200 TWh per year between 2023 and 2030.
These trends necessitate a significant optimization effort, which the Bitcoin mining industry continues to undertake by upgrading to more efficient ASIC machines, primarily manufactured by Bitmain, MicroBT, Canaan, Bitfury, Ebang, and others.
Likewise, a good cooling solution will allow your ASIC rig to maintain a lower operating temperature for longer, reducing the need for cooling power consumption in the process, significantly reducing energy consumption. It is estimated that liquid and immersion cooling can reduce Bitcoin mining operational costs by up to 33%.
In the AI energy space, Nvidia's GPUs dominate the market with 65% market share. Nvidia's latest Blackwell GPU microarchitecture is said to reduce energy costs by 25x compared to its predecessor Hopper. When it comes to locally hosted large language models (LLMs), major suppliers to major tech companies such as Mistral, Meta, and Apple are leading the way, and GPU server hosting and related architectures are expected to increase in the future.
But optimizing energy consumption requires more than updating to better chips and tweaking the cooling, and this is where Bitcoin mining in particular comes into play.
The role of Bitcoin mining in energy management
It's simplistic to think that power plants produce electricity and then the output travels to the consumer, where it is transmitted over long distances at high voltages before being converted to a lower voltage for the end user.
In other words, the power grid must balance high output with low input, resulting in transmission and distribution (T&D) losses, which account for an average of 5% according to the EIA.
One way to strike this balance is to rely on energy storage, which can fill the gap between sudden fluctuations in electricity demand and supply. However, battery storage not only has high upfront investment costs, but mainstream lithium-ion batteries are known to be at risk of thermal runaway and susceptible to overheating.
Ultimately, no solution beats the efficiency of being closer to the energy source, which is why bitcoin mining company TeraWulf (Nasdaq: WULF) selected Nautilus Cryptomine as its flagship facility near the 2.5 GW Susquehanna Nuclear Generating Station in Berwick, Pennsylvania.
TeraWulf has positioned itself as the most efficient bitcoin mining operation with zero-carbon energy at 2 cents per kWh by drawing 300 MW directly from the power plant.
More importantly, Bitcoin mining helps balance the power grid by acting as a dispatchable load: because HPC is energy intensive, the load can be adjusted in real time to smooth out fluctuations in energy supply and demand.
As of July 2024, the Electric Reliability Council of Texas (ERCOT) reported that 3GW of its 5.5GW of electricity was dedicated to Bitcoin mining load sharing.
Not only will load dispatch provide on/off ramps in response to local power shortages or surpluses, but Bitcoin mining companies will be incentivized to do so once they start reporting their power sales.
This injects a new element of security into Bitcoin as a tamper-proof currency. Bitcoin mining companies can offset costs by scaling back operations when they sell BTC, allowing them to be compensated for their role in balancing the power grid. As an example, Riot Platforms (Nasdaq: RIOT) received $2.2 million from ERCOT's demand response credits in January 2024.
More directly, Bitcoin miners could recover stranded energy by using up flared gas burned in oil and gas fields, or even collect/recycle the heat generated by BTC mining to heat water or greenhouses.
The Convergence of AI and High Performance Computing (HPC)
So far I've found the following:
AI and Bitcoin mining are both energy intensive. Power grids have friction as a function of power distribution and load balancing. Bitcoin mining can reduce that friction.
But can Bitcoin mining also be integrated with AI data centers?
Although both fall under the umbrella of High Performance Computing (HPC), AI services require a low probability of interruptions. The success of current and future AI apps depends on uptime/response times, and data centers are not well-suited to deploy the same flexible load balancing strategies as Bitcoin mining companies.
At the same time, bitcoin mining companies have a track record of innovation, such as taking advantage of continuous hydro/nuclear power to scale their operations, and as AI-dedicated data centers put strain on the power grid, the miners' flexible load following allows them to quickly respond to that power depletion.
In addition to ERCOT, more states are starting to wake up to this trend: At the end of May, the Oklahoma Legislature introduced HB1600, a bill that would make digital asset mining operations eligible for tax credits, with special provisions for load balancing.
“Mining must take place in certified colocation facilities that have load reduction agreements with retail electricity suppliers.”
As a result, more Bitcoin mining companies are moving to directly host their AI operations.
Hybrid Data Center Strategy
Despite catering to a variety of HPC aspects, Bitcoin mining facilities are also ideal for hosting AI operations, not only because they have skilled manpower but also because of the highly competitive environment they come from, thanks to Bitcoin mining difficulty and halvings.
It's no wonder that a shift from pure bitcoin mining to hybrid data center operations is already underway: Australia's Iris Energy (Nasdaq: IREN) announced a partnership with WEKA last October to provide both the storage and GPU stack for its generative AI.
Bernstein analysts recently predicted that Iris will shift 15% of its power capacity to AI data centers. In June, bankrupt Core Scientific (Nasdaq: CORZ) launched a similar co-hosting model after restructuring from bankruptcy. The company signed a 12-year deal with AI startup CoreWeave to leverage 200 MW of power capacity for AI HPC operations.
During that period, Core Scientific is expected to generate $3.5 billion in revenue in addition to its bitcoin mining operations, which rely on the spot price of BTC. Again, such hybrid strategies boost bitcoin revenues.
By leveraging AI data center operations, more Bitcoin companies are less likely to go bankrupt in a bear market, which will reduce the pressure to sell BTC. As a result, the healthy currency will become healthier year by year. In the long term, it's not hard to see the trajectory. Hybrid data centers can help companies manage digital assets on the one hand, and become the center of a healthy currency on the other.
Conclusion
Fueled by the promise of cognitive automation, the global economy is gaining another layer on top of digital: the high-performance computing (HPC) layer.
Just as Bitcoin relies on energy-hungry HPC infrastructure to make the money tamper-proof, AI data centers are paving the way for new jobs and a surge in productivity. As sophisticated Bitcoin mining operations scale up to GPU stacks alongside ASIC stacks, the merging of the two will be inevitable.
In doing so, an incentive feedback loop is created: Bitcoin miners’ excess energy capacity flows to load response credits and energy-hungry AI companies. Combined with AI agents capable of executing autonomous BTC microtransactions, the synergy creates the exciting beginnings of Hyperbitcoinization.
Latest Alpha Market Report
Source link
