Bitcoin Miners Are Missing the Crypto Rally as Exchanges and Stablecoins Surge

Bitcoin has climbed roughly 22-25% since August 17, 2026, and the stocks most directly tied to trading and stablecoin activity Coinbase, Bullish, Robinhood, and Circle have moved up with it. Bitcoin mining stocks mostly have not. Among the eleven mining and mining-adjacent companies tracked in a September 9 report from The Block's Data and Insights newsletter, only Canaan (CAN), a mining-rig manufacturer, has outperformed bitcoin during the rally. The other ten posted a median return of just 1.8%, and two of them Core Scientific (CORZ) and TeraWulf (WULF) actually lost ground relative to bitcoin, underperforming by 27% and 24%, respectively.

That gap is the subject of this analysis. It is not a story about bitcoin mining becoming unprofitable, and it is not evidence that the miners are being mismanaged. It is, more precisely, a story about capital rotation: where investor dollars are choosing to flow during this rally, and why mining stocks despite sitting closer to bitcoin's underlying economics than almost any other publicly traded asset are having trouble capturing money that is flowing into bitcoin exposure through other channels.

---

Bitcoin's 22% Rally Has Not Lifted Mining Stocks in Proportion

The starting point for this divergence is straightforward market data. Bitcoin has risen about 22% since August 17, 2026, according to The Block, which cross-referenced pricing data as part of its September 9 report on the trend. Exchange-linked equities such as Coinbase, Bullish, and Robinhood, along with stablecoin-adjacent names including Circle, delivered returns broadly in line with that bitcoin move over the same window.

Mining stocks told a different story. Canaan was the sole exception among tracked names to outperform bitcoin during the stretch. The remaining ten a group that includes Hut 8, Riot Platforms, Marathon Digital, and HIVE posted a median return of only 1.8%, according to the same reporting, corroborated independently by Cryptopolitan and Crypto Briefing coverage of the same dataset. Core Scientific and TeraWulf were the clearest laggards, trailing bitcoin's move by 27 and 24 percentage points, respectively.

The following table summarizes the divergence across the categories tracked in the report.

Category	Representative names	Return relative to Bitcoin's ~22% rally
Exchanges	Coinbase, Bullish, Robinhood	Roughly in line with Bitcoin
Stablecoin-linked	Circle	Roughly in line with Bitcoin
Mining (outperformer)	Canaan	Outperformed Bitcoin
Mining (broad group)	Hut 8, Riot, Marathon, HIVE, and others	Median of 1.8%, underperforming Bitcoin
Mining (worst laggards)	Core Scientific, TeraWulf	Underperformed Bitcoin by 27% and 24%

It's worth being precise about what this data does and doesn't show. The 1.8% median and the individual underperformance figures for Core Scientific and TeraWulf are relative-return statistics tied to a specific measurement window August 17 through early September 2026 not year-to-date or trailing-twelve-month figures. Separately, several individual mining stocks have posted large year-to-date gains earlier in 2026 on the back of AI infrastructure announcements, so the underperformance described here is specific to this rally window, not a claim that mining equities have been weak all year.

---

Miners Sold Billions in Bitcoin to Fund an AI Pivot That Now Competes for Investor Attention

The mechanical explanation for the gap starts with what public miners have been doing with their balance sheets. During the first half of 2026, public miners sold an estimated 28,000 to 32,000 BTC worth roughly $1.78 billion at the time of those sales largely to fund a shift toward artificial intelligence and high-performance computing (HPC) infrastructure. That is a materially different use of capital than simply reinvesting in mining rigs and power capacity.

The scale of that pivot is now substantial. CoinShares, the digital-asset investment firm that publishes a closely watched quarterly mining report, estimated that public miners have announced more than $70 billion in cumulative AI and HPC contracts. Individual deals illustrate the size of the shift: Marathon Digital's (MARA) $1.5 billion acquisition of the Long Ridge power facility, and IREN's $3.4 billion cloud-computing contract with Nvidia, are among the larger commitments. CoinShares has projected that some mining operators could generate up to 70% of their revenue from AI and HPC workloads by the end of 2026, up from roughly 30% earlier in the year.

The mechanism connecting this pivot to weak stock performance during a bitcoin-led rally works roughly as follows: AI infrastructure → bitcoin → data → data centers → capital allocated to compute leases and Nvidia hardware rather than mining fleets → revenue increasingly tied to AI-tenant contracts and hyperscaler counterparties rather than bitcoin's price → investors uncertain whether to value the stock as a bitcoin proxy or a data-center operator → a discount for that ambiguity, rather than a premium for optionality, during a rally driven specifically by bitcoin's price.

That last point is the crux of the story. When bitcoin itself is the asset rallying, the cleanest way to gain exposure is bitcoin, a spot bitcoin ETF, or a company whose revenue is tied almost entirely to bitcoin's price and trading volume which describes exchanges and stablecoin issuers far more than it describes a miner that now derives a large and growing share of revenue from long-duration AI leases. Bitcoin miners with split business models effectively straddle two narratives, and during a bitcoin-specific rally, straddling has proven to be a disadvantage rather than a hedge.

---

Core Scientific and TeraWulf Show the Pattern in Its Most Extreme Form

Core Scientific and TeraWulf are useful case studies because both have leaned furthest into AI infrastructure and both have underperformed bitcoin the most during this rally by 27% and 24%, respectively. Both companies have signed large-scale HPC hosting or lease agreements with technology customers, converting portions of their power capacity from bitcoin mining to AI-oriented data-center use. The operational logic behind that shift is not in question: HPC hosting can offer more predictable, contracted revenue than bitcoin mining, whose income depends on a volatile spot price and a periodically adjusting network difficulty. But predictable AI revenue does not track bitcoin's price, so when bitcoin rallies, that predictable revenue stream does not participate in the upside the way pure mining exposure would.

This is not evidence that the AI pivot was the wrong decision. CoinShares' research and multiple industry reports frame the shift as a survival strategy that helped miners endure a difficult stretch of underlying mining economics a point examined in the next section. It is evidence that the pivot has, for now, split investor attention and made these stocks harder to value cleanly as bitcoin proxies, which matters specifically during bitcoin-driven rallies like this one.

---

Mining Economics Are Recovering, But From a Deep Hole

A second, related explanation is that the core economics of bitcoin mining separate from the AI story have not fully recovered even as bitcoin's price has climbed. Hashprice, the standard industry measure of daily mining revenue per unit of computing power, jumped more than 24% in August 2026. That is a genuine improvement, but it follows a period in which hashprice touched multi-year lows earlier in 2026 amid rising network difficulty and thin margins following the 2024 halving.

Forward-looking pricing data from Luxor, a mining-services and hashrate-derivatives provider, shows the market pricing this recovery as partial rather than complete. Luxor's forward contracts spanning September 2026 through February 2027 have averaged approximately $36.98, which remains roughly 27% below the 2025 average for the same forward-contract product. In practical terms, that means the market for hedging future mining revenue is still pricing bitcoin mining as a meaningfully less profitable business than it was across 2025, even after August's rebound. For a reader unfamiliar with the term, hashprice functions similarly to a "price per unit of work" metric for miners: it combines block rewards, transaction fees, and bitcoin's dollar price into a single daily revenue figure per petahash of computing power, and it is the closest thing the mining industry has to a real-time profitability gauge.

That backdrop matters for interpreting the rally-period underperformance. Investors weighing whether to add mining-stock exposure during a bitcoin rally are not just asking "has bitcoin gone up" they are also asking whether the underlying business generating that exposure has itself become more profitable. With forward hashprice still running below 2025 levels, the answer has been a qualified "somewhat," not a clean "yes," which likely tempers enthusiasm relative to exchanges and stablecoin issuers, whose revenue models (trading fees, and reserve/float income, respectively) scale more directly and immediately with rising crypto asset prices and volumes.

---

Institutional Capital Is Flowing Into ETFs and Exchanges Instead

The other half of the capital-rotation story is where money has actually gone. U.S. spot bitcoin ETFs pulled in roughly $3.8 billion in net inflows over three consecutive weeks through early September 2026, including a single-day inflow of about $731 million on September 3 the largest daily total since mid-January, led by BlackRock's iShares Bitcoin Trust. That inflow run marked the strongest three-week stretch for the ETF category so far in 2026, according to The Block's reporting on the September 3 session.

This flow of capital toward spot ETFs, rather than toward mining equities, illustrates the rotation directly. An investor who wants bitcoin price exposure without picking individual mining operators, evaluating their AI contract books, or underwriting execution risk on multibillion-dollar data-center buildouts can now simply buy a spot ETF. That product did not widely exist in its current form for most of bitcoin's history, and its growing liquidity and scale combined net assets across U.S. bitcoin ETFs have run above $100 billion in 2026 have given capital an easier, lower-friction path into bitcoin exposure than mining stocks offer. It is worth noting that ETF flows are not one-directional: the three-week inflow streak was followed by a stretch of net outflows in the following week, according to fund-flow data tracked by Farside Investors, underscoring that these flows can reverse quickly and should be read as a snapshot rather than a settled trend.

A secondary, more technical signal cited in market commentary during this period is a "death cross" in USDT dominance a pattern in which the 50-day moving average of Tether's share of total crypto market capitalization fell below its 200-day moving average. Some market participants interpret a declining USDT dominance trend as a sign that capital is rotating out of stablecoins and into higher-risk assets like bitcoin, which would be broadly consistent with the rally described here. This should be treated as one technical observation among several, not a confirmed causal driver moving-average crossovers are backward-looking by construction and have an inconsistent record as standalone predictive signals.

---

Risks, Headwinds and Open Questions Facing the Mining Sector

Several distinct risks are worth separating out, since they operate through different mechanisms and carry different levels of certainty.

Execution risk on AI buildouts is a confirmed, ongoing risk. Miners transitioning fleets and facilities toward HPC hosting are undertaking capital-intensive construction and long-duration contracting with hyperscale technology customers. Delays, cost overruns, or counterparty issues in these buildouts could weigh on the stocks independent of bitcoin's price, and would do so precisely because the AI narrative is now central to how these companies are valued.

Dual-narrative valuation ambiguity is a structural, ongoing risk. As long as a mining company generates a meaningful share of revenue from AI/HPC contracts and a meaningful share from bitcoin mining, investors must value it against two different comparable sets data-center operators and pure-play miners that often carry different valuation multiples and different investor bases. This ambiguity is the most direct mechanism connecting the AI pivot to the rally-period underperformance documented above.

Bitcoin sales to fund the pivot represent an emerging, monitorable risk. The estimated 28,000-32,000 BTC sold by public miners in H1 2026 reduced their on-balance-sheet bitcoin holdings, a asset base that has historically been part of the investment case for holding miner equity as a leveraged bitcoin proxy. Continued sales to fund further AI capital expenditure could further dilute that thesis, though the extent of any future sales is not yet known and would depend on financing needs and capital-markets access going forward.

Underlying mining economics remain a scenario-dependent risk. With Luxor's forward hashprice curve still running roughly 27% below 2025 levels, the pure-mining side of these businesses continues to operate on thinner margins than in the recent past. A renewed hashprice decline from rising network difficulty, falling bitcoin prices, or rising energy costs would pressure the segment of revenue not covered by fixed-price AI contracts.

The USDT dominance "death cross" is a speculative, technical-level signal, not a confirmed fundamental driver, and should be weighted accordingly by readers trying to assess the durability of the broader rally.

None of these risks amounts to evidence that bitcoin mining is structurally unprofitable or that the AI pivot was a mistake; CoinShares' research frames the shift as having helped several miners survive a difficult stretch of the bear market. The risks instead explain why, in the specific context of a bitcoin-price-driven rally, mining equities have a harder time translating bitcoin's gains into their own share prices than businesses more directly levered to bitcoin trading and stablecoin activity.

---

What This Divergence Means for How Investors Read Mining Stocks

The practical takeaway is less about any single company and more about how to interpret bitcoin mining equities going forward. As a growing share of the sector's revenue shifts toward AI and HPC infrastructure, mining stocks are increasingly behaving as hybrid instruments part bitcoin proxy, part data-center operator rather than as clean, leveraged bets on bitcoin's price. That hybrid status can work in miners' favor during periods when AI infrastructure demand or individual contract announcements are the dominant story, as several miners' earlier-2026 year-to-date gains show. It has worked against them during this particular rally, where bitcoin's price move itself is the primary catalyst and capital has favored assets and equities most directly tied to that move.

For readers trying to track this story going forward, several data points are worth monitoring: whether Luxor's forward hashprice curve continues closing the gap to 2025 levels, whether spot bitcoin ETF flows sustain their recent pace or continue reversing as they did in the second week of September, whether individual miners report AI/HPC revenue actually materializing at the scale of their announced multibillion-dollar contract backlogs, and whether the USDT dominance technical pattern resolves into a sustained trend or proves to be short-lived noise. Each of these will help clarify whether the current divergence between bitcoin miners and the rest of the crypto-equity complex is a temporary function of this specific rally or the start of a more durable repricing of what a mining stock actually represents.