31 October 2026 | Information cutoff: 06:44 GMT
Morning View
Friday’s selloff marked an important narrative change: investors did not dispute that AI demand is strong; they rejected the idea that strong demand automatically produces attractive equity returns. The Nasdaq fell 2.9%, its largest decline since June, while the semiconductor index lost 7.2%. Intel fell 28.6%, Advanced Micro Devices 10.9%, Dell Technologies 10.8% and Nvidia 4.2%. The common denominator was capital intensity, execution and the widening interval between infrastructure spending and free-cash-flow conversion.
The fundamental evidence beneath that repricing remains unusually strong. Tokyo Electron raised its profit outlook on sustained wafer-fabrication demand, Digital Realty signed record leases, Quanta Services ended the quarter with $63.7bn of backlog and utilities in Virginia and the Midwest disclosed tens of gigawatts of data-center demand. The cross-sector conclusion is not that the buildout is slowing; it is that power, financing, construction and semiconductor equipment are capturing more of the economics while the owners of compute absorb depreciation and balance-sheet risk.
The morning is therefore constructive for delivered infrastructure and security, but negative for leverage and undifferentiated AI exposure. Nvidia’s reported willingness to guarantee billions of dollars of Nscale debt shows that suppliers are increasingly financing the customers that create their demand. Binance’s $6.3bn hot-wallet compromise provides the opposite lesson: scale without compartmentalized operational control can create liabilities large enough to overwhelm years of fee income. The market is moving from rewarding announced capacity to pricing who funds it, who controls it and who ultimately bears the loss if assumptions fail.
1. Friday’s AI rout repriced returns on capital, not end demand
The Nasdaq fell 2.9% on October 30, its largest one-day decline since June 5, and ended 7.8% below its October 6 record. The semiconductor index dropped 7.2%; Intel lost 28.6%, Advanced Micro Devices 10.9%, Dell Technologies 10.8%, Hewlett Packard Enterprise 9.6% and Nvidia 4.2%. Amazon recovered much of its premarket loss but could not prevent the market from treating its free-cash-flow outflow and Intel’s weak outlook as evidence that AI demand can coexist with poor near-term shareholder economics. This is a material new development beyond the underlying earnings already covered: selling spread from company-specific misses into accelerators, servers, equipment and hyperscalers. The prior narrative allowed investors to capitalize capacity commitments years ahead of cash returns; Friday’s tape demanded a higher hurdle rate for depreciation, financing and utilization. Bulls can argue that indiscriminate liquidation creates opportunity because AWS growth, server demand and Asian equipment commentary remain strong. Bears can argue that the sector has not yet incorporated the full depreciation and funding burden of the 2026–2028 build. Nvidia, Advanced Micro Devices, Broadcom, Dell Technologies, Hewlett Packard Enterprise, Super Micro Computer, TSMC, ASML and memory suppliers are the principal transmission channel. The next test is whether Apple, Advanced Micro Devices and Palantir can beat elevated expectations next week without adding another leg to capital intensity. Sources: Reuters semiconductor selloff analysis; Reuters market close report.
2. Nvidia’s reported Nscale guarantee makes vendor financing part of the AI demand stack
Nvidia is in talks to provide billions of dollars of credit guarantees supporting debt for Nscale’s 460MW West Virginia data center for Anthropic, according to reporting by The Information carried by Reuters. Nvidia, Nscale and Anthropic did not confirm the talks, and no final guarantee has been disclosed. The proposed support directly addresses the financing gap highlighted in Nscale’s development plan: Anthropic has contracted to spend roughly $45bn on capacity at the first Monarch building, but contractual demand does not build the power and shell. This is genuinely new evidence following the financing concerns discussed yesterday. The bull case is that Nvidia’s balance sheet lowers funding costs, accelerates Vera Rubin deployments and converts conditional backlog into executable capacity. The bear case is circularity: a chip supplier guarantees the debt used to build a site that buys its chips, increasing contingent exposure if Anthropic, Nscale or residual GPU values disappoint. CoreWeave, Crusoe, Lambda, Nebius, private-credit lenders and data-center developers face the same question at smaller scale. The key unresolved points are the guarantee amount, whether it covers construction, equipment or residual value, the collateral and termination structure, and whether the obligation will be consolidated or disclosed as contingent exposure by Nvidia. Sources: Reuters; The Information.
3. Tokyo Electron’s outlook says equipment demand is accelerating even as AI equities derate
Tokyo Electron’s fiscal first-half update produced a sharp contrast with Friday’s public-market selloff. The company forecast an 87% increase in full-year operating profit to ¥1.17tn, above market expectations, said customer interest remains very strong and reported no evidence of slowing AI-related demand. Management also authorized a repurchase of up to ¥150bn. The importance is not simply another supplier beat: deposition, etch, cleaning and coater-developer demand sits several quarters upstream of accelerator shipments, making Tokyo Electron’s order and capacity signals a useful test of whether the hyperscaler pullback feared by equity markets is already reaching fabs. It is not. Bulls see sustained leading-edge logic, high-bandwidth memory and advanced-packaging intensity supporting multi-year equipment growth. Bears see a disconnect between customer forecasts and ultimate cloud returns, with export restrictions and memory-cycle volatility still capable of changing orders before delivery. Applied Materials, Lam Research, ASML, KLA, Screen Holdings, TSMC, Samsung Electronics and SK Hynix receive the clearest read-through. Investors should watch order intake, China mix, capacity additions and whether customers maintain 2027–2028 plans after the latest hyperscaler capex reaction. Sources: Tokyo Electron financial results; Reuters.
4. Digital Realty’s record leasing shows scarce powered capacity is retaining pricing power
Digital Realty reported third-quarter revenue of $1.9bn, up 21% yoy, and core funds from operations of $2.10 per share. The more important figures were record bookings: $855m of annualized GAAP base rent at 100% share, $627m at Digital Realty’s share, including a record $152m from the 0–1MW plus interconnection category. Signed but not commenced backlog reached $2.5bn at 100% share and $1.8bn at company share, while cash renewal spreads were 11.8%. Management raised 2026 core FFO guidance excluding promote income to $8.30–$8.35 from $8.15–$8.20. Friday’s investor discussion added useful context: the small-footprint and interconnection bookings exceeded the prior quarterly record by roughly 40%, showing that demand is not confined to a handful of hyperscale campuses. The bull case is that scarce power, network density and a large development pipeline sustain double-digit earnings growth and renewal pricing. The bear case is that backlog conversion requires heavy development capital and exposes the company to customer concentration and delayed energization. Equinix, Iron Mountain, American Tower’s CoreSite, QTS, CyrusOne and neocloud customers are exposed. The next catalysts are commencement timing, development yields, financing needs and whether renewal spreads remain positive as more capacity enters service. Sources: Digital Realty results; earnings-call report.
5. Dominion Energy’s 54.9GW pipeline exposes the gap between contracted interest and deliverable power
Dominion Energy disclosed on October 30 that data-center capacity in various stages of contracting in Virginia reached approximately 54.9GW at September’s end, up 6.3GW or 13% from December 2025; more than 12GW was covered by signed electric-service agreements. The scale matters because Virginia remains the largest concentrated test of whether regulated utilities can turn AI load requests into permitted generation and transmission without imposing unacceptable costs on residential customers. The headline pipeline is far larger than near-term deliverable capacity, so investors should not treat 54.9GW as a load forecast. Bulls see a multi-decade regulated-capital opportunity across generation, substations and transmission. Bears see speculative queue inflation, political resistance, interconnection delays and rising customer-affordability risk. NextEra Energy, American Electric Power, Duke Energy, Exelon, GE Vernova, Siemens Energy, Eaton, Quanta Services and data-center operators receive read-throughs. The decisive datapoints are how much of the pipeline advances from study to service agreement, the capital allocated per delivered gigawatt and regulatory treatment of costs if customer projects are cancelled. Sources: Dominion Energy results; earnings-call transcript.
6. Quanta Services converts grid scarcity into $63.7bn of backlog
Quanta Services reported third-quarter revenue of $12.01bn and adjusted earnings of $6.06 per share, with both measures growing strongly and total backlog reaching a record $63.7bn. Remaining performance obligations were $41.75bn, up 75.7% from year-end, while total backlog rose 44.9%; approximately 67% of RPO is expected to convert within 12 months. The shares rose as the broader AI complex sold off, an important relative signal that investors distinguish the companies building power and electrical infrastructure from the companies funding accelerators. The bull case is that transmission, generation interconnection and data-center electrical work are multi-year bottlenecks with limited skilled-labor capacity, supporting pricing and visibility. The bear case is that large, complex projects can produce working-capital volatility, execution losses and delayed revenue when permits or customer financing slip. MasTec, EMCOR, MYR Group, Primoris, Eaton and utility-equipment vendors receive favorable demand read-throughs. The next evidence is 2027 adjusted earnings growth, backlog quality following acquisitions and whether margins hold as project scale and complexity rise. Sources: Quanta Services results; Form 10-Q.
7. Alliant Energy’s next 1GW turns Midwestern power into a hyperscaler growth asset
Alliant Energy said on October 30 that it expects to sign contracts for roughly 1GW of additional data-center demand within six months, on top of 3.4GW already contracted with Google, Meta Platforms and Blackstone-backed QTS. The new load is expected to ramp between 2029 and 2031. Alliant raised its 2027–2030 capital plan by 19% to $15.9bn and expects peak system demand to increase more than 60% by 2031; management said earnings growth could exceed its 8% target if further capital is required. The prior debate centered on whether Midwestern utilities could attract enough committed load to justify generation and transmission investment. The disclosed pipeline suggests demand is arriving, but the lead time means financing and customer-credit protections are critical. Bulls see regulated rate-base growth with anchor counterparties and lower-cost land and power than coastal markets. Bears see residential rate pressure, construction inflation and the risk that AI customers reserve multiple sites before selecting one. WEC Energy, Ameren, Xcel Energy, American Electric Power, Meta Platforms, Alphabet and QTS are the closest read-throughs. Investors should monitor signed service agreements, minimum-demand protections, regulatory approvals and the financing mix for the expanded capital plan. Sources: Alliant Energy results; Reuters.
8. OneSpan’s 27% rally shows profitable identity software can still clear a skeptical tape
OneSpan reported third-quarter revenue of $63.9m, up 12%, with subscription revenue up 16% to $51.5m, annual recurring revenue up 9% to $195.5m and net retention of 105%. Adjusted EBITDA was $19.3m at a 30.1% margin. Management raised 2026 revenue guidance to $254m–$258m from $248m–$252m, ARR guidance to $197m–$201m and adjusted EBITDA guidance to $72m–$76m from $67m–$71m. The shares rose roughly 27% on Friday, a striking divergence from the broad technology selloff. The reaction reflects both low expectations and evidence that digital agreements and authentication can produce growth without sacrificing cash economics. Bulls see bank identity, transaction signing and fraud controls gaining importance as AI increases automated interactions. Bears note that ARR growth still trails subscription revenue, net retention is only modestly above 100% and one major banking expansion helped the quarter. DocuSign, Okta, Ping Identity, Entrust and smaller authentication vendors face the read-through. The next catalysts are ARR exit growth, customer concentration, retention and whether the raised margin outlook survives renewed product investment. Source: OneSpan results.
9. Binance’s $6.3bn hot-wallet breach resets the operational-security benchmark
Binance said its hot wallets were compromised at approximately 20:05 UTC on October 29, resulting in an estimated $6.3bn loss of token assets. The exchange said about $3bn had been frozen on-chain by October 30 and that customer losses would be covered; the FBI and U.S. Secret Service are coordinating with Binance and other partners. The investigation is ongoing, and neither Binance nor authorities have published a complete attack chain. This is the largest disclosed crypto-platform theft by a wide margin and matters beyond digital assets because it shows how authorization, device security and transaction policy can fail at extraordinary scale even without a confirmed database breach. Bulls can argue that rapid freezing, loss absorption and continued solvency demonstrate improved industry resilience. Bears will argue that a platform able to move billions from hot wallets lacked adequate segmentation, limits or independent approval controls. Coinbase, Robinhood, Circle, stablecoin issuers, custodians, Fireblocks, Chainalysis and privileged-access vendors are exposed to regulatory and spending read-throughs. The next evidence is the root-cause report, asset-recovery rate, withdrawal normalization, any reserve impairment and whether regulators impose new custody and hot-wallet limits. Sources: Binance incident update; Reuters.
10. Military AI is moving from pilot budgets to operational dependency
Reuters reported on October 30 that the Pentagon used the Maven system at unprecedented scale during the Iran conflict, integrating intelligence feeds with models and software from Palantir, Anthropic, OpenAI, Google, Microsoft and xAI. U.S. officials described demand from military analysts as insatiable and said classified-compute capacity was rapidly exhausted. The new investor information is not a single contract award but evidence that frontier models have moved into recurring operational workflows where speed, classified deployment and interoperability matter more than benchmark leadership. Bulls see a durable sovereign market with high switching costs, allied demand and expansion across intelligence, logistics and command workflows. Bears see procurement concentration, model-liability disputes and operational errors that could trigger tighter human-oversight requirements or vendor exclusions. Palantir is best positioned at the data and orchestration layer; Microsoft, Google, Amazon, Oracle, OpenAI, Anthropic and xAI compete for classified model and cloud consumption, while cybersecurity vendors benefit from secure identity and audit requirements. The next catalysts are Palantir’s earnings, new Maven task orders, allied deployments and clearer procurement rules governing model-provider control over military use. Sources: Reuters analysis; Palantir program references.
What to Watch
Apple reports on November 2, followed by Advanced Micro Devices and Palantir on November 3 and Qualcomm on November 4. The central issue is whether revenue and backlog can clear a newly higher return-on-capital hurdle without another step-up in spending. Investors should also watch any definitive disclosure from Nvidia or Nscale on the West Virginia guarantee, Binance’s root-cause and asset-recovery updates, and October U.S. employment data on November 6 because long-duration infrastructure valuations remain highly sensitive to the rate path.
Bottom line
Friday’s market reaction did not invalidate the AI build; it repriced who deserves to own the economics. Tokyo Electron, Digital Realty, Quanta Services, Dominion Energy and Alliant Energy show that semiconductor equipment, powered land, grid interconnection and skilled electrical capacity remain scarce. Nvidia’s reported Nscale guarantee shows that the chip layer may need to finance the infrastructure layer to keep demand converting into shipments.
The durable winners should be companies with contractual protection, pricing power and cash conversion before residual-value risk becomes visible. The vulnerable group is no longer defined by weak demand; it is defined by heavy funding needs, long asset lives and counterparties whose commitments can change before capacity is energized.