This node covers the demand side of the power-scarcity chain: cloud platforms, AI compute providers, enterprise software clouds, and data-center REITs that request large blocks of electricity for servers, cooling, networking, storage, and customer deployments. The paper-led rerank puts energized or contracted AI capacity first: CRWV, MSFT, ORCL, AMZN, and GOOGL are core; NBIS and META are option-tier; DLR and EQIX stay as landlord watch rows until power, lease, AFFO, and funding evidence confirms per-share conversion.
What the stack is: demand originators are the customers and operators that turn AI, cloud, colocation, enterprise software, and data-center leasing plans into physical power load.
What it does: it converts compute demand into server halls, GPU clusters, network equipment, cooling, backup power, data-center leases, customer contracts, cloud backlog, active power, and utility connection requests.
Main operating pieces: cloud regions, AI training and inference clusters, GPU clouds, data-center campuses, colocation halls, interconnection fabric, power purchase agreements, utility service requests, leases, customer prepayments, and remaining performance obligations.
Where it sits: demand begins inside hyperscaler and cloud-capacity planning, then shows up at data-center sites, colocation campuses, utility interconnection queues, powered-shell developments, and customer contracts.
How the parent theme uses it: downstream utilities, generators, equipment suppliers, grid contractors, cooling vendors, campus developers, and fuel suppliers need real load before their own revenue, rate recovery, backlog, or capacity economics can be underwritten.
Terms used later: RPO means contracted revenue not yet recognized; backlog means contracted revenue or capacity still to be delivered; active power means energized capacity already available; contracted power means future capacity rights; colocation means third-party data-center space and interconnection sold to customers.
Report boundary: this node is a tactical report layer. It ranks the current value-chain basket, setup labels, confirmation triggers, invalidation levels, and chart provenance. Durable research, claim IDs, raw source registries, and sector thesis maintenance stay in the linked knowledge pages. Price and volume context comes from read-only discovery daily_ohlc through 2026-07-02; static setup thresholds remain from the 2026-05-22 package.
Current Setup
Demand proof gateLoad matters when contracts, capacity, and power access line up.
Hyperscaler capex alone is not enough. The strongest read combines cloud or AI revenue growth, RPO/backlog, signed leases or committed capacity, utility power access, and cash conversion after infrastructure spending.
Positive proofThe rerank separates direct load proof from landlord corroboration.
CRWV, MSFT, ORCL, AMZN, and GOOGL lead; NBIS and META are option rows; DLR and EQIX remain watch rows until lane depth improves.
Conversion gateContracts must become revenue and usable megawatts.
Watch cloud RPO, backlog, power availability, lease starts, and revenue after capex.
Primary constraintFunding and power can absorb demand value.
Debt, lease obligations, equity issuance, capex, power costs, and customer concentration decide per-share economics.
Demand sourceAI, cloud, colocation
CommitmentRPO, backlog, leases
Physical gatePower + interconnection
Proof pointRevenue, margin, FCF
New power demand matters only when planning-case load becomes signed customer commitments, funded interconnection work, leases, PPAs, power-reservation terms, energization schedules, or capex that converts into cloud, AI, colocation, or industrial revenue. This page keeps the upstream demand proof separate from the downstream monetization pages: utilities need approved recovery, merchant power needs capacity and cash conversion, equipment suppliers need backlog shipments, and campus operators need deliverable power.
The local knowledge lanes show large cloud and AI spending plans at MSFT, AMZN, GOOGL, and META; direct AI infrastructure and power-capacity evidence at CRWV; and demand-route evidence for ORCL, NBIS, EQIX, and DLR that still needs deeper company-level source work. These names confirm the theme when RPO, backlog, data-center capex, leases, customer contracts, active or contracted power, cloud or AI revenue, and cash conversion move together.
The setup weakens if demand remains a forecast, LOI, non-binding queue entry, broad capex headline, or gross booking without customer credit, power access, timing, margin, and economic detail. CRWV and NBIS have direct capacity language but carry funding, customer, and execution risk; ORCL, EQIX, and DLR still need local lane depth before the page can make stronger company-level claims.
Static setup labels and thresholds in the sidecar package were built from older discovery rows. Read-only discovery now shows daily OHLC through 2026-07-02 for all nine active tickers, so the static thresholds are stale and should be refreshed before use as current trading evidence. Fundamental claims route to the knowledge theme page, sector lanes, linked security lanes, or bounded official-source checks.
Basket
This basket is reranked as a demand-proof monitor, not a clean-beneficiary list. The papers make signed or energizable load the key input, so direct active/contracted MW, connected-power targets, cloud RPO, AI backlog, leases, customer commitments, capex funding, and FCF after capex move ahead of generic hyperscaler size. CRWV ranks first because the local lane has active and contracted power. MSFT, ORCL, AMZN, and GOOGL follow for durable cloud or RPO evidence. NBIS is added as an option because it has direct connected and contracted power targets. META validates load but has less direct MW/RPO disclosure. DLR and EQIX are watch corroborators rather than demand originators.
CoreWeave is the most direct AI compute and power-capacity proxy, but debt, leases, customers, and margin keep it below the mega-cap originators.
Market cap$60.3B
Next earningsNot confirmed
Latest qtr revenue$2.078B
Role in stack
CoreWeave sells committed and on-demand AI cloud capacity. It turns theme pressure into revenue when contracted GPUs, networking, data-center fit-out, active power, and customer commitments become available capacity at acceptable margin after debt and leases.
Revenue mix
The local lane treats reported revenue as CoreWeave Cloud Platform revenue because public disclosure does not provide a richer segment split. Q1 revenue was 98% tied to customer commitments.
Latest qtr revenue
Q1 2026 revenue was $2.078B. The CRWV lane cites $98.8B RPO, $99.4B revenue backlog, active power above 1 GW, contracted power above 3.5 GW, Q1 capex of $7.695B, $24.859B debt, and about $10.050B operating lease liabilities.
Azure, commercial RPO, AI capacity, and Microsoft Cloud revenue make Microsoft the cleanest source-backed demand originator.
Market cap$3.08T
Next earningsNot confirmed
Latest qtr revenue$82.886B
Role in stack
Microsoft buys and builds data-center, power, networking, and AI capacity for Azure, Microsoft 365, GitHub, security, and first-party AI workloads. Theme pressure becomes economics when revenue-ready Azure capacity, commercial RPO conversion, Copilot adoption, and cloud margin support the capex plan.
Revenue mix
Q3 FY2026 revenue was concentrated in Productivity and Business Processes at $35.013B and Intelligent Cloud at $34.681B, with More Personal Computing at $13.192B. Azure and other cloud services grew 40%.
Latest qtr revenue
Q3 FY2026 revenue was $82.886B. The security lane cites Microsoft Cloud revenue of $54.5B, commercial RPO of $627B, and nine-month property and equipment additions of $80.146B from the Q3 FY2026 10-Q and earnings-call sources.
Oracle now has local OCI, RPO, and data-center lease evidence, but funding quality and conversion timing remain the gating issues.
Market cap$563.6B
Next earningsNot confirmed
Latest qtr revenue$17.190B
Role in stack
Oracle creates demand through OCI regions, AI infrastructure, database cloud, multicloud deployments, data-center leases, and power obligations. The gate is whether RPO and OCI growth convert into revenue without worse FCF, debt, lease, or ATM-dilution evidence.
Revenue mix
Q3 FY2026 revenue was mostly cloud and software at about 87.45% of total revenue. Cloud infrastructure revenue was $4.888B, cloud applications revenue was $4.026B, hardware was $714M, and services was $1.443B.
Latest qtr revenue
Q3 FY2026 total revenue was $17.190B. The ORCL lane cites OCI growth of 84%, RPO of $552.6B, FY2026 capex guidance of $50B, negative $24.736B trailing-four-quarter FCF, and $261B of additional data-center lease commitments.
AWS backlog, AI capex, and data-center buildout give Amazon direct load evidence, with retail cash flow as the funding bridge.
Market cap$2.93T
Next earningsNot confirmed
Latest qtr revenue$181.519B
Role in stack
Amazon creates power load through AWS regions, generative AI infrastructure, custom silicon, data-center capex, and customer commitments. The conversion gate is whether AWS growth, utilization, margin, and retail operating cash flow justify the infrastructure and power spend.
Revenue mix
Q1 2026 revenue was split across North America at $104.143B, International at $39.789B, and AWS at $37.587B. AWS was about one-fifth of sales but the majority of segment operating income.
Latest qtr revenue
Q1 2026 consolidated net sales were $181.519B. The AMZN lane cites AWS operating income of $14.161B, TTM operating cash flow of $148.531B, and TTM property and equipment purchases of $151.003B from Q1 2026 filing and transcript sources.
Google Cloud, TPU capacity, Search cash flow, and AI infrastructure financing make Alphabet a major demand and funding test.
Market cap$4.86T
Next earningsNot confirmed
Latest qtr revenue$109.896B
Role in stack
Alphabet turns AI and cloud demand into load through Google Cloud data centers, TPU and Gemini infrastructure, Search and YouTube funding, and enterprise AI capacity. The gate is whether Cloud, Search, and TPU revenue absorb capex, leases, debt, and issuance.
Revenue mix
Q1 2026 revenue included Google Services at $89.637B, Search and other at $60.399B inside Services, Google Cloud at $20.028B, and Other Bets at $411M.
Latest qtr revenue
Q1 2026 revenue was $109.896B. The GOOGL lane cites Google Cloud operating income of $6.598B, Cloud margin of 32.9%, Q1 capex of $35.674B, and about $10.116B of FCF after capex from Q1 filing and release sources.
Option demand-origin and capacity proof row. Nebius has more direct connected and contracted power disclosure than most hyperscalers, but first-occurrence evidence, capex, debt, ATM funding, and execution risk keep it below the larger self-funding clouds.
Revenue mix
The NBIS lane routes demand through AI Cloud capacity, Meta-related orders, contracted power, connected-power targets, customer advances, and infrastructure funding.
Proof burden
Confirm with connected MW ramping toward the 2026 target, Meta orders starting on schedule, AI Cloud ARR becoming revenue, and financing that does not overwhelm common equity. Weaken if power delivery slips, customer concentration rises, or debt/ATM dilution absorbs the demand signal.
Family of Apps cash flow funds a large AI data-center cycle, making Meta a demand confirmer with a self-funding test.
Market cap$1.55T
Next earningsNot confirmed
Latest qtr revenue$56.311B
Role in stack
Meta creates demand through AI training and inference infrastructure, data centers, finance leases, cloud capacity, and long-duration commitments. The conversion gate is whether advertising and AI product evidence fund the capex without confirmed dilutive financing.
Revenue mix
Q1 2026 revenue was dominated by Family of Apps at $55.91B, while Reality Labs produced $402M and remained loss-making. Advertising revenue was $55.02B.
Latest qtr revenue
Q1 2026 revenue was $56.311B. The META lane cites 41% consolidated operating margin, FoA operating income of $26.900B, FY2026 capex guidance of $125B-$145B, and $237.67B of non-cancelable commitments from Q1 2026 sources.
Digital Realty offers scarce powered-capacity exposure, but the node needs DLR-share backlog conversion and funding proof.
Market cap$68.6B
Next earningsNot confirmed
Latest qtr revenue$1.635B
Role in stack
Digital Realty owns and develops data-center capacity for hyperscale, cloud, AI, and interconnection customers. Theme pressure becomes economics when signed bookings and powered capacity commence as DLR-share rent, NOI, Core FFO, and AFFO after capex, debt, and equity issuance.
Revenue mix
The DLR lane frames the mix by region and capacity: Americas had $2.494B of DLR-share annualized rent, EMEA $1.631B, and Asia-Pacific $435M at 3/31/26. Northern Virginia is a major high-occupancy market.
Latest qtr revenue
Q1 2026 total operating revenue was $1.635B. The DLR lane cites Core FFO/share of $2.04, signed bookings of $423M at DLR share, signed-not-commenced backlog of $1.0B at DLR share, and raised 2026 Core FFO/share guidance of $8.00-$8.10.
Equinix converts AI and hybrid-cloud demand through recurring colocation and interconnection, with AFFO and power delivery as proof gates.
Market cap$105.7B
Next earningsNot confirmed
Latest qtr revenue$2.444B
Role in stack
Equinix sells data-center colocation, interconnection, and managed infrastructure through IBX campuses and xScale ventures. Theme pressure becomes economics when power-backed cabinets, AI bookings, utility costs, and customer deployments convert into recurring revenue and AFFO/share.
Revenue mix
Q1 2026 revenue was 95% recurring, with Americas at $1.091B, EMEA at $827M, and Asia-Pacific at $526M. The company also disclosed ecosystem density above 2,000 network service providers.
Latest qtr revenue
Q1 2026 revenue was $2.444B. The EQIX lane cites Q1 adjusted EBITDA of $1.245B, raised FY2026 revenue and AFFO guidance, 51,900 sellable cabinets under construction, and Q1 capex of $1.256B.
Market caps use read-only discovery instruments.market_cap values queried on 2026-06-13; the table has no source timestamp in last_updated_utc. Next earnings are labeled Not confirmed because bounded checks of Nasdaq earnings pages on 2026-06-13 returned unavailable earnings-date data; local knowledge lanes still indicate late-July windows for several names and June/FY calendar windows for ORCL, but this page does not promote those to confirmed card dates. Latest-quarter revenue metrics use each linked security lane's latest filed quarter and source registry.
What Confirms Or Weakens
AreaWhat confirmsWhat weakens or invalidatesWatch next
01Node thesisDemand becomes real load
What confirms
Cloud RPO, AI backlog, contracted compute, signed leases, customer credit support, power reservations, and active or contracted MW connect to named data-center or cloud-capacity projects.
What weakens or invalidates
Demand remains planning-case load, LOIs, non-binding queue entries, capex headlines, or bookings without contract term, customer credit, cancellation language, power access, or timing.
Watch next
MSFT RPO and Azure capacity
AMZN AWS commitments
GOOGL Cloud and TPU timing
META capex funding
02Economics after capexRevenue and cash conversion
What confirms
Cloud, AI, ad, OCI, colocation, and data-center rent growth support operating income, AFFO/Core FFO, margin, and free cash flow after data-center capex, leases, depreciation, interest, and power costs.
What weakens or invalidates
Capex, leases, interest, depreciation, power costs, customer concentration, or equity issuance rise faster than revenue, utilization, operating income, AFFO/Core FFO, or FCF.
Watch next
Hyperscaler FCF after capex
CRWV adjusted operating margin
ORCL FCF and lease commitments
EQIX/DLR per-share conversion
03Customer and contract qualityCredit, duration, concentration
What confirms
Backlog and bookings have identified customers, duration, credit support, committed usage, cancellation protection, service-level economics, and conversion windows that match build schedules.
What weakens or invalidates
Revenue depends on a few customers, non-binding awards, opaque RPO, customer-linked financing, availability credits, delayed deployments, or receivable concentration without adequate disclosure.
Watch next
CRWV top-customer shares
ORCL RPO recognition
DLR signed-not-commenced backlog
EQIX AI deal mix
04Funding and balance sheetCapital remains available
What confirms
Debt, leases, customer prepayments, vendor financing, equity issuance, and joint ventures fund capacity at terms that still leave margin, FCF, AFFO, or Core FFO per share for equity holders.
What weakens or invalidates
Financing cost, preferred claims, ATM use, debt maturities, customer prepayments, or lease obligations grow without matching revenue conversion and return evidence.
Watch next
GOOGL and META issuance commentary
CRWV DDTL/debt
ORCL gross borrowings
DLR ATM/debt coupon
05Power and supply constraintPhysical capacity is deliverable
What confirms
Power availability, utility agreements, PPAs, interconnection funding, active MW, contracted MW, cabinet delivery, GPU supply, and construction schedules allow capacity to become usable revenue.
What weakens or invalidates
Grid delays, equipment bottlenecks, GPU or memory scarcity, utility-cost inflation, permitting delays, project slips, or power procurement gaps leave contracted demand under-energized.
Watch next
CRWV active and contracted power
EQIX power-backed cabinets
DLR under-construction GW
Hyperscaler capacity commentary
06Downstream conversionDemand reaches suppliers
What confirms
Demand evidence appears in utility load agreements, merchant PPAs or capacity markets, electrical equipment backlog, grid-service work, powered-campus leases, cooling orders, backup-power orders, or fuel demand.
What weakens or invalidates
Downstream pages show delayed recovery, weak PPA economics, equipment cancellations, grid interconnection slippage, campus delivery problems, cooling order gaps, or fuel-cost pressure.
Watch next
Utilities and grid recovery
Merchant power
Electrical equipment
Cooling and backup
07Stale conditionStatic package must refresh
What confirms
The page remains usable only after source lanes, earnings releases, filings, daily OHLC coverage, setup thresholds, and API chart metadata are refreshed to the same as-of date.
What weakens or invalidates
Any new trading session, earnings release, filing, financing, power agreement, lease, data-center update, rating action, or source-lane refresh arrives before this static node package is rebuilt.
Read-only DuckDB checks used ../discovery/data/discovery.duckdb, table daily_ohlc, grouped by ticker for CRWV, MSFT, ORCL, AMZN, GOOGL, NBIS, META, DLR, and EQIX.
The assigned static node package still records chart artifacts through 2026-05-22 with weekly aggregation from daily rows: first open, maximum high, minimum low, final close, and summed volume.
Read-only discovery now shows daily_ohlc coverage through 2026-07-02 for all nine active tickers. Static setup labels and threshold values from the 2026-05-22 sidecar are stale relative to current local data.
The selected-security right rail calls the report API from page chart triggers with data-chart-ticker. Expected API route shape is /api/securities/TICKER/chart?frequency=weekly&window=3y&as_of=latest.
Right-rail chart assumptions: weekly OHLC sourced from daily_ohlc, three-year visible horizon where data exists, 20-week EMA, 100-week EMA where enough history exists, and volume subgraph. CRWV public OHLC begins 2025-03-28 and lacks a full three-year/100-week history.
Market caps use read-only instruments.market_cap queried on 2026-07-04; last_updated_utc is null for those instrument rows.
Earnings-Date Checks
Bounded web checks on 2026-06-13 opened Nasdaq earnings pages for MSFT, AMZN, GOOGL, and META; the pages loaded but reported earnings-date data as unavailable. The page therefore uses Not confirmed rather than an estimated date.
Additional web searches for CRWV, ORCL, EQIX, and DLR did not produce a company-confirmed next earnings date suitable for the card metric. Local knowledge lanes include expected evidence windows, but those are not promoted to confirmed dates.
Handoff for review/coordinator: refresh next-earnings dates from official IR event pages, company releases, exchange/company calendars, Nasdaq/NYSE pages, or another credible calendar before using this node as a live trading checklist.
Latest-Quarter Revenue Sources
MSFT: Q3 FY2026 revenue of $82.886B from the MSFT lane, citing the Q3 FY2026 10-Q and earnings-call source registry.
AMZN: Q1 2026 consolidated net sales of $181.519B from the AMZN lane, citing the Q1 2026 10-Q and transcript source registry.
GOOGL: Q1 2026 revenue of $109.896B from the GOOGL lane, citing the Q1 2026 10-Q and earnings-release source registry.
META: Q1 2026 revenue of $56.311B from the META lane, citing the Q1 2026 10-Q and earnings-release source registry.
CRWV: Q1 2026 revenue of $2.078B from the CRWV lane, citing the Q1 2026 10-Q and transcript source registry.
ORCL: Q3 FY2026 total revenue of $17.190B from the ORCL lane, citing the Q3 FY2026 10-Q and Q3 release source registry.
EQIX: Q1 2026 revenue of $2.444B from the EQIX lane, citing the Q1 2026 10-Q and Q1 release source registry.
DLR: Q1 2026 total operating revenue of $1.635B from the DLR lane, citing the Q1 2026 10-Q, earnings supplement, and presentation source registry.
Generator Drift
scripts/build_power_scarcity_nodes.py owns this node family but still emits the old table-based basket, visible setup paragraph stack, body setup table, and body chart section. Do not rerun it until it can emit the current card-based format and paper-led ranks.
The parent page, parent metadata, node manifest, static HTML, and sidecar JSON now carry the accepted paper-led ranks. Static setup thresholds remain stale and should be refreshed separately before use as live trading levels.
Known Gaps
Discovery supports price, volume, weekly OHLC, EMA, and coverage facts only. It does not prove power-load causality without source-routed commitment, lease, filing, funding, power, or revenue evidence.
The page intentionally avoids relative-return or causal attribution claims versus SPY, QQQ, XLK, VGT, VNQ, or rates proxies because benchmark checks were outside this implementation worker's scope.
Next-earnings dates remain missing for all nine card metrics until a review pass finds credible company-confirmed or clearly estimated dates.
Static setup thresholds in the sidecar package remain stale versus read-only discovery coverage through 2026-07-02. Refresh chart/setup metadata before relying on those thresholds.
ORCL, DLR, and EQIX now have local lanes, but the node still requires power, lease, AFFO, funding, and cash-conversion evidence before raising the landlord rows.