AI Compute

Grayscale Investments has renamed its Bitcoin Miners ETF (MNRS) into the Grayscale AI Compute ETF (GCPU), replacing an index built on block-solving with one built on data centers and power. The switch matters to institutional allocators because it converts a niche crypto-mining vehicle into a bet on AI infrastructure scarcity, a theme Goldman Sachs projects will draw more than $1 trillion in annual capital spending.

• Grayscale renamed MNRS to GCPU on September 22, 2026, tracking the new Indxx High Performance Computing Index.
• Roughly half of GCPU’s portfolio targets native GPU cloud and AI-hosting firms, the rest are miners and other HPC operators pivoting to AI.
• North American data centers hold about six months of spare capacity, per CBRE, against a two-to-five-year timeline to build new facilities.

• $1T Goldman Sachs’ projected annual AI capital spending figure
• 6 months data center spare capacity versus years-long build times
• 94% Riot Platforms’ 2026 stock gain versus 21% for AI-pivot miners broadly

Grayscale confirmed the change in a September 22, 2026 announcement, retiring the Indxx Bitcoin Miners Index in favor of the Indxx High Performance Computing Index. GCPU will not hold digital assets directly or track any coin’s price, according to the filing. Grayscale is instead packaging equity exposure to the physical buildout behind AI models, a trend first detailed by Cryptopolitan.

Grayscale Splits GCPU Roughly 50-50 Between AI-Native and Ex-Mining Firms

The new index captures what Grayscale describes as two distinct paths to compute capacity. About half the portfolio’s weight targets companies built around GPU cloud and AI-hosting from inception. The remainder consists of high-performance computing operators, including Bitcoin miners, that have publicly disclosed or executed a shift toward AI workloads.

The index rebalances quarterly, meaning miners that formalize AI contracts or expand hosting capacity can gain weight each quarter, while those that do not can lose it. That mechanical rule replaces the previous MNRS methodology, which weighted companies purely on mining output and hash rate regardless of revenue mix.

Vanourny Calls Compute the New Scarce Asset

Steve Vanourny, Grayscale’s Head of Index, framed the rebrand as an extension of the firm’s original digital-asset scarcity thesis rather than a departure from it.

“Grayscale has spent over a decade giving investors access to scarce, transformational technologies, such as digital assets, and we believe compute belongs in that same category. GCPU is a direct extension of that thesis, giving investors access to the companies closing the gap between AI’s demand for compute and the physical infrastructure needed to meet it.”

Steve Vanourny, Head of Index, Grayscale

Grayscale backs that scarcity argument with CBRE data showing North American data centers at record-low vacancy, with roughly six months of spare capacity in reserve. Building fresh capacity is not a fast fix; new facilities typically take two to five years to bring online. Goldman Sachs’ $1 trillion annual capex forecast, mostly directed at physical plant, underpins the case that supply cannot close the gap quickly.

What Changes in Practice for Miners That Cross Over

For crypto-native investors, the practical effect is that GCPU no longer functions as a Bitcoin mining proxy, even though several of its constituents still run mining operations.

Cryptopolitan’s reporting shows why the pivot pays: AI cloud services fetch a median of roughly $940 per megawatt-hour, against $113 to $179 for Bitcoin mining, and miners carrying AI or HPC contracts trade at about 12.9 times forward sales versus 3.7 times for those without.

TeraWulf is up near 73% year to date and Riot Platforms roughly 94%, while AI-integrated miners overall have gained about 21% against a broader Bitcoin market that has struggled. Core Scientific booked $136.7 million in colocation revenue in the second quarter of 2026, more than five times the $27.5 million it drew from mining in the same period.

The document Grayscale filed leaves one question unresolved for holders: what happens to constituents that announce an AI pivot but fail to execute it before the next quarterly rebalance.

OpenAI’s $280 Billion Compute Push Sets the Backdrop

Cryptopolitan’s analysis ties the miner repricing to a roughly $280 billion compute buildout tied to OpenAI and a 2,600-gigawatt U.S. grid interconnection backlog, conditions that make an already-permitted, grid-connected site more valuable than the mining rigs sitting inside it. That dynamic is already reshaping the mining sector beyond Grayscale’s index, with public miners collectively holding an estimated $70 billion to $100 billion in announced AI and HPC contracts.

Grayscale’s own MNRS-to-GCPU conversion effectively formalizes what miner management teams had already been telling shareholders on earnings calls.

The CCS read. Grayscale is not calling a bottom on Bitcoin mining economics, it is arbitraging a repricing that already happened in equity markets, where AI-contract miners trade at more than triple the multiple of pure hashers. For allocators who track bitcoin treasury strategies like the one covered in our report on Strategy’s bitcoin purchases, GCPU signals that public markets increasingly separate mining-as-coin-exposure from mining-as-infrastructure-exposure, and reward the latter.

The first quarterly rebalance under the new Indxx High Performance Computing Index methodology will show whether Grayscale’s roughly 50-50 split between AI-native firms and transitioning miners holds, or whether the fund tilts further from Bitcoin mining as more operators disclose AI contracts, a shift already visible in coverage of miners’ financing needs such as our report on crypto-backed lending trends.