Microsoft tightens oversight and trims Neocloud deals amid persistent AI data center delays
Microsoft is conducting a comprehensive internal review of its multi-billion-dollar computing contracts with specialized "neocloud" providers, scaling back several third-party agreements. The strategic pivot follows widespread construction delays, power connection bottlenecks, and rising capital costs that forced Microsoft to record a $5 billion shortfall in anticipated AI infrastructure spending during the previous quarter.
The recalibration marks a cooling period in Big Tech’s aggressive reliance on external GPU-leasing startups, with Microsoft increasingly prioritizing its own proprietary data center builds and balance-sheet-backed infrastructure.
Scaling Back Commitments and Winding Down Partnerships
The shift has led Microsoft to reevaluate or walk away from several high-profile capacity agreements with specialized infrastructure vendors:
Nscale Expansion Decline: Microsoft opted not to exercise an option to expand its computing capacity agreement with UK-based Nscale at a separate site in Ward County, Texas, and previously withdrew as an anchor customer for Nscale’s proposed data center in Loughton, Essex, after power-connection delays pushed expected electricity availability into the 2030s.
Oracle Cloud Adjustments: Microsoft is winding down portions of its arrangement to rent external AI computing capacity from Oracle, reallocating resources toward internal data center footprints.
Increased Scrutiny on Nebius and IREN: Projects tied to Nebius and IREN have faced tighter oversight and delivery postponements. IREN confirmed that certain AI infrastructure deliveries were pushed back following network redesign requests initiated by Microsoft.
Impact on Capital Expenditures and Delivery Timelines
During Microsoft's fiscal first-quarter earnings call, Chief Financial Officer Amy Hood acknowledged that third-party capacity deliveries had slipped, pushing capital expenditures and finance lease payments to $45.2 billion.
While approximately three-quarters of those outlays went toward short-lived AI hardware, primarily CPUs and GPUs, the shifting delivery schedules caused spending to lag initial projections. Leadership noted that many of these delayed capacity blocks have now shifted into the second half of the fiscal year, driving projected capital expenditures past $50 billion for the upcoming quarter.
Broader Market Pressure on Neocloud Providers
The tightening of Microsoft's wallet has created immediate downstream pressure across the neocloud ecosystem. Specialized GPU providers and infrastructure startups rely heavily on multi-year contracts with hyperscale anchor tenants to secure debt financing for massive server acquisitions.
Following news of Microsoft's contract reviews and scaling-back efforts, valuations for specialized cloud operators and supporting hardware partners have faced heightened volatility. The friction underscores a broader maturation phase in the AI boom, where hyperscalers are demanding strict execution discipline, reliable power supply timelines, and strict cost efficiencies before deploying capital into external cloud networks.
