From a Reddit post to $120M ARR, from crypto mining to a $23B IPO — how people actually get into the data center business.
Founded 2017 by Michael Intrator, Brian Venturo, Brannin McBee. All three were commodities/energy futures traders. Intrator co-founded Hudson Ridge Asset Management (natural gas hedge fund) and spent 15+ years at Natsource managing climate-focused capital. They bought their first GPU for crypto mining in 2016, closed the hedge fund, launched Atlantic Crypto in 2017, renamed to CoreWeave after the 2018 crypto crash, and pivoted GPUs to cloud compute.
Founded 2018 by Chase Lochmiller (CEO) and Cully Cavness (President). Lochmiller: Stanford MS in AI, MIT math/physics, quant researcher at GETCO and Jump Trading, GP at Polychain Capital. Cavness: Oxford MBA, geology BA from Middlebury, oil/gas investment banking at Petrie Partners. Childhood friends who conceived the idea during a hiking trip.
Founded 2012 by Stephen Balaban. CS + Economics from University of Michigan. First engineering hire at Perceptio (acquired by Apple). Built face recognition for iPhone GPUs. Started as ML tools company, expanded to GPU cloud.
Led by Wes Cummins, who grew up on a 200-person Idaho potato farm. 20+ years as technology investor. Pivoted company from Bitcoin mining to AI DCs in 2022 (before ChatGPT). The company had bounced between industries for 20+ years before finding its niche.
Founded 2015 by Andrew Feldman and four others. Feldman previously co-founded SeaMicro (acquired by AMD for $357M), VP at Force10 Networks (sold to Dell for $800M). Stanford BA + MBA. All five founders worked together at SeaMicro.
| Company | Founder(s) | Background | Key Insight |
|---|---|---|---|
| Equinix | Jay Adelson, Al Avery | Netcom (early ISP), Digital Equipment Corp, BU film/CS | Carrier-neutral interconnection |
| Digital Realty | GI Partners | PE firm; acquired 21 DCs from bankruptcy at 20-40% discount | Distressed asset acquisition post-dot-com |
| QTS | Chad Williams | Quality Group (real estate family office) | Real estate development background |
| Aligned | Jakob Carnemark | 11 years at Skanska Mission Critical | Construction expertise → patented cooling |
| Background | Examples |
|---|---|
| Finance / Trading | CoreWeave, Applied Digital, Crusoe |
| Previous exits in hardware | Cerebras (SeaMicro $357M) |
| Network engineering / ISPs | Equinix |
| Real estate / Construction | QTS, Aligned, Digital Realty |
| Crypto mining pivot | CoreWeave, Applied Digital, RunPod |
| Deep tech / ML engineering | Lambda, Cerebras |
| Role | Experience | Salary Range |
|---|---|---|
| DC Technician I | 0-2 years | $45K-$57K |
| DC Technician II-III | 2-5 years | $57K-$80K |
| DC Technician IV / Senior | 5-8+ years | $81K-$147K |
| DC Engineer | 3-7 years | $80K-$130K |
| Network Engineer | 3-7 years | $75K-$130K |
| Facilities Manager | 5-10 years | $90K-$140K |
| Operations Manager | 5-10 years | $75K-$160K |
| DC Architect | 8-15 years | $155K-$281K |
| VP of Infrastructure | 12-20 years | $200K-$350K+ |
| CTO / SVP Engineering | 15+ years | $250K-$500K+ |
| Domain | What Matters | Why |
|---|---|---|
| Electrical Engineering | UPS, power distribution (N+1, 2N), switchgear, generators | 40-45% of total DC build cost |
| Mechanical / Cooling | Airflow, HVAC, liquid cooling (D2C, immersion), heat rejection | AI racks >100 kW make air cooling inadequate |
| Networking | InfiniBand, 400G/800G Ethernet, BGP/OSPF, fiber optics | Internal fabric + customer connectivity |
| Systems / Compute | GPU architectures, NVMe, Kubernetes, bare metal provisioning | Managing AI infrastructure |
| Domain | What Matters | Why |
|---|---|---|
| Real Estate Development | Site selection, zoning, entitlements, construction management | Many successful founders came from real estate |
| Energy Procurement | PPAs, utility tariffs, grid interconnection, behind-the-meter | Power is the #1 competitive advantage |
| Financial Modeling | CapEx/OpEx per MW, IRR, project finance, REIT economics | $8M-$20M+ per MW to build |
| Project Management | 18-36 month builds, MEP coordination, commissioning | Transformer lead times: 24-40 weeks |
| Budget | Approach |
|---|---|
| $5K-$50K | GPU rental/resale, managed services layer |
| $50K-$500K | Small GPU cluster in colocation |
| $500K-$5M | Multi-rack deployment, edge datacenter |
| $5M-$50M | Small colocation facility or significant GPU cloud |
| $50M+ | Wholesale datacenter development |
Zhen Lu and Pardeep Singh started as Comcast developers mining Ethereum with ~$50K in rigs. Built a GPU cloud in 3 months, launched on Reddit. $1M revenue in 9 months. Revenue-share partnerships with existing DCs. Today: $120M ARR, 500K developers.
You do not need to build a datacenter. Start as a software layer on top of existing colocation.
AI DCs generate up to 25x more revenue per kWh than Bitcoin mining. Converting mining halls to GPU racks takes weeks. Miners can earn $1-4M per MW annually (hosting) or $10M+/MW (operating GPUs directly).
Start with one rack ($79-$599/month per unit). Build managed services on top. Retail colo is the most lucrative segment per-rack.
10-100 racks. Power density 10-15 kW/rack for AI edge. Target latency-sensitive workloads in underserved markets. Modular units deploy in months.
Provision bare metal from providers like OpenMetal, deploy OpenStack multi-tenant cloud. Offer dedicated hardware, GPU isolation (MIG), compliance features.
Classification varies: light industrial, commercial, or special use. Key issues: setbacks, height restrictions, noise limits (typically 60 dBA day / 55 dBA night), buffer yards, parking ratios.
Structural, electrical, mechanical, fire safety approval required. Server areas: Group S (Storage). Office areas: Group B (Business).
Air quality permits (generators), water restrictions, stormwater management, NEPA if federal involvement.
6 months in DC-friendly jurisdictions to 2+ years in constrained markets. Emerging markets (rural Texas, Idaho, Louisiana) attract investment partly due to faster permitting. See Regulation for details.
| Partner Type | Why Critical | Key Players |
|---|---|---|
| Utility / Power | Power is the moat. Energization delays cost ~$1.5M/MW/month | Regional utilities, PPA providers |
| Fiber / Connectivity | Multiple diverse fiber paths for redundancy | Lumen, Zayo, Crown Castle, regional providers |
| Equipment Vendors | GPU/server supply, cooling, power systems | NVIDIA, Dell, Supermicro, Vertiv, Schneider, Eaton |
| Construction | Mission-critical is specialized | Holder, DPR, Fortis, Skanska |
| Anchor Tenants | De-risks project, enables financing | Hyperscalers, CoreWeave |
| Financial Partners | Capital-intensive business | Macquarie, Blackstone, KKR, Brookfield |
Critical: high-power transformers face 36-month backlogs. Chillers/breakers: 24-40 week lead times.
| Tier | Uptime | Redundancy | Target |
|---|---|---|---|
| I | 99.671% | None | Small business, dev |
| II | 99.741% | Partial | SMB, non-critical |
| III | 99.982% | N+1 | Enterprise, finserv |
| IV | 99.995% | 2N | Mission-critical, government |
| Customer | Required Certs |
|---|---|
| Startups/SMBs | Basic security + SLA |
| Mid-market enterprise | SOC 2 Type II + Tier III |
| Financial services | SOC 2 + ISO 27001 + Tier III + PCI DSS |
| Healthcare | HIPAA + SOC 2 + ISO 27001 |
| Government | FedRAMP + FISMA + SOC 2 |
| AI/ML workloads | Performance SLAs matter more than traditional certs |
| Market | Advantage | Why Emerging |
|---|---|---|
| West Texas | Abundant land, cheap power | Hyperscaler expansion. Low cost/MW ($9.3M in San Antonio) |
| Idaho | Cheap hydroelectric | New entrants seeking power |
| Louisiana | Industrial power infrastructure | Energy sector pivot to AI |
| Oklahoma | Low energy costs, fiber | Central US latency advantages |
| Indianapolis | Growing tech ecosystem | Available power |
| Kansas City | Fiber hub (Google Fiber) | Connectivity + power combo |
| Region | Key Markets | Opportunity |
|---|---|---|
| Southeast Asia | Malaysia (Johor), Indonesia, Thailand, Vietnam | Fastest growing globally. Microsoft $2B in Malaysia. $3B Indonesia market by 2026 |
| Latin America | Brazil (Sao Paulo), Chile, Colombia, Mexico | Brazil Redata reducing tariffs to 2% |
| Africa | South Africa, Kenya, Nigeria | $10-20B investment needed (McKinsey). Energy constraints main challenge |
| Central/Eastern Europe | Poland, Baltics, Czech Republic | Significantly underserved despite EU investment. $7.92B by 2029 |
Central and Eastern Europe is a significant gap. Most of CEE remains blank spots despite favorable conditions, EU sovereignty push, and growing demand. EU planning EUR 20B for AI gigafactories, with Poland and Baltics pushing to secure investment.