Microsoft weighs fewer neocloud suppliers amid delivery risks | Magica
Microsoft weighs fewer AI cloud suppliers as delivery risks test big contracts
Editorial Team
••📖8 min read
Microsoft is evaluating fewer neocloud partners while expanding its own infrastructure. Delivery-dependent contracts underpin supplier financing, while IREN reports rising revenue alongside a $6 billion funding target for its broader expansion.
Microsoft is evaluating a smaller roster of specialist AI cloud suppliers; it has not decided which neoclouds to stop using.
Delivery problems have different causes, including power access and a Microsoft-requested networking upgrade.
Supplier contracts support financing, while IREN reports rising revenue and customer prepayments alongside substantial further funding needs.
Microsoft, the software company behind the Azure cloud platform, is considering concentrating its neocloud purchases among fewer suppliers while expanding its own data centers, a report citing people familiar with the matter said on November 6. Two people said the evaluation had not yet produced a decision on which neoclouds it would stop using.
Neoclouds are specialist providers of AI computing capacity. Microsoft turned to them after ChatGPT’s 2022 launch drove demand for graphics processing units, or GPUs. The review tests a business model in which a major customer’s commitment helps suppliers finance infrastructure that must arrive on schedule before generating the expected revenue.
The evidence does not establish a broad cancellation of existing contracts or a retreat from AI spending. Microsoft reported $45.2 billion in capital expenditures and finance leases in its latest quarter and said spending would accelerate in the second half of its fiscal year, the report said. Nor do delivery problems fall entirely on suppliers: Microsoft’s changing requirements have contributed to at least one delay.
IREN’s company-reported funding plan for remaining FY27 spending. The $9 billion includes restricted cash, undrawn facilities and contracted prepayments not yet received; the additional $6 billion is targeted and subject to market conditions. Source: IREN earnings release.
Power shortages and changing specifications
Microsoft has bought capacity from companies including CoreWeave, Nebius, IREN and Nscale, as well as the larger cloud provider Oracle. Two people cited in the report said Microsoft opted not to renew an Oracle deal. That account does not establish an end to all business between the companies.
Microsoft said its infrastructure strategy combines owned data centers, leased facilities and third-party partnerships. Its spokesperson said it regularly reviews investments against customer demand, tracks delivery and adjusts plans as circumstances change.
Internet scans and leaked databases expose the spread of ARTEX, a security-testing framework used in cyberattacks. Researchers documented largely autonomous compromises, but the records also show failed cashouts and uncertain counts of independent deployments.
Editorial Team
Some projects have already fallen away. Microsoft abandoned an intended Nscale site in England because of electricity access, its U.K. and Ireland chief told CNBC in October, according to the November 6 report. That report also relayed a Financial Times account that Microsoft decided against renewing an Nscale contract in Texas. Nscale, a U.K.-based neocloud that filed to go public in September, had said in its prospectus that it expected Microsoft to become a significant customer.
IREN presents a different problem. The company supplies data centers, computing and software for AI training and inference, while retaining a Bitcoin-mining business. It owns underlying land, power and data center infrastructure. In its November 2025 announcement, IREN described a roughly $9.7 billion, five-year Microsoft GPU services contract with a 20% prepayment. The planned deployment covered four liquid-cooled facilities at Childress, Texas, supporting a combined 200 megawatts of critical IT load. Microsoft cited IREN’s integrated infrastructure and secured power capacity as reasons for the partnership.
IREN’s November 5 release said Horizon 1 had gone into service for Microsoft, with Horizon 2–4 in late-stage construction. The November 6 report’s account of IREN’s results, however, said capacity was handed over in August but a Microsoft-requested networking equipment upgrade pushed the infrastructure’s expected online date into January or February 2027. Those descriptions leave unclear precisely how much capacity is operating and how much awaits the upgrade.
Separately, IREN’s earnings release moved its broader target of more than $4 billion in annualized run-rate revenue from year-end 2026 to January 2027, potentially extending through February, citing a customer-requested change and supplier timing. That operating target assumes contracted GPU pricing and utilization; it is not recognized annual revenue. Commissioning, testing and customer acceptance still determine the revenue ramp.
Microsoft faces constraints in its own cloud business, too. Its finance chief said on the October earnings call that some GPU cloud contracts had slipped into the first half of 2027, affecting Azure growth, according to the November 6 report.
A customer commitment also finances construction
Nebius illustrates why the review matters beyond future sales. The Amsterdam-headquartered, Nasdaq-listed company builds AI cloud infrastructure and tools for developing and running models. Formerly named Yandex N.V., it changed its name in August 2024 after divesting its Russian assets. It was building its specialist AI cloud business before securing Microsoft as a major dedicated-capacity customer.
The September 2025 Microsoft agreement covered capacity at Vineland, New Jersey. In its announcement, Nebius said contract cash flow and debt secured against the agreement would finance associated capital expenditure, with borrowing terms helped by the customer’s credit quality.
Its contract filing valued the base agreement at about $17.4 billion through 2031, subject to deployment and availability, with additional purchases potentially raising it to about $19.4 billion. Capacity was scheduled in tranches during 2025 and 2026 over a five-year term.
The filing also disclosed Microsoft’s right to terminate an affected GPU service if Nebius misses delivery dates beyond a grace period and cannot provide alternative capacity. It includes damages for late delivery, and the parties’ obligations were to begin once Nebius confirmed it had secured any additional financing required. These are contractual conditions, not evidence that Microsoft has exercised termination rights.
The market backdrop is already difficult. CoreWeave, Nebius and IREN shares were each down roughly half over the preceding month, even after November 6 gains, the report said. It described rising borrowing costs and concern about heavy debt loads across the sector.
Customer concentration adds another vulnerability. CoreWeave, an AI infrastructure provider with cryptocurrency-mining roots, said at its 2025 flotation that Microsoft supplied 62% of its 2024 revenue, according to the November 6 report. That historical share shows its earlier dependence, rather than its current customer mix.
There are other buyers. OpenAI, the company behind ChatGPT, used only Azure at the chatbot’s launch; its cloud suppliers now also include Amazon, Cerebras, CoreWeave, Google and Oracle, the report said. Those alternatives broaden the customer landscape without establishing that any supplier can replace Microsoft’s purchases on equivalent terms.
IREN’s company-reported quarterly AI Cloud revenue rose from $71 million to $200 million. September revenue includes $135 million of services and $65 million of lease revenue; the prior quarter had no lease revenue. Source: IREN earnings release.
IREN’s growth comes with a funding distinction
IREN’s results resist a simple story of collapsing demand. For the quarter ended September 30, 2026, it reported $200 million in AI Cloud revenue, versus $71 million in the preceding quarter. The composition matters: services revenue rose to $135 million from $71 million, while a new lease-revenue contribution supplied $65 million. Total AI Cloud growth therefore includes a change in revenue mix.
IREN nevertheless reported a $263 million company-wide net loss. Strong sales growth does not establish that its expansion is profitable or self-financing.
Its remaining planned fiscal 2027 capital expenditure was $15 billion, measured from September 30. IREN identified $9 billion in cash, committed GPU financing and prepayments, and targeted another $6 billion from GPU financing, customer prepayments, data center financing and other corporate sources, subject to market conditions. The $9 billion category includes restricted cash, undrawn facilities and contracted prepayments not yet received; it is not all cash available without conditions.
IREN also reported $20.3 billion in revenue backlog, about two-thirds with investment-grade counterparties or guarantors. Its definition includes remaining performance obligations, lease arrangements and other estimated future revenue under committed contracts, subject to delivery and service availability. Backlog is not revenue already earned.
That broader expansion plan differs from the company’s statement that its more than $4 billion run-rate target was fully contracted and funded. The latter covers a specific 0.3-gigawatt IT program and remains subject to funding conditions. It does not mean every future capacity target is financed.
Customer terms offer some counterweight. IREN reported estimated annual revenue per IT megawatt rising 11% on three-year contracts and 12% on five-year contracts signed in the September quarter versus the preceding quarter. Those comparisons use company assumptions about deployment power consumption and are weighted by annualized run-rate revenue. After quarter-end, it signed a five-year deal with customer prepayments covering 90% of associated estimated GPU capital expenditure. These disclosures show higher contracted revenue per megawatt and customer-backed funding, but do not establish realized margins or terms available across the industry.
Delivery and financing will determine the next commitments
The immediate unanswered question is which suppliers Microsoft will retain and how much additional capacity it will buy. A decision against renewal, a delivery-related termination and a customer-requested redesign have different implications; the reported review does not quantify a cut in external capacity purchases.
CoreWeave and Nebius are scheduled to report quarterly results the following week, according to the November 6 report. Delivery milestones, customer concentration and financing disclosures will help test whether suppliers can turn contracted demand into operating capacity. For IREN, the next evidence is clarity on operating versus upgrade-dependent capacity and customer acceptance against its revised timetable, alongside funding for expansion beyond its already-funded program. For Microsoft, the test is whether owned facilities and external purchases together supply Azure customers when promised.
Meta's strategy chief and former CFO is joining Fireworks AI ahead of a reported fundraising push targeting at least a $30 billion valuation. The inference company announced a $17.5 billion valuation in July, making the hire part of a much larger expansion than its older funding figures suggest.
Sand.ai says its $200 million Series A+ round lifts its valuation above $1 billion. Co-lead investor Wangsu plans inference and deployment cooperation, while its filing shows substantial losses behind the startup’s video AI ambitions.
An advocacy group estimates Interior Secretary Doug Burgum’s AI holdings at $18.1 million–$64.4 million as his department advances data centers on public land. His ethics agreement allows retained investments with recusals, while pending permits will determine both the application of those safeguards and the public’s return. Interior denies his decisions are driven by personal investments.
Physical Superintelligence says its AI system found a proof of a conjecture limiting quantum query advantages on random inputs. Public proof files enable scrutiny, but do not establish the private system’s cost or repeatability.
OpenAI is expanding its relationship with Hacktron, whose researchers previously reported an attack chain into employee accounts, to test third-party connections with greater access to cybersecurity models.
Google has acquired Crusoe’s southeast Wyoming data center site, where it plans a long-term $17 billion investment. Crusoe describes completed pre-development work; independent reporting points to customer setbacks. Power delivery and carbon capture remain key tests.
The team behind document-parsing service SoMark has released OXR-1.0 code and weights. A benchmark retest supports a strong result, but speed and accuracy were measured separately, and users above funding or revenue thresholds need a commercial license.
OpenAI is reportedly developing a Business Hub with daily mention and click charts, product updates and business agents. The unannounced feature could give verified businesses direct visibility into ChatGPT responses, without revealing prompts or competitor rankings.
Palantir and Nvidia have specified an integrated stack for running AI in customer-controlled environments. The architecture builds on existing software and infrastructure offerings, with published configurations spanning 32 to 640 GPUs, while pricing and workload economics remain unproven in the launch materials.
Democrats’ incoming congressional majorities put hearings and investigations within reach, but AI rules and crypto legislation still face bipartisan bargaining and a Republican president.
Microsoft reportedly declined to renew Oracle server rentals as older chips became less useful to it. The decision comes despite continuing capacity shortages, while Oracle says demand for earlier-generation GPUs remains strong.
An investigation puts Nvidia’s future commitments and guarantees at $530.5 billion. Its disclosures show rapidly expanding supplier commitments and phased support for customers whose infrastructure could generate further chip sales, with cash payments depending on timing, demand and contract terms.
TSMC is reportedly planning a possible move into 3D DRAM at Nvidia’s request. The foundry already helps integrate AI processors with memory, but a manufacturing expansion has no disclosed factory or production date.
Physicl has raised a $10 million seed round and introduced Studio, a beta app for generating targeted physical AI data. Its reported customer gains concern 3D reconstruction, leaving broader robot reliability and commercial economics to be demonstrated.
New cloud partners in Southeast Asia must enroll in Nvidia’s Fleet Intelligence system, according to a report, linking preferred chip access to monitoring while leaving data sharing and existing partners’ obligations unresolved.
Airbnb sold intellectual property to a private AI lab controlled by CEO Brian Chesky in a $7.1 million arrangement that also reimburses expenses. It obtained a license to certain lab technology and future investment rights, but disclosed no asset valuation or detailed license terms.
Chip shares recovered part of Thursday’s losses in early Friday trading. Fears about data-centre construction face a test against AI-chip supplier Nvidia’s strong reported sales, growth forecast and conditional infrastructure-financing plans.
SoftBank shares fell 10.8% as default-insurance costs reached a four-year high. Delayed listings and unfinished data centers complicate its AI investment outlook, while a $40 billion bridge facility has a March 2027 maturity and a repayment plan based on existing assets and other financing.
U.S. prosecutors allege a global fraud combined AI impersonation, sham confidential deals and cryptocurrency laundering. A blocked bank transfer shows why convincing an employee and moving the money are separate hurdles.
