Friday, 13 November 2026 | Information cutoff: 06:48 GMT
Morning View
The dominant change is leadership risk at the center of the AI capital cycle. Jensen Huang’s immediate resignation removes the executive most closely identified with Nvidia’s product cadence, ecosystem financing and policy strategy just days before earnings. The overnight results nevertheless show that end demand has not broken: Cisco more than doubled its fiscal-year hyperscaler AI revenue expectation from last year’s level, Applied Materials guided above consensus, Cerebras raised its outlook and SMIC operated at 97.3% utilization.
The tension is now between demand visibility and execution risk. Nebius beat the quarter but delayed capacity, raised 2026 capital expenditure to about $27bn and outlined $40bn-$45bn for 2027; Microsoft suffered a global productivity outage; and HubSpot replaced its chief executive while promising a “refounding.” Investors are being asked to fund more infrastructure at the same time that founder dependence, power availability, component costs, service reliability and governance are becoming harder constraints.
The morning is positive for near-term technology revenue but mixed for positioning. Networking, wafer-fabrication equipment, alternative inference and Chinese mature-node capacity all have upward estimate support. The offset is a higher discount rate for businesses whose valuation depends on uninterrupted leadership or external financing. The market is likely to reward suppliers with near-term orders and cash conversion, while demanding more proof from asset-heavy platforms and long-duration AI optionality.
1. Nvidia loses its founder-CEO at the moment continuity carries the highest premium
Nvidia announced on 12 November that Jensen Huang resigned immediately as president, chief executive and director; chief financial officer Colette Kress is serving as interim chief executive while retaining her finance role, and the board has begun an internal and external search. The SEC filing says the board accepted the resignation on 11 November. This is a fundamental event rather than a reputational footnote: Huang shaped Nvidia’s accelerated-computing roadmap, developer ecosystem, customer financing and Washington-China positioning, and the company is due to report fiscal third-quarter results on 17 November. Nvidia said operations, customer commitments and the product roadmap should continue without interruption, but investors cannot yet assume that capital-allocation practices or supplier and customer relationships will be unchanged. The bull case is that Kress and the existing engineering team institutionalized execution well before the transition; the bear case is that the market had capitalized Huang-specific vision and influence into both Nvidia’s multiple and the financing of the wider AI buildout. Advanced Micro Devices, Broadcom, Taiwan Semiconductor Manufacturing Company, CoreWeave, Nebius and every Nvidia-dependent server and power supplier are exposed. The shares’ roughly $600bn weekly value loss reported by Reuters looks severe, but not obviously excessive until the board identifies a permanent leader and clarifies financing and China strategy. Sources: Nvidia; SEC filing; Reuters.
2. Cisco turns the AI networking supercycle into a $9bn revenue commitment
Cisco reported fiscal first-quarter revenue of $18.5bn, up 24% and above the approximately $18.14bn consensus, with adjusted earnings of $1.40 a share versus about $1.33 expected. More important, it lifted fiscal 2027 revenue guidance to $74.8bn-$75.8bn from $72.2bn-$73.4bn and raised expected AI-infrastructure revenue from hyperscalers to $9bn from $7.5bn, more than twice fiscal 2026. The print provides the cleanest overnight evidence that accelerator deployment is transmitting into Ethernet switching and optics rather than remaining concentrated in GPUs. Record adjusted operating margin of 37.8% shows operating leverage, but adjusted gross margin fell to 65.6% from 68.1% a year earlier as memory and component costs rose. Bulls can raise networking estimates and argue that Cisco’s installed base adds enterprise AI upside beyond hyperscalers. Bears should question how much of the 24% growth reflects acquisitions and whether lower-margin systems mix caps incremental earnings. Arista Networks, Broadcom, Marvell, Coherent, Celestica, Super Micro Computer and optical suppliers receive positive demand read-throughs, while Cisco’s security portfolio still must prove that infrastructure pull-through can lift recurring growth. Watch AI orders, component inflation and the conversion of the $9bn expectation into revenue without further gross-margin erosion. Sources: Cisco; Reuters.
3. Applied Materials confirms that AI spending is broadening into process intensity
Applied Materials delivered fiscal fourth-quarter revenue of $10.41bn and adjusted earnings of $4.16 a share, ahead of roughly $10.30bn and $4.06 expected. Revenue rose 53% year over year, adjusted gross margin expanded 250 basis points to 50.6%, operating cash flow reached $3.30bn and the company guided the January quarter to $11.1bn plus or minus $500m, above the approximately $10.84bn consensus, with adjusted earnings of $4.35 plus or minus $0.20. This matters because it converts the AI thesis from unit demand into higher process complexity across advanced logic, high-bandwidth memory and packaging. The prior concern was that export restrictions and customer concentration would dilute the equipment cycle; the print instead supports another leg of estimate revisions. Bulls will emphasize calendar-2026 Semiconductor Systems growth above 40%, margin expansion and fifteen announced EPIC research engagements. Bears should focus on the sustainability of exceptional growth, China exposure and the risk that customer capital intensity peaks before new fabs reach normalized utilization. ASML, Lam Research, KLA, Tokyo Electron, ASM International, memory makers and advanced-packaging suppliers receive positive read-throughs. Shares rose only modestly after hours, an appropriate response to a good print that was partly anticipated. The next tests are foundry and memory capital budgets and whether January-quarter strength persists into the second half. Sources: Applied Materials; earnings materials.
4. Nebius proves demand and simultaneously raises the financing hurdle
Nebius reported third-quarter revenue of $980.8m, above the roughly $907m consensus, and adjusted EBITDA of $465.8m versus about $352m expected; AI Cloud revenue rose 643% to $973.8m. The quality of the backward-looking beat was overshadowed by capacity timing and capital intensity. Management now expects 2026 revenue at the low end of its $3.0bn-$3.4bn range because third-party capacity slipped into the first half of 2027, raised 2026 capital expenditure to about $27bn from $20bn-$25bn and plans $40bn-$45bn in 2027. The company expects $12bn-$13bn of 2027 revenue and pre-tax breakeven by year-end, with customer prepayments and financing against investment-grade contracts covering 75%-85% of next year’s spending. Bulls can point to longer contracts, higher prices and prepayments that covered more than 100% of associated third-quarter capital expenditure. Bears will argue that a two-month delay to the more-than-$7bn run-rate target illustrates how construction and third-party delivery can overwhelm demand visibility. The 16% share decline reflects financing and execution risk rather than a demand miss. CoreWeave, Crusoe, Lambda, Nvidia, Vertiv and data-center lenders face the clearest read-through. Investors should monitor funding terms, energization dates and the bridge from contracted revenue to free cash flow. Sources: Nebius; Reuters.
5. Cerebras turns inference speed into revenue, but contracted power becomes the constraint
Cerebras reported third-quarter GAAP revenue of $210.8m and core revenue of $230.9m, with core cloud revenue rising 272% to $144.9m. It raised 2026 adjusted revenue guidance to $940m-$960m from $880m-$890m, disclosed two new hyperscaler wins and a Microsoft cloud agreement, and said it now has 700 megawatts of data-center capacity under contract. Remaining performance obligations of $26.1bn are largely tied to OpenAI, while management expects to triple revenue in 2027. The shares rose more than 17% after hours, a rational re-rating because profitable cloud growth and hyperscaler validation reduce the probability that wafer-scale systems remain a niche architecture. The bear case is concentration and delivery: contracted capacity is not energized capacity, and the company says data-center availability constrains deployments. Advanced Micro Devices, Groq, Nvidia, Oracle and inference-focused neoclouds are exposed to the possibility that customers increasingly mix architectures by workload. Power, cooling and networking suppliers benefit if Cerebras converts its backlog, while OpenAI’s commitments raise counterparty concentration. The next catalysts are the identity and scale of the hyperscaler wins, Microsoft deployment timing, capacity energization and evidence that gross margins survive rapid cloud expansion. Sources: Cerebras; Reuters.
6. SMIC’s 97% utilization shows AI scarcity extending beyond leading-edge accelerators
Semiconductor Manufacturing International Corporation reported third-quarter revenue of $3.19bn, up 33.9% and above the roughly $3.09bn consensus, while net profit of $428.4m exceeded about $418m expected. Gross margin reached 29.9%, well above the 26%-28% company range, and utilization rose 3.6 points sequentially to 97.3%. On 13 November, co-chief executive Zhao Haijun said AI-related semiconductor supply should remain tight through 2027, driven by data centers and private clouds, while some equipment lead times exceed 24 months. This changes the debate from whether China can reproduce the most advanced Nvidia accelerators to how AI demand absorbs mature-node power management, connectivity, control and memory capacity. Bulls can raise foundry utilization and domestic-substitution assumptions; bears should account for 30% expected depreciation growth, equipment restrictions and a fourth-quarter gross-margin guide returning to 26%-28%. Hua Hong Semiconductor, UMC, GlobalFoundries, Chinese equipment vendors and analog and power-chip suppliers are the principal read-throughs. The market may be underestimating the second-order uplift to mature nodes, but it should not equate near-full utilization with leading-edge competitiveness. Watch equipment deliveries, product mix and whether pricing holds as capacity expands. Sources: Reuters results; Reuters management comments.
7. Microsoft 365’s global outage reopens the concentration-risk discount
A Microsoft service degradation that began on 12 November continued into Asia-Pacific business hours on 13 November, affecting Outlook, Exchange Online, Teams, Copilot and related services. Downdetector recorded more than 130,000 US Outlook reports at the peak, while government agencies and businesses across the US, Australia, Japan and India reported disruption. Microsoft said traffic redistribution and server restarts produced only partial recovery and reported no evidence of malicious activity. The event is financially relevant because Microsoft 365’s moat rests on suite breadth and operational reliability; an outage that simultaneously disables communications and AI assistants demonstrates that bundling also concentrates failure risk. The bull case is that customers absorb a transient service-level credit rather than undertake expensive migration. The bear case is that repeated incidents support multi-cloud, sovereign-cloud and best-of-breed procurement, limiting Copilot attach and giving Google Workspace, Zoom, Slack, Okta and continuity vendors an opening. OpenAI and Anthropic were among the affected organizations, underscoring that frontier labs themselves depend on incumbent enterprise infrastructure. The immediate earnings impact should be limited, but the market may underprice cumulative procurement scrutiny. Watch the root-cause report, duration, service credits and any customer or regulator response. Sources: Microsoft service status; Reuters.
8. HubSpot’s founder return signals that incremental AI features are not enough
HubSpot said on 12 November that co-founder Brian Halligan returned as chief executive immediately, replacing Yamini Rangan, who also left the board; the company separately named Adobe finance executive Karen Burgess to succeed retiring chief financial officer Kate Bueker in February. HubSpot reaffirmed fourth-quarter and full-year guidance, so the change is not being presented as a near-term financial reset. The phrase that matters is Halligan’s plan to “refound” the company across product, go-to-market and culture. Under Rangan, customers increased from 135,000 to more than 311,000 and revenue grew from $1.3bn in 2021 to an expected $3.7bn in 2026, making the abrupt transition more about the next platform architecture than failed scaling. Bulls will see a founder capable of rebuilding the suite around agents before Salesforce, Microsoft and private AI-native CRM vendors disintermediate seats. Bears should treat simultaneous chief executive and finance transitions as execution risk and question whether AI compresses rather than expands per-seat software economics. Salesforce, Adobe, Intuit, ServiceNow, monday.com and private customer-support agents are exposed. Watch product and pricing changes, sales productivity, retention and whether “refounding” brings restructuring charges or lower near-term margins. Sources: HubSpot; SEC filing; Reuters.
9. A frontier-model injunction request converts safety rhetoric into operating risk
Four nonprofits sued Anthropic, Google, Meta Platforms, OpenAI and SpaceX’s AI division in San Francisco Superior Court on 12 November, seeking to halt development and deployment of more advanced models under California unfair-competition and public-nuisance theories. The plaintiffs cite developers’ own warnings about catastrophic risks and alleged examples of models escaping tests or autonomously attacking systems. An injunction of the breadth requested remains a low-probability outcome, and the allegations are not judicial findings. The investable change is that state-law litigation is becoming a potential constraint while federal policy remains comparatively permissive; Florida has separately asked a court to impose outside oversight on OpenAI model development. Bulls will argue that standing, causation and preemption hurdles make the case more headline risk than terminal-value risk. Bears will note that discovery could expose safety tests, delay releases and impose governance obligations even without a final injunction. Microsoft, Amazon and Oracle face indirect cloud-consumption exposure, while cybersecurity, evaluation and model-governance vendors could gain. Anthropic’s prospective listing makes disclosure of these contingencies especially important. Watch the defendants’ motions to dismiss, any temporary-relief hearing and whether courts consolidate or coordinate the growing cases. Sources: Public Citizen; Reuters.
10. Kalshi’s reported $65bn funding discussion puts a price on compute hedging
Reuters reported on 12 November that Kalshi is discussing a new financing at a $65bn valuation, only days after finalizing a $1bn round at $40bn, with proceeds intended partly for a compute-futures platform. The talks are not a completed financing and the valuation should be treated as reported, not confirmed. Even so, the product direction matters: GPU rental prices, power constraints and delivery timing have become material input risks for AI labs, neoclouds and enterprises, but the market lacks deep standardized hedging instruments. Bulls will argue that compute derivatives can reduce earnings volatility, improve project finance and turn Kalshi into market infrastructure rather than an event-betting venue. Bears should question regulatory approval, benchmark integrity, liquidity and a 63% valuation step-up before commercial adoption. CoreWeave, Nebius, Crusoe, Lambda, OpenAI and hyperscalers could use forward curves for procurement and financing; exchanges, commodity brokers and cloud marketplaces face potential competition. Nvidia and data-center owners would gain a transparent pricing signal but could lose some bilateral pricing power. The next catalysts are CFTC treatment, contract specifications, liquidity commitments and whether the financing closes near the reported terms. Sources: Reuters; Kalshi; Kalshi compute markets.
What to Watch
Over the next seven days, the highest-value catalysts are Nvidia’s 17 November earnings, where interim leadership must defend product cadence, customer financing and the fiscal 2028 outlook; Palo Alto Networks’ results on 18 November for platform consolidation and AI-security demand; the first market response to Cisco, Applied Materials and Cerebras in regular trading; Microsoft’s root-cause report and full recovery from the Microsoft 365 outage; and any court timetable in the frontier-model injunction case. Investors should also track whether Nebius provides firm energization dates and whether Kalshi files compute-futures specifications with regulators.
Bottom line
AI demand remains strong enough to lift estimates across networking, semiconductor equipment, alternative inference and Chinese foundry utilization. The valuation framework is changing faster than the demand curve: leadership continuity, capacity delivery, financing structure, service resilience and legal permission now determine which companies can convert the same secular demand into durable free cash flow. The strongest near-term setups are suppliers already monetizing orders; the weakest are businesses asking investors to capitalize distant capacity or founder-dependent strategy without equally visible execution.