Over the past two quarters, ten of the twelve publicly listed Bitcoin miners tracked by CoinShares have repriced between 70% and 195%. That is not a mining rally. That is an AI infrastructure repricing dressed in mining clothing. And on October 12, one Amsterdam-listed operator added €800 million of institutional capital to the pile at a €6.1 billion implied valuation — bringing cumulative equity raised since March to more than €1 billion in roughly seven months.
I audit the code, not the charisma. Here there is no code. So I audited the capital structure, the power contracts, and the contract-to-revenue ratio instead. What I found is a sector-wide gap between what is being signed and what is actually being billed — and a company-level information vacuum that makes the €6.1 billion number impossible to verify.
The Asset Is the Interconnection, Not the Algorithm
Strip away the AI narrative and the actual asset is unglamorous: interconnected power capacity across the US and Europe exceeding 4GW, acquired through a share purchase and restructuring completed in August. The technical work is a retrofit — converting Bitcoin mining sites into AI/HPC data centers. There is no cryptographic innovation here. The moat is the grid interconnection queue, which in many US and European jurisdictions can run years. A new-build data center cannot buy that queue position. A miner that already holds it can repurpose it. That is the entire scarcity argument.
The economics are where it gets uncomfortable. CoinShares data puts AI retrofit capex at $8–15 million per megawatt, against $0.7–1 million per megawatt for a mining buildout — roughly 10 to 15 times the capital intensity. The payoff is roughly 3 times: about $1.5 million annualized profit per megawatt post-retrofit versus roughly $500,000 for mining. Read that ratio carefully. Capex scales 10–15x while profit scales 3x, which means the payback period stretches dramatically and every month of execution delay compounds against the return.
Yields are calculated, not guaranteed. So is retrofit ROI. And here is a detail the announcement does not address: ASIC miners have essentially no residual value in an AI retrofit. This is not a conversion — it is a teardown. Any balance sheet carrying mining hardware at book value faces a potential impairment that no one has disclosed.
The Contract-to-Revenue Chasm
This is where I stop reading press releases and start reading delivery data. Across the covered miner cohort, disclosed AI compute contracts exceed $100 billion. The annualized revenue those contracts currently support is approximately $1.1 billion. That is a roughly 90-fold gap between headline contract value and billed revenue.
On capacity: contracted volume exceeds 4GW, while billed capacity sits near 550MW. Utilization is approximately 14%. The industry is signing far ahead of building. Some of those "contracts" may be framework agreements or letters of intent — non-binding instruments that inflate the headline and evaporate on delivery. I have seen this structure before. In 2021, crypto protocols reported TVL that included incentivized deposits with no organic demand. When emissions stopped, the TVL vanished. Liquidity dries up faster than hope. Contract value behaves the same way when it is not backed by billed capacity.
The valuation math reflects the gap, not the delivery. AI infrastructure comparables trade near 19.4x forward sales. Pure mining comparables trade near 6.6x. That 3x premium is entirely a bet on the pivot succeeding. Every dollar of that premium is priced against revenue that has not yet been billed.
What the Filing Does Not Say
Here is the contrarian read, and it is not about AI demand — that is real. It is about disclosure quality.
The company is publicly listed in Amsterdam. Listed entities operate under disclosure obligations. Yet the identity of the institutional investors behind the €800 million is undisclosed. The allocation of those proceeds between GDA retrofit, general operations, and debt service is undisclosed. The retrofit timeline, cost curve, and revenue schedule are undisclosed. There is also an unresolved legal thread: the FTX/Alameda recovery trust previously sought to claw back funds transferred to GDA. A dismissal was recorded — dated June 2026, which I flag as a timeline anomaly requiring verification, because a future-dated dismissal in a current financing story is either a data error or a signal the timeline extends further than presented.
Verify the source, trust no one. When a listed company publicizes the size and valuation of a raise but withholds the counterparties, the use of proceeds, and the schedule, the disclosure is optimized for the equity narrative, not for the investor. Seven months, over €1 billion raised, integration of a major acquisition completed in August, and a fresh raise announced in October. That cadence suggests capital consumption running ahead of internal cash generation. Financing-driven growth is not fraud, but it is fragile. It depends on the capital market staying open and the narrative staying hot.
The bull case rests on real power assets in real interconnection queues. The bear case rests on the fact that this team has no disclosed AI/HPC operating history, no disclosed GPU supply relationship, and no disclosed milestone. Execution capability is the single largest unverified variable, and it is the one that determines whether the 19.4x multiple holds or reverts toward 6.6x.
Diversification is the only safety net — and here, the exposure is concentrated in one narrative, one execution team, and one financing channel.
Levels to Watch
Track two numbers, not the price. First, billed megawatts: if 550MW does not move materially toward the 4GW contracted base within two quarters, the contract-to-revenue gap is structural, not timing. Second, disclosure: if the next filing still omits investor identity and capex allocation, treat the €6.1 billion valuation as narrative, not fundamentals. The trade is not long or short the pivot. The trade is long verified delivery and short unverified announcements. Watch which one arrives first.
Strategy beats speculation every time. The miners who convert power into billed compute will survive the repricing. The ones who converted press releases into valuation will not.