A Bitcoin Fund That No Longer Tracks Bitcoin
The Valkyrie Bitcoin Miners ETF (NASDAQ: WGMI) has delivered a 97% return over the past year. Bitcoin, the asset it was built around, has lost 46% of its value over the same period [1]. That gap is not a paradox or a market anomaly—it is the direct result of a strategic transformation quietly underway inside the fund's largest holdings. The miners have stopped being miners.
What WGMI now holds, in practice, is a basket of companies that have pivoted from validating blockchain transactions to becoming landlords for artificial intelligence infrastructure. Understanding that distinction is no longer optional for anyone with money in the fund.
Mining Economics Forced the Pivot
The shift did not happen out of ambition alone. The underlying economics of bitcoin mining have deteriorated sharply. Riot Platforms (NASDAQ: RIOT) reported that its cost to mine a single bitcoin reached 70% of that coin's production value last quarter—up from 50% just a year earlier [1][2]. That margin compression, playing out across the sector, made the status quo unsustainable. Mining alone, at those cost ratios, would sink these companies.
AI leasing offered a way out—and a way up. The same physical assets that power bitcoin mining operations—large power contracts, purpose-built facilities, dense electrical infrastructure—turn out to be exactly what hyperscalers need to run GPU clusters. The pivot required a change in tenants, not a change in real estate.
The Deals That Rewrote the Fund's Identity
The scale of the contracts signed by WGMI's top holdings explains why the fund's behavior has diverged so dramatically from bitcoin's price.
Core Scientific (NASDAQ: CORZ) signed a 15-year lease with AMD worth more than $14 billion in base contracted revenue across 530 megawatts [1][2]. The company's total potential contracted revenue across 1.1 gigawatts now exceeds $24 billion [2]. Core Scientific CEO Adam Sullivan framed the opportunity bluntly, telling analysts: "There's a lot of GPUs sitting on the ground, and those GPUs still need to be plugged in" [1].
IREN (NASDAQ: IREN) secured a five-year, $3.4 billion NVIDIA cloud contract alongside a separate NVIDIA equity investment of up to $2.1 billion that vests as GPUs are deployed [1][2]. IREN CEO Dan Roberts confirmed the demand picture from the supply side: "All of our operational capacity is fully contracted" [1]. The company's shares have risen 152% over the trailing year [2].
Riot Platforms added a 20-year, 191-megawatt lease with a frontier AI lab on top of its existing AMD deal, pushing its total contracted data-center revenue to $9.8 billion [1][2]. RIOT shares are up 66% over the trailing year, while CORZ has gained 43% [2].
What Actually Drives WGMI Now
For investors accustomed to using WGMI as a leveraged proxy for bitcoin price movements, the fund's new sensitivity profile is a significant change. The macro variable that matters most over the next 12 months is AI hyperscaler capital expenditure—specifically, the pace at which companies like AMD, NVIDIA, and Microsoft continue writing multi-billion-dollar power-and-compute checks [1][2].
A slowdown in AI capex guidance would hit these stocks directly, because the bull case is now written in 15-year lease agreements rather than hash rates [2]. Conversely, continued hyperscaler spending sustains the contracted revenue streams that have made the fund's top holdings behave more like infrastructure REITs than crypto plays.
There is also an operational variable worth tracking. Core Scientific was awaiting an ERCOT Pecos load-study result as of early August 2026, and delays in that process push back AMD GPU delivery timelines [2]. Texas grid interconnection updates from ERCOT have become a material data point for the fund's performance outlook.
The Fund-Level Signal: Active Management Risk
WGMI is actively managed, which introduces a layer of risk that a passive index fund would not carry [2]. The 97% trailing-year return was generated by the fund's concentration in AI-pivoted names. If the manager rotates toward pure-play bitcoin miners—perhaps chasing mean reversion in BTC price—the fund's correlation with bitcoin would snap back, and the decoupling that produced the outperformance would reverse with it [2].
Monitoring the fund's monthly holdings disclosures for the relative weight of RIOT, CORZ, and IREN versus pure miners is the most direct way to track whether the AI infrastructure thesis remains intact at the portfolio level [2].
Wrong Vehicle, Wrong Exposure
For investors who entered WGMI seeking straightforward bitcoin exposure, the fund no longer delivers that. A spot bitcoin ETF such as the iShares Bitcoin Trust (NASDAQ: IBIT) provides direct commodity exposure without the AI infrastructure overlay [1][2]. That distinction has never mattered more than it does now: WGMI's biggest holdings are effectively landlords collecting rent from GPU operators, not entities whose fortunes rise and fall with BTC price discovery.
Bitcoin currently trades around $63,000, down 28% year-to-date, while WGMI shares sit near $53, up 38% over the same window [1][2]. Those two numbers, moving in opposite directions, are the clearest possible signal that the fund and the asset in its name have parted ways.
What to Watch Next
The next round of quarterly capex guidance from NVIDIA, AMD, and Microsoft will serve as the primary macro indicator for WGMI's trajectory [2]. Any reduction in projected AI infrastructure spending would undermine the contracted revenue streams that have driven the fund's top holdings. Alongside that, ERCOT interconnection updates—particularly any delays affecting Core Scientific's Texas facilities—could shift delivery timelines and, with them, near-term revenue recognition [2].
At the fund level, WGMI's next published holdings disclosure will reveal whether the manager is maintaining or diluting its concentration in the AI-pivoted names. If either the macro signal or the portfolio composition shifts against the infrastructure thesis, the 46% bitcoin decline that WGMI has so far shrugged off will start to matter again [2].