Bitcoin Mining to AI: SWI Group’s €800M AI Data Pivot 🚀
Discover SWI Group Genesis Digital’s €800M pivot from bitcoin mining to AI data centers—what it means for AI data center infrastructure, power access & institutional investment.

Why SWI Group Is Turning a Former FTX-Backed Bitcoin Miner Into AI Data Centers
SWI Group is raising €800 million to expand its digital infrastructure business - and a former FTX-backed bitcoin miner is part of the plan. The Amsterdam-listed investment firm says it will repurpose U.S. and Swedish facilities belonging to Genesis Digital Assets for AI workloads, while selling the miner's sites in Kazakhstan.
This is more than a corporate makeover. Bitcoin mining operators have spent years building out sites around electricity, cooling, and network capacity. Those same ingredients can also matter for AI and high-performance computing customers, whose demand for data centers has been rising fast. For miners facing volatile bitcoin revenue, that can look like a path to steadier contracts - but the move isn't automatic. Facilities, power arrangements, and the overall economics still have to work for the new workload.
Below, we break down what SWI has raised, what it plans to do with Genesis Digital, and why the miner-to-AI pivot is drawing attention from investors. We also dig into the practical questions behind the headline: what can be reused, what likely needs upgrades, and what would make the strategy succeed - or fail.
What SWI Group raised and acquired
SWI Group said it raised €800 million, or about $894 million, from institutional investors to expand its data-center and artificial intelligence projects in Europe and the United States. The company says the capital increase values it at €6.1 billion. That figure reflects the valuation implied by the raise, not a separate statement about how much cash is going specifically into converting Genesis Digital.
Genesis Digital Assets - the bitcoin mining business at the center of the plan - joined SWI's portfolio in 2026. In June, SWI announced it would buy about 40% of the company for $500 million. The transaction closed in August, giving SWI 70% of the voting rights. The business was later renamed SWI Digital.
It matters that SWI has voting control without necessarily owning an equivalent share of every economic interest. For readers trying to understand how much control SWI really has, the 70% voting stake helps explain why it can steer decisions. But it doesn't, on its own, tell the full story of ownership, governance, or how proceeds will be allocated going forward.
The acquisition also comes with an FTX-related history. Alameda Research - owned by former FTX CEO Sam Bankman-Fried - invested more than $1 billion in Genesis Digital starting in 2021 and reportedly held about 20% of the miner. After FTX filed for bankruptcy in November 2022, the estate reviewed that investment. According to the reporting behind this deal, the estate dropped a $1.15 billion clawback suit against Genesis Digital after SWI took control. That background is part of why the transaction has been covered beyond pure infrastructure strategy.
So far, SWI hasn't described the €800 million raise as a dedicated conversion budget for Genesis Digital. The company says the money is meant to support the group's broader data-center and AI expansion. It plans to use Genesis Digital's U.S. and Swedish facilities for AI workloads and to sell the Kazakhstan sites. The announcement doesn't spell out how much capital will go to specific locations, new equipment, or power upgrades.
That missing detail is significant. A big fundraising headline tells you SWI can access capital. It doesn't tell you whether any particular site can be converted on terms that make sense. The real questions are operational: which facilities have suitable power and cooling, how quickly they can be upgraded, and whether customers' contracts can cover the cost.
Why bitcoin miners are targeting AI workloads
Bitcoin mining and AI computing are different businesses, but they overlap on key infrastructure needs. Both require reliable access to electricity, data-center space, cooling, and day-to-day operational discipline. Miners have built facilities and secured power connections to run specialized machines around the clock. For some operators, those sites could become a starting point for serving AI or high-performance computing customers.
The business case also has to do with revenue predictability. Bitcoin mining revenue depends on the price of bitcoin, network competition, energy costs, and mining difficulty. But a site's power bill and other operating expenses don't automatically fall when mining revenue drops. Long-term hosting or capacity contracts can change the revenue profile, depending on the terms - and on the customer's credit quality.
That said, AI hosting isn't automatically "better." The drivers are different. A miner looking at the switch needs to model expected returns after conversion costs, hardware choices, power pricing, customer requirements, and the risk of relying on one or a small number of customers. A predictable contract is still a bad deal if it doesn't cover what you have to invest and what it costs to keep operating.
There's also plenty of industry interest. Reporting on this shift says bitcoin miners have signed more than $70 billion in AI and high-performance computing contracts as they look beyond mining. Still, that number shouldn't be treated as a guarantee that every miner has landed profitable work or that all announced capacity is already live and earning money.
SWI's strategy fits a wider trend among publicly traded miners, including TeraWulf and Core Scientific, to turn power access and data-center capacity into AI hosting businesses. The advantage isn't just "having land" or "owning equipment." It depends on whether an operator can deliver the power, uptime, cooling, and technical environment that customers want - at a price that still leaves room for margin.
SWI CEO Max-Hervé George told the Wall Street Journal that he had "zero interest" in bitcoin mining. That's a blunt signal of intent: SWI seems to be treating these sites as digital infrastructure, not as a base for continuing bitcoin production. Selling Kazakhstan and focusing on the U.S. and Sweden makes the direction clearer, even though the company hasn't published detailed site-by-site conversion plans in the material referenced here.
The asset is not just the mining equipment
Mining machines are designed for a specific job. You don't "convert" them into AI servers with a software flip. In many cases, the more transferable assets are the site itself - especially the power connection, electrical infrastructure, cooling systems, the operating team, and local approvals. Even then, those assets only help if they align with what customers require and can be upgraded without turning the project into a money pit.
A facility built around a large fleet of bitcoin mining rigs may have a different layout, power density, cooling architecture, or resilience plan than a site intended for AI workloads. Some infrastructure may be reused; other parts may need significant changes. The economics depend on the condition and design of each facility - not on a sweeping assumption that all mining capacity can be repurposed.
That's why the phrase "miners are becoming AI companies" can be misleading as a shortcut. Some firms may end up as data-center operators or hosting providers. Others may sell or lease sites, sign power agreements, or continue mining while developing a separate AI business. The right model depends on what's already in place, what conversion will cost, and whether customers can be secured in time.
What changes when a mining site becomes an AI data center
Shifting from mining to AI is an infrastructure project, not a simple rebrand. Mining operations run optimized stacks built around purpose-built machines. AI workloads can require different equipment, higher-density racks, different cooling needs, and tighter expectations around availability and service. The exact requirements vary by customer and workload, so a facility that works for one use case may not work for another.
Power is the first and biggest constraint. A strong grid connection can be valuable, especially in markets where new capacity is hard to secure. But the headline number of megawatts isn't enough. Operators need to know what power is actually available, what the terms are, what the price is, and how reliable it is. A connection that looks attractive on paper may become less useful if upgrades, restrictions, or delays limit what can be delivered.
Cooling is the next major issue. Bitcoin mining sites often rely on relatively straightforward cooling approaches matched to their equipment and operating model. AI systems can generate a lot of heat, and cooling requirements depend on the hardware and configuration. Converting a site may mean installing new cooling equipment, changing airflow designs, or adding liquid-cooling capability. Those upgrades cost money - and can slow readiness for customers.
Then there's the customer and the contract. A data-center operator needs to be clear about what the customer is buying: powered space, managed hosting, a specific amount of computing capacity, or a broader service package. Each model shifts responsibilities and capital needs. Contract terms - length, pricing structure, uptime obligations, and provisions around equipment or power-cost changes - can matter as much as the overall size of the deal.
A practical conversion checklist
For an investor or operator looking at a miner-to-AI plan, the sequence should be pretty direct:
- Verify the power position. Confirm actual capacity, pricing, reliability, grid constraints, and the cost of any required upgrades.
- Assess the facility against a defined workload. "AI-ready" isn't a spec. Compare electrical design, cooling, network, and physical layout to what a specific customer needs.
- Build the full conversion budget. Include equipment, construction, cooling, network connectivity, permitting, staffing, and operating costs during the transition.
- Secure customer demand before scaling. A signed agreement with clear scope and economics is stronger evidence than generic "interest" in AI demand.
- Model the downside. Test delays, higher energy costs, lower utilization, and customer concentration risk. If the plan only works in a best-case scenario, it's fragile.
The order matters because infrastructure capital gets committed long before revenue starts. If you build first and only then hunt for customers, you can end up with a technically capable facility that sits underused. On paper, a grid connection can make a site look attractive. In practice, the question is whether the conversion can turn that advantage into a contract with returns that justify the investment.
SWI's plan to focus on U.S. and Swedish locations while divesting Kazakhstan suggests portfolio selection, not a universal judgment that every market is equally ready. Local power conditions, regulation, customer proximity, and asset quality can differ. The source information doesn't provide enough detail to rank specific facilities, so it would be premature to declare which sites are best positioned.
The risks and questions investors should watch
One risk is assuming AI demand will soak up every available site. Demand for computing is real, but individual projects can still miss customer requirements or fail to deliver acceptable returns. The real question isn't whether AI needs data centers. It's whether SWI can convert the specific assets it's buying into capacity customers will pay for - under terms that make the investment worthwhile.
Another risk is capital intensity. Repurposing a site isn't always as simple as swapping servers. Electrical systems, cooling systems, and network setups may need upgrades, and construction can take time. If costs rise or delivery slips, the business can carry expenses before it earns meaningful revenue. SWI has raised a lot of money, but the announcement doesn't explain how the capital will be allocated across projects - or how much of the Genesis Digital estate needs investment.
Power is both an advantage and a constraint. Data-center economics can be sensitive to electricity prices and availability. Competition for power can make suitable locations more valuable, but it can also push costs higher and complicate expansion. Operators should separate what exists today from what's actually reachable as capacity requirements grow.
Contracts also need careful reading. Long-term agreements can stabilize revenue, but they can also create strict obligations around uptime, service levels, and delivery dates. Customers may hold leverage - especially if they bring equipment or represent a large share of a site's expected revenue. A headline contract value won't tell you the margins, payment protections, or who pays if requirements change.
Then there's execution and governance. SWI has majority voting control of SWI Digital. At the same time, Genesis Digital's history includes Alameda Research's prior investment and scrutiny after FTX's bankruptcy. The estate's reported decision to drop its clawback suit changes part of the legal backdrop described in the reporting, but it doesn't remove the need for investors to understand ownership structure, remaining obligations, and what financial disclosures exist.
Finally, there's strategic focus. SWI's CEO has said the company has no interest in bitcoin mining, and the plan is to sell the Kazakhstan sites. A decisive pivot can make priorities clearer. It also means SWI is choosing to lean more on digital-infrastructure and AI economics than on mining. That choice should stand on its own merits, not just as a way to "move from one hot sector to another."
A practical way to track future updates is to separate four milestones:
- Capital raised: Is the money available, and what projects is it meant to support?
- Capacity converted: Which sites have finished the technical work?
- Customers contracted: What capacity is committed, for how long, and under what commercial terms?
- Revenue delivered: Are the facilities operating and generating cash flow on the expected economics?
Progress on one milestone doesn't automatically mean progress on the others. A funding round isn't a completed data center, and a converted site isn't automatically profitable.
Conclusion
SWI's fundraise puts substantial institutional capital behind a clear shift: treat Genesis Digital's facilities as digital infrastructure for AI rather than as a bitcoin mining operation. SWI's control of SWI Digital, its plans for the U.S. and Swedish sites, and its intent to sell Kazakhstan give the pivot a defined direction.
The harder test is execution. Power access can help, but it only becomes a lasting advantage if the site meets customer requirements and produces returns after conversion and operating costs. For investors, the most useful signals are specific: disclosed spending, facilities brought online, customer commitments, and realized revenue - not just the size of the raise or the overall growth of AI demand.
If you're evaluating similar crypto-to-infrastructure pivots, start with the site-level economics. Check power terms, conversion requirements, and customer contracts before assuming a mining facility is already ready for AI. For background on the deal, read the original report from CoinDesk.