NextFin News - The crypto-treasury trade is losing its premium, and a growing number of public companies are trying to replace it with an AI infrastructure story before investors walk away entirely. What began as a balance-sheet bet on digital assets is now turning into a race to own power, data-center capacity and GPU revenue, because the market no longer seems willing to pay for crypto hoards alone.
The shift is visible in company disclosures. Bitdeer Technologies Group said in a June 2026 filing that its AI Cloud annualized run-rate revenue was about $69 million at 90% GPU utilization, while its May update showed 921 BTC mined and roughly 3.0 GW of global electrical capacity across online and pipeline projects. Bit Digital said in a March 2026 filing that it had continued its transition toward an Ethereum-focused treasury strategy while maintaining AI infrastructure exposure through WhiteFiber, its majority-owned public company. In both cases, the pitch has moved from token accumulation to operating assets.
That matters because it changes what investors are being asked to underwrite. The treasury model depended on rising coin prices and a market willing to treat the balance sheet itself as the product. The AI model asks for a different judgment: can a miner-converted data-center operator turn electricity, land and cooling into durable contracted revenue? The answer is not the same across companies. Some have real assets and real customer demand. Others are simply attaching the hottest market label to a model that was already under pressure.
The central question is whether this is a cyclical rebrand or a structural pivot. The near-term move is cyclical: firms are rushing into AI because the treasury trade is being repriced and capital is tighter. But the underlying change looks structural, because the assets that once supported mining - power interconnects, large sites, cooling and engineering - now have a second life in high-performance computing. That does not guarantee success, but it does mean the business model can evolve rather than merely fade.
In practical terms, the market is shifting from asking “How many coins do you hold?” to “How much power can you monetize, and for how long?” That transition is important because it replaces mark-to-market exposure with operating leverage. It also creates a second-order effect: as more treasury companies chase AI, the market may start valuing the scarcity of real power and signed capacity agreements instead of the mere promise of a repositioning. The winners will be the firms that can prove conversion, not just announce it.
Bitdeer’s disclosures provide a useful example of the mechanism. In its June 2026 filing, the company said it had roughly $69 million of AI Cloud ARR and 90% GPU utilization. In the same period, it said its mining and infrastructure footprint included 65.5 EH/s of self-mining hashrate, 8.4 EH/s of co-mining hashrate, 218,000 self-owned rigs and 3.0 GW of global electrical capacity. That combination shows how the market narrative is changing: investors are no longer being asked to value only mined Bitcoin, but also the capacity to host compute and sell recurring services.
Bit Digital’s filing shows a different route to the same conclusion. The company said it had concentrated capital and staking activity around Ethereum and was maintaining exposure to AI infrastructure through WhiteFiber. That is not a pure treasury business anymore. It is an attempt to turn a crypto-linked balance sheet into a broader asset-allocation platform with a compute arm attached. Whether that succeeds will depend less on the next token move than on execution, contract quality and dilution discipline.
The revaluation problem is straightforward. Treasury companies are often priced as leveraged exposures to volatile digital assets, but leverage cuts both ways. If coin prices stall or financing costs rise, the equity premium disappears quickly. AI infrastructure offers a way out because contracts and utilization can support a steadier cash-flow story. But that story only works if the company can fill the space it builds, and do so before capital costs or competitive pressure erode returns.
That is why the pivot is both compelling and fragile. It is compelling because AI demand is real, and the power-rich sites miners already own are hard to replicate. It is fragile because the same firms now chasing AI are also the ones most exposed to the pressure that made the treasury model less attractive in the first place. The market is effectively asking whether these businesses can become industrial infrastructure owners before they become stranded balance-sheet stories.
Why The Treasury Premium Collapsed
The treasury model worked when investors were willing to pay for direct exposure to digital assets plus optionality on higher prices. That premium became harder to defend when the market began to focus on dilution, financing and the absence of operating cash flow. Once that happened, simply holding more crypto stopped looking like a strategy and started looking like balance-sheet beta.
The mechanism is simple. A company that owns a large pile of Bitcoin or Ether can look exciting when token prices rise faster than financing costs. But if the company repeatedly issues shares, converts debt or spends capital just to maintain that exposure, shareholders are effectively underwriting a levered trade with corporate overhead attached. When the token side stops compounding fast enough, the valuation gap closes.
That is the environment in which AI became attractive. In crypto mining, the scarce input was hash rate; in AI infrastructure, the scarce input is power and capacity that can be sold into a market with multi-year demand. The shift is not just a label change. It is a change from owning a volatile financial proxy to owning physical infrastructure that can be leased, contracted or cloud-sold.
Bitdeer is useful because it already straddles both worlds. The company said in May and June 2026 disclosures that it mined 921 BTC in May, held around 171 BTC at month-end, and had AI Cloud ARR of roughly $69 million at 90% utilization. It also said its global electrical capacity was about 3.0 GW, including projects in Norway, Ohio and elsewhere. Those numbers matter because they show why investors may still be willing to listen: there is an actual operating base beneath the AI pitch.
But the same facts also show the limits of the move. Even a well-positioned company needs customers, execution and funding. If a data center has capacity but no tenant, the asset remains a cost center. If a company has a GPU cloud but low utilization, the economics fall apart. If capital is raised too aggressively, the stock can lose value even if the project eventually works. That is why the pivot is not a free lunch.
The better comparison is not between crypto and AI as industries. It is between two forms of scarcity. The old model relied on scarcity of coins. The new model relies on scarcity of usable power and deployable compute. Those are not interchangeable. One is priced by market sentiment. The other is priced by infrastructure economics.
Why The AI Story Has Real Substance
The strongest argument for the pivot is that AI demand is not a temporary narrative. It is a structural buildout of compute capacity, and that buildout needs the very assets many crypto miners already control. Large power blocks, grid access, cooling systems and industrial sites are hard to assemble quickly. That gives miners a real asset advantage.
This is why the second-order implication matters. The first-order view is that AI is a better story than the treasury model. The second-order view is that the market will start sorting winners by asset quality rather than by how loudly they announce the pivot. A company with a signed tenant, high utilization and a credible power pipeline will likely receive a different valuation than a company with only a rebrand and a slide deck.
That distinction turns the story from narrative to mechanism. A miner that converts a 50 MW or 100 MW site into contracted AI colocation changes the revenue profile from highly reflexive to more recurring. Once the company has tenant demand, the asset can support financing on better terms. That lower cost of capital can then support further buildout. The feedback loop is what makes the move potentially structural.
Bitdeer’s reported 90% GPU utilization is the kind of metric investors now care about because it signals that capacity is being used rather than merely installed. Its 3.0 GW footprint is equally important because scale matters in a market where power availability is often the binding constraint. Bit Digital’s WhiteFiber ownership matters for a different reason: it offers exposure to AI infrastructure even as the company shifts its treasury mix. In both cases, the market is being asked to believe that the company has more than one way to monetize its assets.
The strongest counter-thesis is that this is still mostly a financing workaround. Skeptics would say the AI pivot is what companies say when the treasury story stops working, and that many firms will still struggle to produce enough contracted revenue to justify the capital they are deploying. That is a fair challenge because it targets the core of the thesis: whether the shift creates value or only postpones the reckoning.
“In parallel, our self-mining hashrate continues to improve, reaching approximately 65.5 EH/s, up over 400% year-over-year, while our co-mining hashrate has grown to an incremental 8.4 EH/s, reinforcing Bitdeer’s position as one of the largest publicly listed Bitcoin miners by total hash rate under management and reflecting the continued ramp-up of our SEALMINER fleet.”
That statement shows why the market is willing to listen: these are not empty firms, and their infrastructure has scale. But the falsifying signal is just as clear. If AI cloud revenue, utilization and contract wins do not keep rising over the next two to four quarters while capex and dilution remain elevated, the market will likely conclude that the pivot is a cyclical disguise rather than a durable reset.
For now, the best reading is that the AI shift is structurally real but cyclical in its timing. Companies are moving because the treasury premium vanished, not because they suddenly discovered compute in a vacuum. Yet the assets they are buying, building or converting can outlast the current market mood if they are attached to actual demand.
What It Means From Here
The near-term beneficiaries are the firms with real power, existing sites and early AI revenue. They can present themselves as infrastructure owners rather than pure crypto proxies, and that gives the market a clearer model to value. The exposed names are the firms that lack tenant demand, need repeated capital raises or have only a weak path from mining to cloud economics.
In the short term, the pivot can support sentiment because it aligns these companies with one of the market’s strongest themes. In the medium term, the key test is whether AI-related ARR, utilization and contract duration improve faster than capital spending. In the long term, the only firms likely to keep a premium are those whose revenue can stand on compute demand even if crypto prices stay range-bound.
The base case is a split outcome. A handful of companies will use power assets, data-center conversions and GPU clouds to become credible AI infrastructure businesses. Others will fail to convert capacity into durable revenue and will remain trading vehicles tied to crypto volatility. That split will matter more than any one headline because it will determine whether the category survives as a business model or just as a temporary rebranding exercise.
The upside case is that AI leasing and cloud services scale quickly enough to replace the lost treasury premium, allowing the best-capitalized operators to monetize their sites at better returns than mining ever offered. The downside case is that financing tightens, projects slip and investor patience runs out before the AI revenue becomes meaningful. If that happens, the pivot will be remembered as an expensive attempt to escape a broken trade.
The signals to watch are specific: AI ARR, GPU utilization, lease signings, power additions, debt terms and dilution. If those numbers improve, the pivot can justify itself. If they do not, the market will stop treating the AI label as a solution and start treating it as an apology.
The old trade was about owning coins. The new one is about owning wattage. That is a more serious business - and a much harder one to fake.
