How CoreWeave, Nebius, Lambda, Crusoe, Applied Digital, Together AI, Voltage Park, Fluidstack, and AI-native GPU-cloud, AI-factory.
AI neocloud, AI factory, GPU-backed project finance is the most capital-intensive category in tech — CoreWeave $CRWV IPO Mar-2025 ($23B open, $86B backlog with Microsoft, OpenAI, Meta, IBM), Nebius $NBIS relisted Oct-2024 (Yandex-NV spin-out, $1.5B PIPE Nvidia-Accel-Orbis Dec-2024), Lambda ($480M D Feb-2024 USIT-Andra Capital, $500M GPU-financed debt Macquarie Apr-2024, IPO-track), Crusoe ($600M D Dec-2024 Founders Fund-Fidelity-Long Journey, $3.4B JV Blue Owl-Primary Digital Infrastructure for Abilene TX Stargate campus 2025), Applied Digital $APLD ($5B Macquarie perpetual-preferred Sep-2024, Ellendale ND HPC campus), Together AI ($305M B Feb-2024 Salesforce Ventures-Coatue-Nvidia), Voltage Park (Jed McCaleb-funded 24k H100 cluster launched 2024), Fluidstack (Google-anchored, Mistral partner), Vultr ($3.5B valuation Dec-2024 LuminArx-AMD), Cirrascale, GMI Cloud, DataCrunch, Massed Compute, Runpod, Paperspace-DigitalOcean ($111M acquisition Jul-2023), Genesis Cloud, Latitude.sh, TensorWave (AMD MI300X-anchored), reset the compute-supply-chain so that every frontier-lab, enterprise, sovereign-AI-program bypasses hyperscaler-reserved-instance queues and procures GPU-cloud, AI-factory capacity direct from neocloud, colo, power-first operators. Capital-stack, infrastructure-fund, hyperscaler-anchor — Blackstone $BX (QTS acquisition $10B 2021, AirTrunk $16B Sep-2024, Blackstone Data Center Fund, $70B+ AI-infra pipeline announced 2024), BlackRock-Microsoft-MGX-Global Infrastructure Partners AI Infrastructure Partnership $30B Sep-2024 (expandable to $100B), Apollo $APO ($25B AI-infra allocation 2024, Wintergreen Renewable Energy JV), KKR $KKR ($50B+ AI-infra allocation 2024, CyrusOne $15B), Brookfield Infrastructure $BIP, Brookfield Renewable $BEP ($20B+ AI-infra, power PPAs), Stonepeak (CoreSite, AIP, $15B AI-infra fund 2024), DigitalBridge $DBRG (DataBank, Vantage, Vertical Bridge, Zayo, $80B AUM AI-infra), Blue Owl $OWL (Primary Digital Infrastructure-Crusoe $3.4B JV Jul-2024, IPI Partners $10.5B acquisition 2024), IPI Partners, GDS Holdings $GDS, VNET $VNET, KIO Networks, GLP, Ares $ARES, Antin Infrastructure, I Squared, EQT Infrastructure, Macquarie Asset Management ($5B Applied Digital, $500M Lambda GPU-financed debt), Fifth Wall, Generation Investment Management, TPG Rise, Warburg Pincus, General Atlantic, Silver Lake, Bain Capital, plus GPU-backed ABS, private-credit — Blue Owl-Morgan Stanley-Voyager Space GPU-ABS $1.8B Aug-2024 (first-ever GPU-collateralized asset-backed security for CoreWeave), Lambda-Macquarie $500M GPU-financed debt Apr-2024, CoreWeave-Blackstone-Magnetar-Coatue $7.5B debt May-2024, $650M revolver Nov-2024, $650M IPO, $2.3B extension Nov-2024, Crusoe-Blue Owl-Primary Digital Infrastructure $3.4B JV Jul-2024 for Stargate Abilene, Applied Digital-Macquarie $5B perpetual-preferred Sep-2024, plus hyperscaler take-or-pay — OpenAI-Stargate $500B commit Jan-2025 (SoftBank-OpenAI-Oracle-MGX, Crusoe-Oracle Abilene TX first campus), Microsoft-CoreWeave $10B+ take-or-pay 2023-2030, Microsoft-G42 UAE partnership Apr-2024, Meta-CoreWeave $6.5B extension Nov-2024, Google-Fluidstack partnership, Google-Anthropic $2B, $10B compute-commit Nov-2024, Amazon-Anthropic $8B additional Nov-2024, Nvidia-CoreWeave $1.3B lease-back 2023, Nvidia-Lambda, Nvidia-Nebius, Nvidia-VAST, Nvidia-DDN partner-network. Power, siting, interconnection — ERCOT Texas, PJM Interconnection, MISO, SPP, WECC, SERC, NERC queue-backlogs (2027-2032 typical, 500 GW nationwide), Talen-Amazon $650M Cumulus Susquehanna nuclear-PPA Mar-2024, Constellation-Microsoft Three Mile Island restart Sep-2024 (Sep-2028 target, 835 MW), Vistra Comanche Peak-Microsoft 2024, Meta-Entergy nuclear-RFP Dec-2024, Google-Kairos SMR PPA Oct-2024 (500 MW by 2035), Amazon-Dominion nuclear, Amazon-X-Energy SMR Oct-2024 ($500M), Oklo-DOE demonstration, IPO May-2024 $OKLO, TerraPower-DOE Natrium Kemmerer WY, NuScale $SMR-DOE, plus behind-the-meter gas, solar, storage, geothermal (Fervo-Google, Sage Geosystems-Meta), curtailed-wind stranded-gas Bitcoin-to-AI conversion (Crusoe origin story, Applied Digital, Iris Energy $IREN, Hut 8 $HUT, TeraWulf $WULF, Bit Digital $BTBT, Cipher Mining $CIFR).
Four unlocks: (1) OpenAI-Stargate $500B commit Jan-2025 (SoftBank-OpenAI-Oracle-MGX, Crusoe-Oracle Abilene TX first campus), BlackRock-Microsoft-MGX-GIP AI Infrastructure Partnership $30B Sep-2024 (expandable to $100B), Blackstone AirTrunk $16B Sep-2024, Apollo $25B, KKR $50B+ AI-infra allocation 2024 created the largest infrastructure-capital wave since 2000s telecom, 1990s power-plant buildouts. (2) CoreWeave $CRWV IPO Mar-2025 ($23B open, $86B backlog), Nebius $NBIS relisting Oct-2024, Cerebras S-1 Sep-2024 validated neocloud, AI-factory as a public-market category and gave every downstream operator (Lambda, Crusoe, Together AI, Applied Digital) a listed comp for private-round pricing, IPO-track. (3) Blue Owl-Morgan Stanley-Voyager Space $1.8B GPU-ABS Aug-2024 (first-ever GPU-collateralized ABS), Lambda-Macquarie $500M GPU-financed debt Apr-2024, CoreWeave-Blackstone-Magnetar-Coatue $7.5B debt May-2024, Applied Digital-Macquarie $5B perpetual-preferred Sep-2024, Crusoe-Blue Owl-Primary Digital Infrastructure $3.4B JV Jul-2024 opened a new asset-class — GPU, hyperscaler-take-or-pay-collateralized private-credit, ABS — that unlocks 60-80% debt-financing on new campuses vs. 30-50% for legacy colo. (4) Nuclear-PPA, SMR, behind-the-meter power renaissance (Constellation-Microsoft Three Mile Island restart Sep-2024, Talen-Amazon $650M Cumulus Mar-2024, Meta-Entergy RFP Dec-2024, Google-Kairos Oct-2024, Amazon-X-Energy $500M Oct-2024, Oklo $OKLO IPO May-2024, Vistra-Microsoft Comanche Peak), Fervo-Google, Sage Geosystems-Meta geothermal, curtailed-wind stranded-gas conversion (Crusoe, Applied Digital, Iris Energy, Hut 8, TeraWulf, Cipher Mining) solved the 500 GW ERCOT, PJM, MISO, SPP interconnection-queue bottleneck for 100-1000 MW campuses on 24-48 month timelines vs. 5-10 year utility-grid connections.
Seed, Series A $10-50M for founding-team, first 5-20 MW campus, 1-5k H100 cluster, 1-2 hyperscaler-frontier-lab anchor-tenant, interconnection-queue-position, Nvidia Preferred Partner tier (Voltage Park Jed McCaleb-funded 24k H100 cluster launched 2024, Fluidstack Google-anchored, TensorWave AMD MI300X-anchored as benchmarks). Series B $50-300M for 20-100 MW online, 5-25k GPU, $50-300M contracted-revenue-backlog, 3-5 hyperscaler-frontier-lab anchor-tenants, published PUE, WUE, carbon-intensity, SOC 2, FedRAMP-track (Together AI $305M B Feb-2024 Salesforce-Coatue-Nvidia $3.3B, Lambda $480M D Feb-2024 USIT-Andra $1.5B+ as benchmarks). Series C+ / infrastructure-round $300M-$3B for 100-500 MW online, 25-100k GPU, $1-10B contracted-backlog, 5-15 hyperscaler-frontier-lab, F500 anchor-tenants, GPU-backed-debt, ABS access (Crusoe $600M D Dec-2024 $2.8B valuation, Nebius $1.5B PIPE Dec-2024 Nvidia-Accel-Orbis, VAST Data $9.1B Nov-2023 Fidelity-Sequoia-Nvidia, Weka $140M E May-2024 $1.6B). Growth, infrastructure JV, GPU-ABS, private-credit $3B-$30B+ for 500-2000 MW, 100-500k GPU, $10-100B contracted-backlog, IPO-track (CoreWeave-Blackstone-Magnetar-Coatue $7.5B debt May-2024, $650M revolver Nov-2024, $2.3B extension Nov-2024, $CRWV IPO Mar-2025 $23B open, Crusoe-Blue Owl-Primary Digital Infrastructure $3.4B JV Jul-2024 for Stargate Abilene, Applied Digital-Macquarie $5B perpetual-preferred Sep-2024, Blue Owl-Morgan Stanley-Voyager Space $1.8B GPU-ABS Aug-2024). Non-dilutive, government: US DOE Loan Programs Office ($20B+ Title 17 AI-infra, Advanced Nuclear, long-duration-storage), US DOD Defense Production Act Title III, CHIPS Act (indirect via Intel, TSMC, Samsung, Micron, GlobalFoundries), UK AI Growth Zones, AI Opportunities Action Plan Jan-2025 (£14B private-commit), EU AI Factories Initiative Sep-2024 (€10B, seven EuroHPC AI factories), Japan METI AI-Compute Subsidy (¥725B), Korea AI-Compute (₩9T), India IndiaAI Mission (₹10,372 Cr), UAE MGX, G42 sovereign-AI fund $100B target.
Building a neocloud without signed hyperscaler, frontier-lab, F500 take-or-pay contracts before construction — spec-built GPU capacity gets stranded when Nvidia allocation-schedule slips or hyperscaler demand-forecast rolls off, and infrastructure-fund, GPU-backed-debt, ABS underwriters require 60-95% contracted-revenue-visibility (5-10 year WACL) as gating for capital. Ignoring the Nvidia Preferred Partner tier, direct-allocation vs. Supermicro-Dell-HPE-Lenovo-channel dynamics — sub-Preferred-Partner operators wait 6-12+ months for H200, B200, GB200, Blackwell Ultra allocation while CoreWeave, Nebius, Lambda, Crusoe, Together AI, G42-Cerebras get first-in-line, and Nvidia NVentures strategic-investment is the fastest signal-boost. Underestimating power, interconnection-queue, PUE, WUE, local-air, water-use, FERC, NERC-CIP, EPA permitting — 500 GW nationwide ERCOT, PJM, MISO, SPP, WECC, SERC, NERC interconnection-queue backlogs push 100-1000 MW campus energization to 2027-2032 without behind-the-meter gas, nuclear-PPA, SMR, geothermal, curtailed-wind stranded-gas conversion, and every failed permit or PUE > 1.4 disclosure decertifies enterprise, F500 procurement. Building a bitcoin-miner-to-AI-conversion play without HPC-grade power, cooling, network, operational-readiness — legacy PoW mining sites are optimized for cheap-power, low-uptime, air-cooled ASIC racks, and HPC-grade AI-factory conversion requires 100-250 kW/rack liquid-cooling, Nvidia InfiniBand NDR/XDR, Nvidia Preferred Partner, SOC 2, ISO 27001, FedRAMP, IL5/IL6, HITRUST, PCI-DSS, HIPAA-BAA compliance (Applied Digital, Iris Energy, Hut 8, TeraWulf, Cipher Mining, Core Scientific, Galaxy Digital-Amazon Helios TX have made the pivot; MARA, Riot, CleanSpark are earlier). Ignoring the BIS Diffusion Framework Jan-2025, ITAR, EAR export-controls on H100/H200/B200 to China, Russia, Tier 1/2/3 country, entity restrictions — a single BIS export-license violation or Entity List placement kills hyperscaler, sovereign, F500 tenant trust. Underestimating the GPU-depreciation, Blackwell-Rubin-refresh cycle — H100, H200 residual-value curves are still uncertain (target 25-40% residual at year-5 vs. 15-30% risk-case), and GPU-backed-debt, ABS structures require conservative-residual, tenant-diversification, refresh-CapEx-reserve to survive downside cases. Ignoring the sovereign-AI, national-AI-factory, EU AI Factories, Japan METI, Korea, India IndiaAI, UAE MGX, Saudi PIF pattern — most non-US neocloud, AI-factory demand will be sovereign-anchored, and US-only operators without EU, Middle East, Asia, LATAM regional-partnership, data-sovereignty, FedRAMP, IL5/IL6, EU AI Act, China-decoupled-supply-chain posture will lose the $50B+ sovereign-AI-buildout wave.
Investor directory · Fundraising library · Articles A–Z · Company funding database