Elon Musk’s SpaceX has turned one of many world’s largest synthetic intelligence clusters right into a industrial compute product, creating a brand new problem for Bitcoin miners racing to recast themselves as AI infrastructure corporations.
Anthropic stated it reached a deal to make use of the total computing energy of SpaceX’s Colossus 1 facility in Memphis, Tennessee, giving the Claude maker greater than 220,000 Nvidia processors and 300 megawatts of recent capability inside a month.
The added capability helped Anthropic double Claude Code fee limits for paid plans, take away peak-hour utilization caps for Professional and Max accounts, and sharply improve developer request quantity for its Claude Opus fashions.
The settlement offers SpaceX a marquee AI buyer because it tries to indicate traders that its infrastructure ambitions lengthen past rockets and satellites.
It additionally lands instantly out there Bitcoin miners have been making an attempt to enter: the race to safe energy for knowledge facilities for AI companies that want electrical energy quicker than the grid can ship it.
For miners, the issue is now not solely Bitcoin’s worth, community problem, or the subsequent halving. The brand new query is whether or not they can compete with expertise giants, neoclouds, and Musk-linked infrastructure platforms within the race to transform electrical energy into AI income.
Miners transfer towards compute
Bitcoin miners have spent the previous yr arguing that their future will probably be formed much less by block rewards and extra by powered websites, long-term leases, and AI compute demand.
That shift accelerated after the 2024 Bitcoin halving, which minimize the block subsidy paid to miners and tightened an already troublesome margin construction.
CoinShares stated the fourth quarter of 2025 was essentially the most troublesome interval for miners for the reason that halving, as Bitcoin’s worth correction and near-record hashrate pushed hashprice to five-year lows.
The agency stated hash worth fell additional within the first quarter to about $29 per petahash per second per day, extending stress on operators with older machines and better energy prices.
Consequently, BTC mining economics have pushed a number of public miners towards AI and high-performance computing.
CoinShares stated listed miners might generate as a lot as 70% of their income from AI by the tip of this yr, up from roughly 30% in the present day. The agency additionally stated that public miners have introduced greater than $70 billion in combination GPU colocation and cloud service agreements with hyperscalers and AI prospects by way of 2025 and early 2026.
That transition is already seen within the sector’s company map. BTC miners like TeraWulf, Core Scientific, Cipher, and Hut 8 have more and more grow to be data-center operators that also mine Bitcoin.
Different miners, together with IREN and Bitfarms, are utilizing mining as a bridge into high-performance computing, whereas some operators stay extra intently tied to Bitcoin mining and low-cost power methods.
The excellence has grow to be central to investor valuations. CoinShares stated miners with secured HPC contracts commerce at enterprise-value-to-next-12-month gross sales multiples of 12.3 occasions, in contrast with 5.9 occasions for pure-play miners.
The result’s a sector cut up between infrastructure corporations with AI publicity and mining corporations whose earnings nonetheless transfer extra instantly with Bitcoin’s worth and hash worth.
Energy turns into the commerce
In the meantime, the miner pivot has gained traction as a result of AI demand has uncovered a bottleneck that mining corporations perceive higher than most: entry to large-scale electrical energy.
AI builders want chips, however chips are solely helpful when they are often put in in services with energy, cooling, and grid connections. That has shifted market consideration towards energized websites able to supporting dense computing hundreds.
Artemis, a blockchain evaluation agency, has argued that the AI commerce could also be extra about energy than chips, pointing to a projected roughly 50-gigawatt US data-center energy deficit by way of 2028.
The agency additionally described BTC miners reminiscent of IREN, Core Scientific, and TeraWulf as AI infrastructure corporations hiding in plain sight.
On the similar time, Artemis famous that the Bitcoin miner AI theme rose 56% over the previous month, forward of baskets tied to AI chips, knowledge facilities, energy, and different infrastructure segments.
That worth motion displays a market more and more keen to worth miners for his or her energy portfolios fairly than just for their Bitcoin manufacturing.
Modular Capital’s analysis factors to the identical constraint. The agency stated AI workloads require sustained high-density energy at a scale that the present grid interconnection course of can’t ship rapidly.
It estimated that knowledge facilities, which now account for about 3% to 4% of whole US grid consumption, might attain 12% by 2028 as hyperscaler capital expenditure runs close to $650 billion this yr.
The grid queue makes the shortage extra acute. Modular stated large-load interconnection timelines can stretch 4 years or extra, whereas ERCOT, the Texas grid operator, has roughly 458 gigawatts of pending purposes.
In PJM, the grid area protecting Virginia, Ohio, Pennsylvania, and far of the Northeast, a brand new large-load interconnection is broadly stalled after out there provide capability fell 20% over 4 years. Massive transformers can take two to a few years to acquire, and substations for hundreds above 100 megawatts can add one other 18 to 24 months.
These delays clarify why BTC miners have grow to be engaging candidates for AI infrastructure. A lot of them had secured energy contracts earlier than the AI buildout intensified. Some have already got land, interconnections, and working expertise with industrial-scale power use.
Nevertheless, a mining web site nonetheless wants vital work earlier than it will probably host superior AI workloads, however essentially the most precious asset could also be its place within the energy queue.
Musk enters the race
SpaceX’s Colossus deal adjustments the aggressive map as a result of it exhibits that the ability commerce is attracting corporations with deeper capital swimming pools and broader expertise platforms.
Neocloud operators purchase or lease giant swimming pools of GPUs and hire computing capability to AI builders. Bitcoin miners have been making an attempt to maneuver into that market by providing powered shells, colocation, and, in some circumstances, cloud providers.
Musk’s ecosystem can strategy the identical market from one other angle by constructing huge AI clusters for inner use, then leasing capability when workloads shift elsewhere.
For context, Musk reportedly stated SpaceX had moved its AI coaching efforts to Colossus 2 and would supply computing capability to different AI corporations making comparable efforts to favor humanity.
This remark suggests Colossus 1 grew to become out there as a result of SpaceX’s personal coaching work had already moved to a more recent web site, permitting the corporate to monetize an current asset with out abandoning its broader AI ambitions.
That may be a totally different form of competitors for BTC miners. A transformed mining web site might provide low cost energy and quicker time-to-market than a brand new data-center undertaking. Colossus gives speedy scale, a frontier AI buyer, and a platform tied to Musk’s broader ambitions in AI, house, and infrastructure.
Anthropic additionally stated it’s keen on working with SpaceX on a number of gigawatts of orbital knowledge facilities, a long-range idea that will use solar energy in house and require main technical and capital commitments.
That broadens the aggressive area for BTC miners as they’re now not pitching AI conversion solely in opposition to different miners. They’re competing with hyperscalers, neoclouds, power builders, infrastructure funds, and expertise platforms that may construct or reallocate capability at huge scale.
