us datacenter outlook 2025-2030

overview

the period 2025-2030 represents the most ambitious private infrastructure buildout in us history, with 99 projects already announced for 2025 onwards totaling $616.0 billion investment and 54.4 gw power capacity, while 146 projects currently under construction or planned will deliver 130+ gw of new capacity. this analysis examines the pipeline, projected growth, technology evolution, and critical challenges facing the industry.

forward pipeline summary

key predictions

1. total market: $2-3 trillion by 2030 (from$1.1t documented today)
2. ai dominance: 50%+ of new capacity ai/ml focused
3. nuclear deployment: 24+ gw smr capacity online 2027-2030
4. gigawatt standard: 1-5 gw projects become normal
5. geographic transformation: beyond traditional hubs to power-rich regions

2025 outlook

announced 2025 projects

already committed: 99 projects, $616.0b investment, 54.4 gw

expected 2025 completions

43 projects expected operational in 2025:

major operational dates:

• q1 2025: 12 projects including microsoft azure regions, vantage va2
• q2 2025: 14 projects including coreweave facilities, databank culpeper
• q3 2025: 9 projects including aws expansions
• q4 2025: 8 projects including powerhouse 95 first phase (150 mw), cleanarc va1 groundbreaking

2025 investment forecast

projected annual investment: $180-220 billion

• q1: $45b (strong start, continued 2024 momentum)
• q2: $55b (peak announcement season)
• q3: $60b (construction season)
• q4: $50b (planning for 2026)

investment drivers:

• ai infrastructure urgency
• nuclear smr partnerships
• geographic diversification
• power capacity race

2025 technology milestones

gpu evolution:

• nvidia b200/b300 volume production
• 200,000+ gpu clusters become standard
• 1 million gpu installations industry-wide

cooling systems:

• liquid cooling universal for ai (100% adoption)
• immersion cooling: 25-30% of new ai capacity
• 300+ kw/rack capability demonstrated

power solutions:

• first smr construction begins (constellation tmi restart)
• on-site natural gas: 60%+ of gigawatt projects
• grid interconnection: still 2-3 year timelines

2025 challenges

power crisis intensifies:

• utility queues exceed 400 gw nationwide
• regulatory reform required but slow
• ratepayer protection movements gain strength
• renewable mandate conflicts with ai demand

construction capacity:

• labor shortages worsen
• material costs inflate 15-20%
• timeline pressure increases (sub-12 months demanded)
• quality concerns emerge

community opposition:

• organized resistance in multiple states
• water usage becomes primary concern
• tax incentive backlash spreads
• local moratoriums attempted

2026-2027: nuclear renaissance

smr deployment timeline

24+ gw committed through smr partnerships:

2026 expected completions

45 projects projected operational:

capacity additions:

• total: 18.2 gw
• ai/ml: 11.4 gw (63%)
• traditional: 6.8 gw (37%)

major 2026 projects:

• multiple gigawatt-scale ai campuses
• nuclear restart preparations
• hyperscale cloud expansions
• colocation capacity doubles

2027 projections

30 expected completions:

nuclear milestone: three mile island restart (837 mw) - first ai-dedicated nuclear power

capacity forecast:

• total: 14.8 gw
• nuclear-powered: 1.2 gw (first meaningful nuclear contribution)
• ai-focused: 10.1 gw (68%)

technology evolution:

• next-generation gpus (post-b300)
• 400+ kw/rack immersion cooling standard
• 800 gbps networking universal
• autonomous datacenter operations pilot

2028-2030: maturation phase

2028 outlook

expected completions: 16 projects

key developments:

• smr deployments accelerate (5+ reactors operational)
• 2-5 gw campuses normal
• liquid cooling universal (100%)
• gpu counts: 500k-1m clusters common

capacity forecast: 8.4 gw total

• nuclear-powered: 2.8 gw (33%)
• ai/ml: 6.1 gw (73%)

2029 projections

expected completions: 6 projects

market maturity indicators:

• consolidation begins (smaller operators acquired)
• standardization emerges (power, cooling, design)
• geographic diversification complete
• regulatory framework stabilizes

capacity forecast: 3.2 gw total

• focus shifts from quantity to efficiency
• retrofit/upgrade market emerges
• nuclear baseload standard

2030 endpoint

expected completions: 6 projects

cumulative 2025-2030 additions: 250-300+ gw

• 2025: 43 projects
• 2026: 45 projects
• 2027: 30 projects
• 2028: 16 projects
• 2029: 6 projects
• 2030: 6+ projects

total us datacenter capacity by 2030: 350-400 gw

• ai/ml: 175-200 gw (50%)
• traditional cloud: 100-125 gw (30%)
• enterprise/colocation: 75-100 gw (20%)

investment outlook 2025-2030

total projected investment

cumulative investment by 2030: $2.0-2.3 trillion (including 2020-2024)

investment composition

by category (2025-2030 cumulative):

investor composition

expected capital sources:

technology roadmap 2025-2030

compute evolution

2025-2026: nvidia b200/b300 era

• 100-200k gpu clusters standard
• power: 700-1,000w per gpu
• rack density: 100-140 kw

2027-2028: next-generation accelerators

• 300-500k gpu clusters
• power: 500-800w per gpu (efficiency gains)
• rack density: 140-200 kw

2029-2030: post-moore innovation

• optical interconnects
• photonic computing pilots
• quantum-classical hybrid
• rack density: 200-300+ kw

cooling technology

2025: liquid cooling universal

• direct liquid cooling (dlc): 70%
• immersion cooling: 30%
• air cooling: legacy only

2026-2027: immersion dominance

• immersion: 60% of new ai capacity
• dlc: 35%
• advanced phase-change: 5%

2028-2030: next-generation

• two-phase immersion standard
• 400+ kw/rack capability
• waste heat recovery (district heating)
• closed-loop systems

power systems

2025-2026: natural gas bridge

• on-site generation: 65% of gigawatt projects
• grid-connected: 35% (where available)
• renewable ppas: widespread but insufficient

2027-2028: nuclear begins

• smr first deployments: 2-3 gw total
• combined cycle natural gas: still majority
• grid improvements: limited progress

2029-2030: nuclear scale

• smr deployments accelerate: 10-15 gw total
• hyperscaler-owned nuclear common
• renewable + storage: niche applications
• grid dependency reduces

networking evolution

2025-2026: 400-800 gbps standard

• 400 gbps ethernet universal
• 800 gbps for large clusters
• infiniband hddr (400 gbps)

2027-2028: terabit networking

• 800 gbps standard
• 1.6 tbps deployments
• silicon photonics integration

2029-2030: next-generation

• 1.6 tbps standard
• 3.2 tbps cutting edge
• optical switching
• co-packaged optics

geographic transformation

emerging leaders (2025-2030 growth)

traditional hubs evolution

northern virginia:

• current: 10.2 gw, 40 gw pipeline
• 2030: 25-30 gw (growth slows due to constraints)
• focus: mission-critical, low-latency applications
• challenges: power limits, community opposition

california:

• current: silicon valley established base
• 2030: modest growth (regulatory constraints)
• focus: innovation, ai research
• challenges: cost, permitting, water

oregon/washington:

• current: traditional hyperscale base
• 2030: steady growth limited by grid
• focus: traditional cloud workloads
• advantages: renewable power, cooling climate

new frontiers

midwest expansion:

• ohio, indiana, illinois, wisconsin
• advantages: manufacturing sites, power access
• 2030 total: 35-45 gw

southeast growth:

• georgia, north carolina, south carolina
• advantages: business climate, incentives
• 2030 total: 25-30 gw

mountain west:

• wyoming, montana, idaho
• advantages: renewable power, land availability
• 2030 total: 15-20 gw

market structure evolution

consolidation trends (2025-2028)

operators:

• top 5 control 60%+ of capacity by 2028
• mid-tier consolidation (10-20 acquisitions)
• specialist ai operators emerge and consolidate
• reit model questioned

financial sponsors:

• infrastructure funds dominate (70%+ of deals)
• sovereign wealth fund entry accelerates
• strategic tech investors (nvidia, others) continue
• traditional pe exits shorten (4-year cycles)

vertical integration

hyperscaler evolution:

• microsoft, google, amazon, meta build own infrastructure
• third-party capacity: specialty ai workloads only
• design standardization
• supply chain integration

operator response:

• powered shell model dominates
• long-term leases standard (15-20 years)
• ai-specific facilities separate business
• technology partnerships required

new business models

capacity-as-a-service:

• gpu clusters as subscription
• power + cooling + compute bundled
• flex capacity for training bursts
• inference optimization services

power brokering:

• datacenter operators as power buyers
• ppa aggregation
• renewable + storage + nuclear portfolios
• grid services revenue

regulatory landscape 2025-2030

federal evolution

energy policy:

• doe ai datacenter task force (2025)
• nuclear licensing reform (2026-2027)
• ferc interconnection streamlining (2027)
• renewable energy integration mandates

environmental:

• epa emissions standards for datacenters (2026)
• water usage reporting requirements (2027)
• waste heat recovery incentives (2028)
• carbon accounting standardization

state responses

pro-growth states (texas, arizona, utah):

• streamlined permitting (sub-6 months)
• competitive incentive packages
• utility cooperation mandated
• workforce development programs

restrictive states (california, virginia):

• environmental review extended
• community approval required
• ratepayer protection strengthened
• growth caps discussed/implemented

local governance

2025-2026: resistance peaks

• moratoriums attempted (many fail)
• special taxation proposals
• water usage restrictions
• nimby movements organized

2027-2030: accommodation

• datacenter overlay zoning districts
• negotiated community benefits
• employment targets
• infrastructure contributions

risk factors

power delivery failure

risk: utilities cannot deliver promised capacity on schedule

• probability: 40-50%
• impact: project delays 2-4 years, $100b+ investment at risk
• mitigation: on-site generation, nuclear partnerships

nuclear delays

risk: smr deployments miss 2027-2030 targets

• probability: 60-70% (some delays inevitable)
• impact: continued natural gas dependency, emissions targets missed
• mitigation: diverse vendor portfolio, traditional nuclear restarts

regulatory backlash

risk: state/local moratoriums, restrictive legislation

• probability: 30-40%
• impact: geographic concentration, cost inflation
• mitigation: community engagement, economic benefits communication

technology disruption

risk: ai efficiency gains reduce infrastructure needs

• probability: 20-30%
• impact: stranded assets, overcapacity
• mitigation: flexible designs, multi-workload capability

economic recession

risk: ai investment bubble deflates

• probability: 25-35%
• impact: project cancellations, funding challenges
• mitigation: hyperscaler demand relatively stable

geopolitical shocks

risk: chip supply chain disruption, energy crisis

• probability: 20-30%
• impact: construction delays, cost overruns
• mitigation: domestic manufacturing, diverse suppliers

upside scenarios

ai demand exceeds projections

scenario: ai capabilities expand faster than expected

• training requirements: 2-3x current projections
• inference explosion: edge + cloud hybrid
• new modalities: video, robotics, scientific computing
• result: 400-500 gw by 2030 (vs 350-400 gw base case)

nuclear breakthrough

scenario: smr deployments accelerate beyond expectations

• regulatory streamlining succeeds
• multiple vendors achieve commercial scale
• costs decline with volume
• result: 40-50 gw nuclear by 2030 (vs 20-25 gw base case)

efficiency gains reinvested

scenario: moore’s law equivalent for ai efficiency

• 2x efficiency every 18 months
• savings reinvested in scale
• new applications enabled
• result: 1,000+ gw by 2035

key takeaways

transformation scale

• investment: $1.0-1.2 trillion 2025-2030 (cumulative$2.0-2.3t by 2030)
• capacity: 250-300 gw additions (350-400 gw total us by 2030)
• ai share: 50%+ of new capacity
• nuclear: 24+ gw smr deployments critical path

critical success factors

1. power delivery: can utilities and nuclear scale?
2. regulatory reform: will permitting and interconnection accelerate?
3. technology: can efficiency gains keep pace with demand?
4. capital: is $1+ trillion investment sustainable?
5. talent: can industry scale workforce fast enough?

strategic imperatives

for hyperscalers:

• secure power capacity now (2-3 year lead times)
• vertical integration vs third-party capacity decision
• nuclear partnerships essential
• geographic diversification required

for operators:

• ai-specific infrastructure differentiation
• power sourcing competitive advantage
• technology partnerships critical
• consolidation inevitable

for investors:

• long-term infrastructure hold (10+ years)
• power-rich geographic focus
• nuclear exposure required
• selectivity essential (not all projects succeed)

for policymakers:

• balance growth with community/environment
• power infrastructure investment urgent
• nuclear licensing reform critical
• workforce development programs needed

---

the period 2025-2030 will determine whether the us builds the infrastructure foundation for ai leadership. success requires unprecedented coordination between private industry, utilities, regulators, and communities. the scale of investment and construction has no peacetime precedent.