Microsoft's $60B Neocloud Bet: Buying Time in the AI Capacity Crisis

December 12, 2025

December 2025 Update: Microsoft has committed over $60 billion to neocloud providers, with $23 billion going to British startup Nscale alone for 200,000 GB300 GPUs. Azure's capacity crunch now extends into mid-2026. The hyperscaler's pivot to renting AI infrastructure signals a fundamental shift in how cloud giants scale.

TL;DR

Microsoft's inability to build data centers fast enough has forced a strategic pivot. The company now rents AI capacity from specialized "neocloud" providers, committing $60B+ across multi-year deals with Nscale, Nebius, CoreWeave, Iren, and Lambda. The Nscale deal alone provides 200,000 GB300 GPUs across four countries. The commitment represents the largest infrastructure outsourcing play in cloud history, and it validates the neocloud model as critical AI infrastructure.

What Happened

Microsoft crossed $60 billion in neocloud commitments in November 2025, according to Bloomberg reporting.¹ The spending doubled from approximately $33 billion in October, reflecting what Microsoft executives describe as "land-grab mode" for AI compute.²

The largest commitment: $23 billion to British startup Nscale for approximately 200,000 NVIDIA GB300 GPUs deployed across the United Kingdom, Norway, Portugal, and Texas.³

"We are in very much land-grab mode in the AI space," stated Microsoft Cloud EVP Scott Guthrie.⁴

The deals accelerated after Microsoft CFO Amy Hood acknowledged during October earnings that Azure demand remains "significantly ahead of the capacity we have available."⁵ She added: "I thought we were going to catch up. We are not."

CEO Satya Nadella diagnosed the bottleneck: "It's actually power that's the bottleneck."⁶ Renting from neoclouds sidesteps years of data center construction by tapping providers who have already secured power and chips.

Why It Matters for Infrastructure

Validation of Neocloud Model: Microsoft's $60B commitment legitimizes neoclouds as essential AI infrastructure, not just niche players. CoreWeave, Nebius, Nscale, Lambda, and Iren now operate at hyperscaler scale with hyperscaler contracts.

Why Neoclouds Move Faster: Neoclouds secured power and land before the 2023-2024 AI surge. Microsoft must now permit, build, and interconnect new facilities from scratch (3-5 year cycle). Neoclouds already have sites with power; they only need to install GPUs (6-18 month cycle). The constraint is physical infrastructure, not capital.

Capacity Crunch Is Real: Despite spending $34.9 billion on capex in Q1 FY25 alone—up $10 billion from the previous quarter—Microsoft cannot build fast enough.⁷ The shortage extends into mid-2026 across key regions including Northern Virginia and Texas.

Power Is the Constraint: Neoclouds win because they secured power agreements before the AI surge. Nscale's Texas campus can scale to 1.2GW—more power than many traditional data center campuses combined.

Multi-Year Commitments: Most deals run five years, locking Microsoft into long-term dependency on third-party infrastructure. The multi-year contracts represent structural change, not a temporary fix.

Technical Details

Microsoft's Neocloud Portfolio

Nscale Deployment Details

The Nscale relationship, Microsoft's largest, spans four countries:⁸

Microsoft holds an option on a second Texas phase of 700MW starting late 2027.⁹

Why Neoclouds Win on Economics

Neoclouds offer significant cost advantages over traditional hyperscalers:¹⁰

• NVIDIA A100 40GB: $1.39/hr (CoreWeave) vs $3.67/hr (Azure/GCP)—62% savings
• Specialized AI infrastructure without legacy general-purpose overhead
• GPU-collateralized financing enables faster capital deployment

CoreWeave Market Position

The neocloud leader provides context for the sector:¹¹

• IPO: March 2025 at $40/share—largest tech IPO since 2021
• Revenue: $1.36B in Q3 2025 (+133% YoY)
• Backlog: $55.6 billion contracted (10 years visibility)
• Debt: $18.8 billion in GPU-collateralized facilities
• Investors: Nvidia owns 6.56% stake (~$2.1B)

The Capacity Crisis Context

Microsoft's neocloud pivot responds to a documented crisis:¹²

Demand Outpaces Supply: Microsoft CTO Kevin Scott called the situation "a massive crunch"—understating, he added, the challenge of building "capacity fast enough since ChatGPT launched."

Regional Constraints: Northern Virginia and Texas Azure regions limit new subscriptions due to space, rack, and server shortages.

Spending Isn't Solving It: Even $85 billion annual capex cannot close the gap.⁷ Physical construction timelines measure in years; AI demand growth measures in months.

Competitors Circling: Analysts warn that AWS and Google Cloud could capture overflow demand, reshaping cloud market share during the constraint period.

What's Next

Q3 2026: Nscale Texas campus delivers first 104,000 GB300 GPUs to Microsoft.

Mid-2026: Azure capacity constraints expected to ease—though Microsoft has missed previous timeline estimates.

H2 2026: Nscale considering IPO, which would create another major public neocloud alongside CoreWeave and Nebius.

Late 2027: Microsoft option on additional 700MW Texas capacity from Nscale.

Structural Question: Will Microsoft ever return to primarily self-owned AI infrastructure, or does the neocloud model become permanent?

Key Takeaways

For infrastructure planners: - Neocloud relationships now operate at hyperscaler scale ($10B+ contracts) - Power availability determines AI infrastructure location more than any other factor - GB300 deployments concentrated in regions with secured power agreements - Five-year contract structures create planning visibility

For operations teams: - Neocloud providers offer 62% cost advantage for comparable GPU access - Multi-region deployment strategies reduce concentration risk - Liquid cooling standard for GB300-class deployments - Monitor Nscale Texas, Nebius, and Iren capacity timelines

For strategic planning: - Microsoft's pivot validates neocloud model for enterprise consideration - Azure capacity constraints extend into mid-2026 minimum; plan accordingly - Power, not chips, now limits AI infrastructure expansion - Neocloud IPO pipeline (Nscale H2 2026) creates new investment opportunities - Consider neocloud partnerships for speed; hyperscalers for long-term stability

References

For AI infrastructure deployment and GPU capacity solutions, contact Introl.