
IOSG|Bitcoin Must Hit $250,000 for Mining to Catch Up with AI Rent
Author: Mario Chow, IOSG Ventures

Stock prices reflect the closing price on November 13, 2026; Bitcoin was priced at $64,400 on November 16. MW in this article refers to critical IT load, meaning the actual power delivered to servers. Gross MW refers to total park electricity consumption, including cooling, and is typically 20% to 40% higher than IT load.
TL;DR
Why the pivot.
CoinShares values completed, leased AI data centers at approximately $27 million per MW, while valuing powered mining capacity without tenants at under $3 million per MW.
This 9x difference isn't entirely due to grid interconnection fees, as the former includes construction costs of $8 million to $15 million per MW. Even after deducting construction costs, the gap remains 4x to 6x.
Based on operating profit per MW, mining nets about $230,000 annually at the current hashprice, while an AI lease yields around $1.9 million. Bitcoin would need to rise to approximately $250,000 to close this gap. Behind this lies an AI power shortage: connecting new grid infrastructure in the most sought-after US markets takes three to four years, whereas miners already hold powered interconnections.
The midterm elections on November 3 further scarce these interconnections, as Texas elected a governor who pledged to halt data center permitting.
How to pivot.
Grid interconnections, substations, land, and permits can be retained, while almost everything else must be demolished and rebuilt at a cost of $8 million to $15 million per MW, comparable to greenfield development. Miners aren't selling buildings; they're selling the two-to-four-year queue time that tenants can bypass.
Who is pivoting.
Eight publicly listed mining companies have signed agreements totaling approximately 6.3 GW of IT capacity. Early on the day following the election, Riot leased out its entire 910 MW Corsicana campus; additionally, Cipher holds 700 MW of gross power. Only about 1 GW is truly operational. CoreWeave alone accounts for 24%, with leases backed by no guarantor.
Riot's AI lab tenant (reported to be Anthropic) now represents 18%, nearly matching CoreWeave's share.
Revenue and Costs.
After deducting operating expenses, disclosed NOI leases range from $1.85 million to $2.15 million per MW annually. Galaxy Helios Phase II's NOI is derived from its disclosed margins, approximately $2.3 million, serving as an outlier. The annual cost return on construction sits between 15% and 20%.
Riot's Corsicana lease breaks this range at roughly 33%, since the tenant constructs the data hall itself on Riot's already-powered land. Differences in nominal rent primarily stem from lease structures rather than compensation for credit risk.
Implications.
The mining holdouts have sold approximately 40% of their Bitcoin reserves, with remaining sell pressure concentrated on MARA. Riot sold 20,264 BTC this year, leaving only 2,418; MARA offloaded 13,597 BTC after the quarter-end to repay its Bitcoin-backed loans.
MARA, CleanSpark, and Riot collectively hold about 36,000 BTC today, down from 60,500 in August. MARA's ~20,400 BTC are the most likely to continue flooding the market. Hashrate dipped and recovered; miner exits haven't weakened the network. Mining equities are no longer mere proxies for Bitcoin.
Post-election performance diverged by state: Riot surged 25% while Bitcoin fell 12%; several other companies with pending deliveries in Texas and Georgia dropped 15% to 18%.
IOSG Perspective
CoreWeave exposure represents a rent-uncompensated risk; Texas constitutes a second risk factor.
Galaxy charges the highest rent in the industry to CoreWeave, yet posts the lowest construction ROI of roughly 15% annually. CleanSpark earns ~16% leasing to Meta, and Hut 8 secures 17% to 21% from investment-grade tenants. Logically, weaker tenant credit should yield higher landlord returns, but reality dictates the opposite.
CoreWeave accounts for 100% of Galaxy's AI business, ~53% of Core Scientific's, and 28% of Applied Digital's. Hut 8, CleanSpark, Riot, and TeraWulf hold zero exposure to it.
CoreWeave's Nov 10 earnings beat expectations, sending shares up 25% the next day. However, recourse debt ballooned by ~$10 billion during the quarter to $41.6 billion, pushing the stock back to $53 by Nov 13.
The election added a second stress test. Companies with approved, leased, and funded MW saw gains, while those awaiting Texas or Georgia approvals saw declines. Galaxy failed both tests: its business relies entirely on CoreWeave, its Helios campus is in Texas, and Helios II remains unleased.
Until CoreWeave's credit stabilizes and Texas's moratorium gets a clear end date, high-exposure landlords should trade at steeper discounts than low-exposure ones. Core Scientific represents the cleanest CoreWeave exposure sample; its stock stabilized post-election because 485 of its 590 MW to CoreWeave is already generating rent.
Mapping these two stress tests against each other segments landlords into three groups.
Group 1: Leased, outside Texas and Georgia, funding secured: TeraWulf (New York and Kentucky, with Google backing project financing for its Anthropic lease), Hut 8's River Bend (Louisiana, non-recourse bonds), and Applied Digital (North Dakota, collateralized notes).
Group 2: Leased, but in Texas or Georgia, with subsequent phases contingent on state permits or unfinalized loans: Riot's Rockdale and Corsicana, Hut 8's Beacon Point, Galaxy's Helios Phase I, and CleanSpark's Sandersville.
Group 3: Unleased or unapproved: Galaxy's Helios Phase II, MARA's Matagorda campus, and Cipher's Texas pipeline assets. Discounts should be smallest for Group 1, largest for Group 3. Galaxy straddles Groups 2 and 3, compounded by its CoreWeave exposure.
As expected, the primary sellers are the mining holdouts, with MARA constituting the bulk of the remainder.
In August, transformed landlords held virtually zero Bitcoin, while MARA, CleanSpark, and Riot collectively sat on ~60,500 BTC needed to fund AI projects. Two of these three later sold. Riot offloaded 20,264 BTC in the first nine months (~$1.5 billion), primarily for data center construction, leaving just 2,418 by late September.
After quarter-end, MARA sold 13,597 BTC at an average of ~$81,000 to pay off $750 million in Bitcoin-backed loans, signaling continued reliance on BTC sales to fund AI operations. The trio now holds ~36,000 BTC (~$2.3 billion), representing roughly a 40% reduction, leaving 60%. Whether they will liquidate further hinges on remaining capital needs.
Riot's remaining 2,418 BTC are immaterial; Corsicana funding is covered by cash and project loans. CleanSpark's 13,530 BTC remain untouched, with Sandersville financed via collateralized notes, making BTC sales least necessary for them; however, they suspended monthly production reports starting in October.
MARA's ~20,400 BTC (~$1.3 billion) constitute the key sell pressure to watch: management stated it will continue selling. If its Texas campus receives power approval, it faces up to $600 million in costs. Additionally, active lease negotiations may eventually require self-construction. Inventory runoff has peaked but hasn't concluded.
Mining equities now function as AI credit trades, with the election adding a layer of Texas permitting speculation.
Transformed landlords move with CoreWeave and Nebius, decoupled from Bitcoin: during Bitcoin's October rally, they fell alongside CoreWeave's bonds.
From Nov 3 to 13, Bitcoin dropped 12%, while Riot surged 25% on a 910 MW Texas lease with approved power. Galaxy, Cipher, and CleanSpark, carrying pending deliveries in Texas and Georgia, fell 15% to 18%.
Buying this basket bets on AI data center financing and state-level permits, with Bitcoin playing only a minor role.
Why Miners Are Pivoting
Higher Prices, Weaker Mining Economics
Hashprice denotes the daily revenue expected per PH/s of hashrate before costs; it determines miner survivability, yet the metric has been steadily declining.

Pre-2024 halving, 1 PH/s earned ~$105 daily. Today it's ~$30, despite Bitcoin prices returning to similar levels. Halving cut block subsidies in half, while global hashrate rose ~70%, squeezing each machine's share twice over.
This nears summer lows: hashprice briefly dipped below $30. CoinShares estimated public miners' average cash cost at $75,500/BTC versus ~$60,000 pricing, forcing the industry to mine at a loss.
Bitcoin rallied to $84,000 in October before falling back to $64,400 this month. Neither move altered fundamentals, given the stark profitability gap versus alternative paths.
Same 1 MW, Two Different Math Models
Chart below uses late-October hashprice of ~$35; at today's $30, mining figures would be even lower.

On an operating profit per MW basis, 1 MW of next-gen ASICs nets ~$230,000 annually at current hashprice after power costs, with next-gen hardware requiring separate capital expenditure.
Leasing that same 1 MW to an AI tenant generates $1.85M–$2.15M in NOI over 15–20 year contracts with annual escalators. Even if Bitcoin hits $150,000 with static difficulty, mining yields only ~$1M annually with zero contractual guarantee.
This comparison favors leases: the $1.9M NOI requires $8M–$15M/MW in CapEx, versus ~$800K/MW for mining rigs. CoinShares' valuation metrics cannot directly dictate conclusions.
It values completed, fully leased AI data centers at ~$27M/MW versus under $3M/MW for powered but unleased mining capacity—comparing two distinct asset classes, not pre- and post-leasing valuation of the same campus.
Deducting construction costs from leased valuations leaves $12M–$19M/MW, still 4x to 6x the unleased baseline. This excludes completion risk and pre-lease runway. The directional trend is clear; precise multiples are approximate.
The most compelling evidence is exit willingness: Core Scientific spent ~$42M to cancel a new 15 EH/s rig order. Keel (formerly Bitfarms) halted mining in June. Galaxy restructured Helios facilities and recorded impairment losses.
Converting Farms to Data Centers
Keep the Outlets, Rebuild Everything Else
Post-retrofit, only infrastructure outside the meter face (grid-side) survives.

Exterior elements transfer over; interior setups become obsolete. ASICs hold zero value for AI, and warehouses typically get demolished. Miners voluntarily shed load when power tariffs spike, historically skipping backup generation and redundant power distribution, failing to meet AI tenant uptime commitments.
A single NVIDIA GB300 rack draws 130–140 kW versus ~9 kW for standard server racks. Heat dissipation mandates direct-to-chip liquid cooling lines.
Hence retrofit costs hit $8M–$15M/MW: TeraWulf guides $8M–$10M, Hut 8 targets $9M–$11M, CleanSpark estimates ~$11.9M, Galaxy's Helios Phase II runs ~$15.4M—effectively greenfield pricing.
The saved asset is never capital; it's time.
Three Ways to Sell MW
Landlords can lease shells, sell compute, or divest campuses outright.

Industry standard is the "powered shell": land, power, structure, and cooling are landlord-provided; GPUs, networking, and ops staff come with tenants. This division allows post-pivot miners to command real estate multiples.
Riot's Corsicana lease pushes further into "powered land": Riot delivers only land, power, and slab platforms; the tenant designs, builds, and operates the data hall.
IREN pursues compute leasing, exemplified by a 5-year, $9.7B Microsoft deal requiring $5.8B in Dell hardware purchases. Per-MW revenue approximates five times traditional leases, but chip depreciation wipes most value within five years, and contract duration covers only one-third of typical lease tenors.
Benchmarking IREN against pure-play landlords using per-MW revenue is the sector's most common analytical error.
Third path: campus divestiture. TeraWulf sold its Abernathy JV equity to Fluidstack for ~$530M, with $250M settling in Q3.
CoreWeave previously attempted a full acquisition of Core Scientific in 2025, rejected by shareholders.
Funding Sources
A 200 MW campus demands ~$2B at $10M/MW, with construction spend lagging lease commencement by roughly two years.

Tightest timing risk remains Riot. Its $573M Morgan Stanley bridge loan matures Dec 31, 2026, yet Rockdale's initial 96 MW won't lease until Dec 2027, with investment-grade guarantees for this $9.1B agreement still being negotiated.
Riot held $950M in cash and BTC at late September, ruling out equity dilution while arranging project finance for Rockdale and Corsicana. These funds must clear before commencement.
Current Landscape
Landlords: Signed Doesn't Equal Built
The Corsicana deal ranks Riot as the second-largest landlord by contracted capacity, though virtually nothing is operational yet.

Contracted MW includes future deliveries; billable MW captures active revenue streams. Galaxy leases 526 MW to CoreWeave, with 133 MW actively billing. The remaining 393 MW aren't idle vacancies—they're pre-leased but under construction.
Core Scientific reported 485 MW billable at Sept 30 versus 395 MW at June 30. Dalton's remaining 100 MW expects commissioning in early 2027.
Interconnection solves one bottleneck. Subsequent hurdles include financing, construction, cooling, redundancy, and tenant acceptance. Low current billable capacity versus large远期 backlogs simply reflects pipeline maturity, not weak demand.
Cipher's 700 MW represents gross campus power, not IT load, excluding it from capacity allocation stats. However, filings show Amazon's Black Pearl lease utilizes 216 MW of IT load, translating to ~$1.7M/MW/year in rent.
Black Pearl generated $5.9M in lease revenue in Q3, marking Cipher's inaugural collection.
The sector's largest single lease emerged the morning after the election. Nov 4 saw Riot lease its entire 910 MW Corsicana campus for 15 years and ~$28B to the same AI lab behind Rockdale.
Prior to that, zero new deals closed since Sept 18. Cipher extended its Barber Lake lease to 20 years, boosting contract value by $5.2B, yet added zero MW.
Tenants
With Corsicana, direct AI lab leasing share more than doubled.

Neocloud providers remain the dominant tenant cohort, comprising ~37% of the 6.3 GW. CoreWeave leases 1,516 MW across Applied Digital, Core Scientific, and Galaxy (24% aggregate), with Fluidstack capturing another 10%.
Major cloud and tech firms account for ~30%, AMD-linked arrangements ~9%, direct AI labs 24% (up from 11% pre-Corsicana).
If Riot's tenant is indeed Anthropic, direct exposure totals ~1.5 GW. Adding River Bend's 245 MW through Fluidstack pushes terminal demand beyond CoreWeave.

Tenant concentration is highly skewed across landlords. Galaxy, CleanSpark, and Bitdeer each hold single-tenant profiles. Riot's capacity is now 96% anchored to one lab. CoreWeave concentration warrants closer scrutiny, as credit dynamics flipped since summer when unaudited AI labs sparked caution.
Since then, Anthropic raised $65B at a $965B valuation. Reported Q3 revenues hit $16.8B. Filed IPO docs Oct 26, currently roadshowing with rumored valuations near $1T.
Market anxiety centers squarely on CoreWeave: B+ rating, $41.6B in recourse debt ($10B incremental quarterly). Mid-Oct bond yields neared 15%, 5Y CDS exceeded 1,000bps, persisting pre-Nov 10 earnings.
Within this leasebook, the heaviest risks stem not from labs themselves, but from intermediaries reselling capacity.
Holding Out: Remaining Miners
Large-scale mining persists only among select names: Bitdeer (82 EH/s), MARA (~71), CleanSpark (50), Hut 8's American Bitcoin (~25). All maintain dual strategies.
MARA's 2 GW Matagorda Texas campus qualified as "studied load" in ERCOT's Batch Zero queue. MARA targets two year-end leases, acknowledging potential carryover into 2027. CleanSpark's rigs sit adjacent to Georgia's Meta lease.
Riot formally maintains mining roots: hash revenue dominates top line, yet 1,151 MW is leased and BTC inventory nearly depleted.
Revenue and Cost Structures
Lease Yield per 1 MW
Nominal rent equals contract value divided by MW count and term. NOI reflects landlord retention post-property OpEx deduction, excluding financing, depreciation, and taxes. Under triple-net leases, tenants bear OpEx, aligning both figures. Service-inclusive models command premium rents to absorb costs, sacrificing 10%–20% of gross inflows.
On IT-load bases, nominal rents span $1.44M/MW/year (CoreWeave @ Core Scientific) to $2.6M/MW/year (CoreWeave @ Galaxy Helios II). Identical tenant bookending the range proves nominal pricing ignores credit differentials.
At Core Scientific, CoreWeave funds majority fit-outs. Economic commitment runs ~$1.5M/MW. Lower rentals correspondingly offset reduced build-outs. Four sampled leases below disclose concurrent rent and NOI figures, illuminating mechanics.

Riot's Rockdale AI lab lease (Anthropic-reported) commands nominal rents 28% above Hut 8's high-rated triple-net counterpart. Post-OpEx, midpoint disclosures reveal only 9% upside, bottom-tier calculations just 4%.
Premiums predominantly offset landlord-incurred costs, leaving negligible risk premiums for credit weakness.
Riot's Corsicana deal marks a product iteration. 910 MW averages ~$1.8B/year, translating to ~$2M/MW/year (inclusive of escalators), yielding 99% NOI margins within Rockdale parameters. The variable shifted to CapEx.
Riot commits ~$6M/MW versus $8M–$15M full-build costs, as tenants construct proprietary halls. This represents the first lease priced on "approved Texas power" rather than construction milestones. Closing timed precisely post-election, a catalyst that tightened Texas supply availability.
Timing supersedes credit ratings. Amazon closed with Cipher late 2025 at ~$1.7M/MW. Meta partnered with CleanSpark nine months later at $1.89M, both triple-net AAA credits.
Direct AI lab leases executed mid-2026 (TeraWulf, Riot Rockdale, Bitdeer) priced at $2.37M–$2.43M.
November Corsicana closure at ~$2M breaks trends; base-period accounting would depress this further, as $2M blends escalators absent from comparables. Reduced build coverage drives variance. These represent structural nominalities; we treat patterns as heuristics, not definitive proofs.
Cost Returns
Build costs track rental inflation synchronously, compressing construction ROI into narrow bands. Metrics calculate annual NOI divided by CapEx, excluding financing structures, thus diverging from equity IRR.

15%–20% annual returns on 15-year anchored properties constitute premium real estate economics. Miners capture this spread by monopolizing scarce inputs. Corsicana quantifies element monetization when tenant-funded construction eliminates CapEx drag. Riot's self-reported 33% assumes successful project financing clearance, currently in negotiation.
Galaxy anchors the range's lower bound despite weakest tenant credit: commanding peak rents and bearing maximum build costs, resulting thinnest margins.
Actual Revenue Generating Assets
Latest disclosures confirm Core Scientific billed 485 MW at Sept 30. Galaxy generated revenue on 133 MW flatlining through Q3, with Phase II targeting H2 2027 launch. TeraWulf commissioned 102 MW by Sept 30, plus 336 MW under construction.
Applied Digital achieved 250 MW ready-for-use status Oct 1, guiding to 300 MW year-end. Cipher collected inaugural Black Pearl rent in Q3 without MW breakdown; Amazon's six delivery tranches run Sept 30 through Feb 28, 2027.
Riot's 25 MW AMD allocation commenced billing in Q2.
Remainder constitutes delivery pipelines.
Bitdeer stages 60.5 MW batches in Q4 2026 and Q1 2027. Hut 8 River Bend launches H2 2027. Riot Rockdale targets Dec 2027. Corsicana phases deliver Sept 2027–July 2028. CleanSpark initiates billing Q4 2027.
Divergent dating methodologies and milestone definitions preclude aggregation into unified industry percentages.
Implications
Bitcoin Impact: Holdout Sellers
Addressing primary bearish narratives: Blockware data shows public miner inventories contracting from ~127k early-year to ~99k mid-August, representing ~$1.8B net proceeds allocated to data center CapEx.
MARA tops holdings lists and loudly championed accumulation policies. March policy shifts enabled reserve liquidation, slashing positions from 53,822 to 35,577 over six months. Blockware correctly attributes this sell pressure to underestimated drag on 2026 BTC performance. Momentum persists.

Transformed landlords exhausted sell-side options long ago: Core Scientific and Cipher each hold <1,000 BTC (~1% market value impact), TeraWulf, IREN, and Applied Digital approach zero. Consequently, liquidity pressure redirected toward remaining miners initiating self-built AI ventures.
Riot liquidated 20,264 BTC YTD, channeling proceeds into facility construction, retaining 2,418 by September. Post-quarter MARA offloaded 13,597 at ~$81k均价, deploying $750M to extinguish collateralized loans, releasing pledged balances. Hut 8 sold ~3,100 for share buybacks.
CleanSpark's 13,530 BTC stand untouched among holdouts, coinciding with suspended October reporting cycles. Trio aggregates ~36,000 (~$2.3B current value), down from 60,500 in August. This reflects a 40% reduction, not total liquidation.
MARA retains ~20,400 BTC pledging ongoing liquidations for AI funding. CleanSpark finances builds via notes. Riot's residual 2,418 carries negligible impact.
Inventories represent isolated supply vectors. Network issues ~450 daily emissions. Miner retention decisions pivot aggressively toward sale amid transformation. MARA updated 2025 mining payout policies, expanded reserves liquidation eligibility in 2026. Riot outpaced internal issuance with external sales volumes.
CryptoQuant logs show Oct 26 miners bridging 21,140 BTC to Binance—sixth >20k daily sweep since February. Reserves approach five-year troughs. Flow metrics encompass private operators and pools. Exchange deposits signal readiness to transact, not realized sales velocity.
Thereafter, BTC retraced from late-October highs near $84k to $64.4k, marking August lows alongside spot ETF outflows, hawkish Fed signals, and NVIDIA executive transition-driven tech selloffs. Miner redistribution injected friction into the decline but didn't initiate the catalyst sequence.

Network absorbed exits seamlessly. Hashrate declined 19% from last Nov peaks to July troughs, rebounding past 1,000 EH/s in October, estimating 965 EH/s normalized to mid-Nov difficulty. Dynamic adjustments compensated: post-shutdown survivor economies captured redistributed block rewards, while external capacity filled gaps.
Long-duration concerns target security budget compression. Fee contributions plateaued at 0.63% of miner revenue—fifteenth consecutive month below 1%. Ecosystems rely heavily on subsidy distributions facing binary reductions in 2028. Adam Back theorizes mining-AI arbitrage convergence points; rising hashrate validates equilibrium mechanisms.
2028 halving constitutes ultimate stress test. Current pricing models project next-gen ASICs unable to cover 0.04/kWh utility rates post-subsidy cuts. Structural energy diversion toward AI accelerates. Historical absorption patterns confirm network resilience through sequential capacity substitution.
Crypto Market Dynamics: Equity Decoupling
Since Aug 3, BTC rallies persistently while every mining counterparty depreciates.

Period Aug 3–Nov 13 witnesses 8% BTC appreciation versus universal mining drawdowns ranging Riot (-18%) to Cipher (-72%). Parallel erosion across active miners and pivoted entities disproves punitive market mechanics. Five drivers explain divergence.
First, timing misalignment: equity cycles peaked June alongside AI infrastructure euphoria, contrasting August bottom formations near $63k. Second, universal funding requirements anchor pricing regardless of operational stance. Third, October CoreWeave solvency shocks contaminated correlated credit proxies. Fourth, marginal upside benefit from spot strength limited by suppressed hashpricing ceilings. Fifth, election outcomes penalized jurisdictional uncertainty exposures.
Neutral colocation benchmarks (Digital Realty, Equinix) bracketing −6% to +2% isolate idiosyncratic mining flaws: execution risk, permitting delays, delivery slippage, and tenant credit deterioration.

60-day rolling correlations peg transformed landlords firmly within 0.6–0.8 proximity to CoreWeave/Nebius trajectories, severing historical Bitcoin linkages that collapsed from spring 0.55 averages to September mid-month 0.15 thresholds.
CryptoSlate derivatives tracking Hut 8 beta contraction 1.18→0.53 and Cipher 1.13→0.58 confirms fundamental re-rating away from crypto sensitivity.
Electoral fragmentation recalibrated subsector rotation. Nov 3–13 periods showcased inverse movements: BTC drops 12% juxtaposed Riot +25%, Core Scientific +2%, TeraWulf +1% versus Galaxy -15%, Cipher -16%, CleanSpark -18%.
Bernstein slashes seven-name target averages 35%→54%, applying forward-year EBITDA 12x multiples to AI custody franchises citing tightened credit windows and Texas regulatory ambiguity.
Morgan Stanley favours TeraWulf, Hut 8, Applied Digital based on geographic diversification, overlooking Hut 8's Beacon Point Texas footprint inclusion.
Contagion initiated among debt-fueled compute lessors expands ecosystem-wide. Aug 3 baselines record CoreWeave -38%, Nebius -46%, NVIDIA -21% cascading post-executive succession announcements.
Oct 14 institutional dumping pressured CoreWeave senior notes coincident with Cipher convertible pricing events. Single-day cipher declines hit 12%, sector averages compress 5%–10%.
Prominent mining vehicles like WGMI pivot top allocations toward Cipher, Nebius, CoreWeave, Hut 8. Dual-zero-hash portfolios redefine public mining baskets from proxy leverage instruments to embedded AI infrastructure credit exposures stripped of underlying native asset correlation.
AI Sector Implications: Real Capacity Deployment
JLL pipelines catalogue 66 GW North American construction volume. BNEF grid interconnection capacity constraints cap annual additions at ~7.1 GW. Pre-2029 energization projections allocate substantial portions to incumbent miner leases leveraging pre-installed interconnection rights.
Midterm electoral shifts amplify scarcity premiums. New executives across Texas, Georgia, Ohio campaigning on construction moratoria elevate stranded asset valuations. Constraints detailed subsequently.
Risk Factors
Texas Jurisdiction.
Concentrated footprints characterize Riot, Galaxy, Cipher AI campuses alongside Hut 8 Beacon Point and MARA developments.
Standalone exposure registers 2.4 GW (38% aggregate contracted volume), excluding Core Scientific Texas divisions and Cipher allocations. Republican gubernatorial administrations implemented preliminary suspension protocols.
Aug 3 directives ordered heavy-load queue auditing triggering ERCOT Batch Zero research halts alongside fresh computational authorization suspensions. Sept 21 administrative freezes encompassed statewide environmental compliance approvals. Consensus anticipated post-victory easing mechanisms.
Electorate installed Democratic successor Gina Hinojosa Nov 3. Universal permitting moratorium pledges extend indeterminate horizon pending legislative reconstruction. Inauguration scheduled Jan 19 follows legislative convening schedules.
Executive authority interfaces constrain unilateral mandate enforcement. Public Utility Commission appointments retain Abbott incumbency; gubernatorial nominees require two-thirds upper chamber ratification maintaining 19–12 partisan majorities. Baseline projection anticipates extension through mid-2027 culminating in enhanced conditional frameworks rather than absolute prohibitions.
Near-term catalysts identify ERCOT audit dissemination Dec 10 alongside regulatory committee sessions Dec 17. Secured power-backed contracted volumes demonstrate temporary stability anchoring pre-approval transactions like Riot Corsicana closures.
Vulnerability vectors target unbuilt dependencies: Galaxy Helios trailing 830 MW unabsorbed square footage. MARA Matagorda fiscal triggers tethered exclusively to grid authorization status. Operational standby generation permit requirements complicate executing agreements.
Georgia jurisdictions hosting CleanSpark Meta assignments anticipate statewide transitional suspension directives mirroring gubernatorial platforms. Regulatory composition transitions democratic majorities projected for 2027.
Workload Allocation Shifts.
Geographic positioning inherently favors training architectures driven by remote utility procurement advantages. Gartner forecasting projects inference spending eclipsing training allocations this cycle ($23.3B vs $19B), escalating trajectory approaching 59% deployment ratios by 2027.
OpenAI restructuring initiatives cancelled Abilene expansion programs March 800 MW allocations April suspends Stargate UK deployments. Microsoft absorbs Norwegian capacity equivalents demonstrating demand migration rather than eradication. Counterarguments emphasize batch inference, agent frameworks, reinforcement learning paradigms exhibiting latency insensitivity blurring workload categorizations.
Multi-decade contractual alignments structurally bet upon persistent workload symbiosis.
Intermediary Solvency Risks.
Subleasing architecture channels ~37% throughput utilizing specialized cloud aggregators redistributing utilization matrices. CoreWeave concentration dominates 24% distribution lacking underlying guarantees accumulating $41.6B recourse liabilities trading distressed conventions October. Executives signal sovereign forbearance suppressing unsecured issuance trajectories.
Fluidstack aggregates 10% private entity status exempted rating protocols. Google obligations selectively support TeraWulf Cipher deliverables excluding comprehensive coverage mandates. Bitdeer Volta dependency preserves conditional banking instrument protections.
Aggregate intermediary collapse scenarios leave landlords possessing unfinished structures singularly occupied facilities lacking direct contractual linkages terminating laboratory entities.
Capital Availability Constraints.
Majority leasebooks lack corresponding monetization events relying either discounted equity sequences experiencing >33% August depreciations or liability structures benchmarked against CoreWeave spread architectures. VanEck diagnostics estimate residual funding shortfalls nearing $50B thresholds. Bernstein publications corroborate tightening capital formation environments.
Liquidity reservoirs contract significantly Hut 8 unrestricted cash balances plummeting $157M quarterly decreases Bitdeer September endpoints securing $214M Tydal residual capital deployment expenditures spanning hundreds millions. Bridge financing maturities December 31 confront imminent settlement obligations.
Conclusion
Fundamental realignment materializes irreversible strategic completions. Incumbent grid interconnections represent paramount AI construction bottlenecks. Post-completion commercialization elevates campus valuations to ~$27M/MW versus sub-$3M空置 baselines; subtracting structural expenditures preserves 4x to 6x contractual premiums. Achieving $250k price parity theoretically offsets operational math discrepancies necessitating dual-strategy positioning even amongst conservative participants.
Electorally driven asset revaluation bifurcates underlying infrastructure valuations. Prohibitively expensive slower permitting processes augment existing cleared leased funded MW multiplicities whilst pending Texas Georgia adjudications compress pricing metrics.
Closing mechanics executed election-morning Corsicana agreements establish transparent pricing benchmarks calculating ~33% annualized cost recovery efficiencies when contractor-managed construction parameters apply.
Execution phases remain incomplete. 6.3 GW contracted volumes contrast functional 1 GW activations requiring capital infusions throughout near-term fiscal horizons. Standardized 15%–20% margin configurations validate enterprise viability; differential pricing structures drive credit tier separations rather than risk premiums mitigating CoreWeave concentration vulnerabilities disproportionately affecting exposed landlords.
Asset realization curves exhibit deflationary momentum exhaustion uncomplete distribution sequences. Transition cohorts liquidate precedentially followed operational holdout disbursements aggregating ~36,000 circulating supplies MARA components projecting sustained market injections. Restored hashing capabilities demonstrate systemic absorption resilience validating sequential participant exodus accommodative network behaviors.
Digital portfolio allocations transform equity classifications abandoning foundational cryptocurrency proxy characteristics replacing derivative dependencies with AI infrastructure default probabilities combined regional regulatory uncertainties embedding minimal native asset sensitivities.
Methodology and Disclosures
Throughput aggregations benchmarked December 2026 disclosure cutoffs exclude IREN GPU cloud licensing arrangements. Contractual allocations incorporate prospective fulfillment timelines. Ready-for-service, delivered, and billable statuses report independently per corporate nomenclature definitions varying timestamp parameters precluding unified industry occupancy rate calculations.
Distribution matrices quantify contracted IT megawatt proportions disregarding revenue attribution or credit grading standards. Anonymous hyperscale provider identifications assume distinct corporate identities transiting AI lab utilization requirements remain categorized under cloud vendor aggregators.
Dual laboratory lease executions omit explicit counterparty designations Bloomberg intelligence sources attribute Rockdale occupancies Anthropic identities Corsicana alignment confirms identical operator classification.
Published 700 MW specifications represent gross electrical throughput rather than critical computing loads excluding inclusion threshold determinations.
February investor prospectus documentation outlines Black Pearl configurations registering 300 MW gross capacity 216 MW critical processing density translating fifteen-year $5.5B agreements into ~$1.7M per IT MW annualized metrics ($1.22M gross normalization factors).
Commercialization activation commences August Black Pearl operations withholding billable throughput quantification reports.
Base-period contract valuations divide critical IT capacity multiplied tenure periods excluding escalator provisions Corsicana anomalies reflect averaged inclusive uplift reporting methodologies. Net operating income extracts exclusively audited organizational submissions sourced Hut 8 CleanSpark Riot filings.
Phase II operational margins extrapolate disclosed ninety percent-plus profitability coefficients applied aggregate receipts. Construction expense recoveries utilize directed capital expenditure roadmaps Corsicana benchmarks incorporate Riot six million dollar per unit calculations offsetting four million tenant reimbursement allowances.
Regional allocation computations encompass Riot Rockdale Corsicana assets Hut 8 Beacon Point installations Galaxy Helios complexes excluding Core Scientific Texas divisions fragmented disclosure limitations Cipher operational footprints.
Hashing economics parameters establish nineteen joules terahash efficiency standards fifty-two point six petahash per megawatt equivalencies utility pricing four cents kilowatt-hour continuous deployment constants difficulty stagnation assumptions restricting arithmetic operations net revenue minus expenditure deductions.
Computational valuation models integrate daily hundred forty-four block issuance routines adjusted total network processing outputs projected earning capacities contemporary snapshots deploy mid-November trillion fourteen point eight threshold difficulties sixty-four thousand four hundred dollar quotation benchmarks. Correlation matrices apply United States securities trading session logarithmic return series. Reserve inventories synchronize latest published organizational statements mapped chart timeline parameters standardized valuation constants.
Lease Schedule Detail
Investment grade designation indicators non-disclosure acknowledgments. Base period contractual sums define baseline valuations. Reporting omissions denote absence extracted net operating income figures within reviewed documentation repositories.












