🚨📁ANALYST REPORT Analyst report by KBW came out yesterday and presented a bull case for miners getting into HPC and why the pipeline multiples may have evaporated for most. Spring and Summer signaled that leasing remained dominated by a few names. ⚡🏗️ $APLD (810 MW) ⚡🏗️ $HUT (352 MW Beacon Point lease + expansion) ⚡🏗️ $WULF (Anthropic at Hawesville) ⚡🏗️ $CLSK (178MW Sandersville + exclusivity for ~610 MW in Texas) ⚡🏗️ $RIOT added a small $AMD expansion ⚡🏗️ Crusoe large Meta deal. Newer participants such as $BTDR (Tydal, Norway phase-1 pending conditions) and $HIVE (Boden LOI) are only beginning to convert pipeline. 🤝 HPC colocation fundamentals remain constructive with a clear rebound in activity: 📈 Net lease signings jumped to 1.19 GW in Q2 928 MW has already been announced in Q3 Hyperscalers and AI labs continue to lock in capacity well ahead of delivery, supporting solid lease economics even as development costs rise. Leasing is still highly concentrated among a handful of proven developers: @APLDdigital , @Hut8Corp , and @TeraWulfInc alone accounted for 87% of the 2.21 GW signed YTD. Hyperscalers continue to favor established partners, though time to power is a key differentiator that can open doors for high quality energized sites from newer players like @CleanSpark_Inc. The bankable tenant pool is expanding but with added model layer risk: ✍🏻Frontier AI labs (with IG hardware provider credit support) are joining $AWS, $NVDA, and others in signing miner leases. This broadens demand, yet it exposes developers more directly to uncertain model outcomes and potentially partial (vs. full) credit backstops, raising the risk of future cap rate expansion relative to pure hyperscaler/IG enterprise leases. 📉The current selloff is from mostly de-rating pipelines rather than expanding cap rates: At a constant ~$17 MM/MW equity creation assumption (6.7% cap rate), current valuations imply only ~1.7 GW of future leases across the coverage universe down sharply from ~80% of visible interconnect-approved pipeline at the June peak to just 41% today. The market is now assigning little to no forward credit even to recent proven signers. 🏦Funding conditions have tightened selectively, with mixed implications: Average miner project bond yields have risen ~96 bps since late June (IG-tenant bonds +58 bps to ~6.51%; non-IG/CoreWeave-backed +172 bps to 8.82%). Higher capital costs favor proven developers and simultaneously raise the cost of hyperscaler self-builds, which supports third-party leasing demand. 💵Hyperscalers themselves are turning more to equity: ($GOOG) @Google's ~$85 B raise + ATM; ($META) @Meta's reportedly considering equity 📃Lease economics have held up well despite higher costs and longer terms: Untrended first year yield on cost has moderated to ~13.2% on recent (ex- $APLD) deals mainly because of higher estimated capex (~$11 MM/MW) and a shift toward 20-year terms. Trended NOI/MW, however, continues to set new highs (~$1.95 MM), and absolute rents remain resilient as IG tenants are still willing to pre-commit. Long term compute demand is intact, but model layer uncertainty and cloud pivots warrant caution: Hyperscaler uncommenced lease obligations have surged (reaching $561 B+ by Q1 across the big four). Miners should still prefer pure IG/hyperscaler leases over lab backed ones. 🔌Multi-year reserved GPUaaS contracts are essentially colocation “in cloud costume” (with miners taking GPU risk), while true AIaaS differentiation remains difficult against hyperscalers and frontier labs. The GPUaaS market itself is becoming crowded: Meta, @SoftBank , xAI (supporting $SPCX).