Micron Falls 12% as IREN and Nebius Extend AI Selloff | Magica
Micron Drops 12% as AI Selloff Hits IREN and Nebius
Editorial Team
••📖8 min read
AI infrastructure stocks fell on November 20 amid cloud financing concerns and a reported NAND demand warning. Customer contracts support the buildout, but IREN’s filing shows the conditions and cash demands behind its expansion.
Micron’s intraday loss reached 12.1%; earlier reports showed Nebius down 7% and IREN down 5%.
Cloud providers face funding and delivery constraints even as they report substantial customer commitments.
Micron’s NAND exposure adds a memory-pricing risk distinct from the cloud builders’ financing problems.
Micron Technology, the memory-chip manufacturer, fell 12.1% to $606.93 by 12:43 p.m. ET on November 20, according to a midday market snapshot. AI cloud providers Nebius and IREN also extended their declines, with an earlier report putting their losses at 7% and 5%, respectively. These are intraday observations taken at different times, rather than closing prices.
The common exposure is the AI infrastructure buildout. The immediate business risks differ: cloud operators must finance and commission expensive computing capacity before collecting the expected revenue, while Micron faces a warning that part of the memory market could swing toward weaker pricing.
The retreat began before the regular session: CoreWeave, Nebius and IREN, all AI cloud providers, were down about 4% in premarket reporting. Anthropic, the Claude chatbot developer, had made its Nasdaq debut the previous day, but the listing failed to sustain the initial enthusiasm for AI infrastructure stocks.
Funding also remains conditional across the sector. The premarket report said AI chip supplier Nvidia’s neocloud commitments through 2032 had fallen to $32 billion from $36 billion in the previous quarter, and required cloud firms to secure financing for infrastructure purchases. Those commitments do not replace the operators’ own funding obligations.
November 20 intraday declines: Micron at 12:43 p.m. ET; Nebius and IREN from an earlier report published at 9:17 a.m. ET. These are not simultaneous observations or closing returns. Source: The Motley Fool.
OpenAI reports sharply lower reasoning-extraction rates in internal GPT-6.2 tests and backs sanctions against illicit foreign distillation. Its own research and rival Anthropic’s experience show why protecting hidden reasoning leaves capability transfer and customer costs unresolved.
Editorial Team
Nebius operates an AI cloud using its own data centers, rented colocation capacity and partnerships in which other parties fund hardware. Its corporate predecessor was Yandex N.V.; it adopted the Nebius name in August 2024 after selling its Russian businesses and related operations in other markets, according to its annual report.
IREN owns the land, power and data centers beneath its AI cloud, following a shift away from digital-asset mining. The comparison of their operating models shows why control matters: owning facilities gives operators more influence over delivery, but does not remove equipment, networking or financing constraints.
A report on Nebius’s November 12 results and earnings call said the company reported third-quarter revenue of $980.8 million, up 571% from a year earlier, but third-party colocation delays pushed its 2026 revenue outlook toward the bottom of its $3 billion to $3.4 billion range. That quarter also included $325 million of partnership revenue that its finance chief said would not recur in the same form, according to the same report. That is roughly one-third of reported quarterly revenue, calculated by dividing $325 million by $980.8 million.
The spending commitments are much larger: Nebius expects capital expenditures of $27 billion in 2026 and $40 billion to $45 billion in 2027. Convertible financing and share sales expose shareholders to dilution. The same report said management attributed the guidance change to timing rather than demand, and said customer prepayments on its third-quarter agreements covered more than all the associated capital expenditure. Management also reported more than $12 billion in contract value signed since the start of the third quarter. Those commitments support the demand case and funding for particular agreements; they do not establish that the entire expansion needs no new capital. Management is steering most new capacity toward company-owned sites to gain more control over delivery, according to the report.
Specialists also compete with established alternatives. In its annual report, Nebius identifies CoreWeave, Crusoe and Lambda as specialized competitors, alongside general cloud providers Amazon, Google, Microsoft and Oracle. The earlier market report put Oracle, the enterprise software company funding its AI expansion with software cash flows as well as debt and equity, down 4%. The retreat extended beyond the specialist cloud operators.
Nebius’s company-reported platform architecture spans data centers and servers through software and AI applications. Source: Nebius annual report. Source: Nebius annual report.
IREN’s funding claim has a defined boundary
IREN’s November 5 earnings release makes a similar distinction between demand, funding and timing. AI cloud revenue reached $200 million in the quarter ended September 30, compared with $71 million in the preceding quarter; total revenue was $228 million. AI cloud therefore supplied about 88% of quarterly revenue, calculated as $200 million divided by $228 million. But the sequential comparison includes a new revenue stream: cloud services revenue rose to $135 million from $71 million, while cloud lease revenue contributed $65 million versus none in the preceding quarter. Bitcoin mining revenue fell to $28 million from $67 million.
IREN reported $20.3 billion in revenue backlog, about two-thirds with investment-grade counterparties or guarantors. That company metric includes estimated future revenue under committed contracts and remains subject to delivery and service availability. Its Horizon 1 cluster had entered service for Microsoft.
The company said its target of more than $4 billion in operating annualized run-rate revenue was fully contracted and funded. But the timing moved from year-end to January 2027, potentially February, after a customer-requested design change and supplier delays. Annualized run-rate revenue annualizes contracted GPU pricing and includes storage and ancillary revenue; it is not annual revenue already earned. Commissioning, testing, utilization assumptions and customer acceptance affect recognized revenue, which IREN cautions could be materially lower.
The funding claim covers the 0.3-gigawatt computing-capacity program supporting that target. It includes undrawn committed financing and contracted customer prepayments not yet received, with conditions governing access to funds. Across its broader plans, IREN estimated $15 billion of remaining fiscal 2027 capital expenditure as of September 30. Cash, committed GPU financing and prepayments covered $9 billion; the remaining $6 billion—40% of the planned amount—was targeted financing and prepayments subject to market conditions.
Actual cash flows show why that distinction matters. During the September quarter, IREN paid $2.131 billion for property, plant and equipment excluding computer hardware, plus $2.693 billion for computer hardware: $4.824 billion combined. Operating cash inflow was $529 million, including a $645 million increase in deferred revenue. IREN reported a $263 million net loss. Its reconciled net-debt measure rose from $216 million at June 30 to $4.43 billion at September 30.
There is also evidence against a simple demand-collapse explanation. After quarter-end, IREN said it signed a five-year contract with estimated annual revenue of about $22 million per megawatt of IT capacity and customer prepayments covering 90% of associated estimated GPU capital expenditure. A separate short-term contract carried estimated annual revenue of $40 million per IT-megawatt. Different contract lengths prevent a direct profitability comparison, and those terms do not guarantee comparable prepayments on future contracts.
Company-reported funding plan for IREN’s remaining FY27 capital expenditure. The $9 billion includes restricted cash, undrawn financing and contracted prepayments not yet received, with funding conditions; targeted sources remain subject to market conditions. Source: IREN SEC earnings release.
Micron faces a different demand test
Micron supplies memory and storage used in AI systems, including high-bandwidth memory, DRAM and solid-state drives. An early trading report citing TechNode described progress at its roughly $7 billion Singapore high-bandwidth-memory packaging facility: the first process tool had been installed, with initial output expected in early 2027. That expansion continued even as the shares fell. Packaging output is a different milestone from wafer production: a separate Singapore wafer fabrication project was expected to begin output in the second half of 2028. The same account said Micron shares had gained more than 140% in 2026 before Friday’s regular session, putting the daily decline in the context of a much larger prior advance.
The more specific negative signal concerned NAND flash, the memory used for storage. A midday analysis relayed research firm TrendForce’s warning, reported through StreetInsider, that major Chinese cloud providers were cutting enterprise SSD orders. It described NAND spot prices as already down 20% as distributors cleared inventory, with further price declines forecast for 2027. This is a reported market forecast, not a decline in Micron’s realized selling prices or a company revenue warning.
The account gives no clear time window for the 20% spot-price decline. Nor does it establish an equivalent deterioration in DRAM or high-bandwidth memory, other products through which Micron supplies AI infrastructure.
The next test is delivery and realized pricing
The next useful evidence is operational. For Nebius, it is whether new sites arrive on schedule and customer prepayments continue to support construction. For IREN, it is delivery of the delayed contracted capacity alongside financing for the wider buildout. A separate account of an Evercore ISI analyst’s comments described capacity and capital as immediate constraints, while identifying the durability of demand as the longer-term financing question.
Micron is scheduled to report earnings on December 17, according to the early trading report. Its pricing and demand disclosures will help test whether NAND weakness is confined to particular markets or spreading through its product mix. Friday’s share losses alone cannot answer that question.
Alberta proposes 600-metre setbacks for large data centres and expanded utility oversight, but exemptions and rules still to be written will determine how much protection communities receive.
President Donald Trump backed Commerce Secretary Howard Lutnick as lawmakers scrutinize alleged delays in restricting a suspected route for AI chips to China. A new congressional inquiry also questions Nvidia’s influence on an existing investigation.
Amazon’s month-old Echo Time is $50 off for Black Friday. Its touchscreen, home hub and optional sleep-aware alarm cost $90 more than the reported Echo Spot price, while Alexa+ access and evidence of better waking remain key buying questions.
Merck will pair its experimental data with Isomorphic Labs’ AI to seek small-molecule cancer drug candidates. The deal follows Isomorphic’s $5 billion financing, but its payments, target and development timetable remain undisclosed.
Thinking Machines Lab released a roughly 20GB open-weights model that reuses its middle layers for reasoning. It brings the Inkling family to local hardware, with competitive coding results but workload-dependent costs and preliminary evidence on oversight.
Apple’s first HomePod mini update since 2020 adds Mac audio and four-speaker surround, but Siri AI remains unavailable, core home-hub features carry over, and reviews disagree on playback reliability.
Treasury Secretary Scott Bessent says an unnamed Anthropic founder will announce a large gift to children's investment accounts within weeks, but the amount, recipients and connection to existing philanthropic pledges remain undisclosed.
Google says air cooling will limit water use at its Visakhapatnam AI hub, but environmental proceedings and residents’ accounts raise broader questions about land, electricity, public incentives and the evidence behind promised benefits.
Nvidia slipped behind Alphabet to fourth in U.S. market value during November 20 trading. The reversal follows a steep selloff despite strong reported revenue growth, with a permanent CEO still to be chosen.
Google will close the Nest app on January 25, 2027 and retire Assistant on home speakers and displays in August. Device transfers are automatic, but continued voice access requires accepting Gemini; basic assistance remains free.
Amazon says Cairn agents exploited a MONAI Label zero-day to take over a university server. Other attacks failed, and maintainers list no patched package release.
Federal Reserve Governor Lisa D. Cook sees early AI pressure on coding, translation and young workers, but says a solid overall labor market and persistent inflation still call for tighter monetary policy.
President Donald Trump threatened a tax on U.S. companies building data centers abroad without specifying how it would work. The proposal raises questions for an AI export program that has received 78 applications and for domestic construction constrained by power and community opposition.
Donald Trump said only older chips should be sold to China while discussing a broader deal with Beijing. His remarks leave unresolved which products would qualify as Washington tightens access to American AI hardware and overseas manufacturing.
Treasury Secretary Scott Bessent expects no change to the US policy permitting limited H200 sales to China despite sparse orders. Nvidia’s earlier filing shows Chinese restrictions and tariff costs also obstruct sales, while the company warns that exclusion helps competitors.
Jensen and Lori Huang’s foundation has committed $350 million to CoreWeave computing for research recipients. The purchase links charitable access to Nvidia’s commercial ecosystem, while pricing, allocation and supplier-selection terms remain unclear.
The mathematicians claim finite-time blow-up without external forcing, a significant advance if verified. Their repository omits the numerical certificate and its checking computation, and says the patched Lean package was not rebuilt.
The Pentagon treated Anthropic’s military-use safeguards as a supply chain threat. Later court rulings diverged under different statutes, while the dispute exposed how difficult it was to replace Claude in classified systems.
A new investigation describes decades of alleged humiliation under Nvidia co-founder and former CEO Jensen Huang. He denies a former HR chief’s claims, while employee support and stock rewards complicate the account of his management.