This page covers the assembly layer that turns accelerators, central processing units, memory, storage, networking cards, power shelves, cooling interfaces, firmware, test plans, and customer bills of materials into deployable AI servers and rack-scale systems. Original equipment manufacturers sell branded systems, electronics manufacturing services providers build and test customer programs, and systems integrators combine trays, racks, cabling, firmware, supply-chain scheduling, and acceptance work so a data-center buyer can install usable compute capacity. The node gets paid when AI server orders, rack programs, and manufacturing backlog ship into accepted revenue, segment margin, operating cash flow, free cash flow, and cleaner working capital. DELL ranks first because AI-server revenue, AI orders, backlog, and cash flow are the clearest scale evidence. HPE ranks second after Q2 FY2026 Cloud & AI, Networking, backlog, and free-cash-flow evidence improved. CLS, JBL, and SANM are manufacturing routes that need program conversion and cash discipline. SMCI has the most direct rack-scale exposure, but inventory, financing, controls, customer concentration, and export risk keep it option-ranked.
What the stack is: server original equipment manufacturers, rack-scale integrators, electronics manufacturing services providers, and hardware-platform specialists that assemble AI compute systems for hyperscalers, AI-cloud providers, enterprises, governments, and infrastructure partners.
What it does in the parent chain: this layer converts chips and components from the accelerator, memory, storage, networking, and power/cooling nodes into shipped servers, racks, integrated manufacturing programs, field-ready configurations, and accepted customer deployments.
Main physical, software, and service pieces: server trays, motherboards, accelerator modules, CPUs, DRAM, storage drives, network interface cards, optical or copper links, power shelves, rack manifolds, liquid-cooling connections, firmware images, test scripts, supply-chain planning, factory labor, and customer acceptance work.
Where it sits: physically at OEM factories, outsourced EMS plants, integration labs, rack-build sites, customer staging areas, and the data-center loading dock before the system moves into a hall, row, rack, or cluster.
How AI infrastructure uses it: hyperscalers and AI-cloud buyers fund the cluster, component suppliers ship inputs, the server or EMS company builds and tests the rack, the customer accepts it, and revenue becomes investable only if margin, receivables, inventory, capex, debt, and control evidence hold together.
Terms used later: EMS means outsourced electronics manufacturing services; HPS means Celestica's hardware-platform solutions route; Cloud & AI is HPE's segment that includes servers and AI systems; backlog is customer demand not yet recognized as revenue; free cash flow is cash left after operating cash flow and capital spending.
Report boundary: this node is a tactical report layer. It ranks the current servers, racks, EMS, and systems integration basket, operating evidence, confirmation checks, and chart provenance. Durable research, claim IDs, raw source registries, and sector thesis maintenance stay in the linked knowledge pages. The right-rail chart API reads local discovery daily_ohlc; stored rows were checked through 2026-06-12, while static setup thresholds in metadata still use completed weekly bars through 2026-06-05 and should be refreshed before use as current trading evidence.
Current Setup
Hardware conversion layerAI orders need to become accepted systems, margin, and cash.
The basket is most useful when backlog and server demand convert into reported revenue without inventory, receivables, financing, capex, or controls absorbing the economics.
Reported proofDELL and HPE have the freshest scale evidence.
Dell reported Q1 FY2027 AI-server revenue and backlog; HPE reported Q2 FY2026 Cloud & AI revenue, AI orders, and FCF.
Manufacturing conversionCLS, JBL, and SANM need clean program economics.
Revenue must show up in segment margin, FCF, customer breadth, and working-capital discipline.
Primary constraintPass-through hardware volume can dilute the read.
Low margin, receivables, inventory, debt, controls, and customer concentration remain the main checks.
Demand sourceHyperscaler + AI cloud
System buildServer, rack, EMS
Accounting proofMargin, OCF, FCF
Live gateCash + controls
This node is where AI capex becomes physical systems that customers can deploy. Hyperscalers, AI-cloud providers, enterprises, and infrastructure partners need servers, racks, storage, networking, liquid-cooling integration, firmware, supply-chain execution, and site-ready shipments. Revenue matters only if it converts into segment margin, operating cash flow, free cash flow, customer-backed backlog, manageable working capital, and filings that investors can rely on.
The strongest current evidence is at Dell, HPE, Celestica, Jabil, and Sanmina. Dell reported Q1 FY2027 revenue of $43.8B, AI-server revenue of $16.1B, AI orders of $24.4B, AI backlog of $51.3B, and FY2027 AI-server guidance near $60B. HPE reported Q2 FY2026 revenue of $10.7B, Cloud & AI revenue of $7.7B, Cloud & AI operating margin of 12.4%, $1.8B of AI Systems orders, and FY2026 FCF guidance of at least $3.5B. Celestica and Jabil still add strong EMS/HPS evidence through CCS, HPS, Intelligent Infrastructure, and AI/data-center manufacturing ramps. Sanmina adds ZT Systems exposure, Q2 FY2026 revenue of $4.01B, and $342M of Q2 free cash flow.
The setup weakens when AI infrastructure revenue looks like pass-through hardware volume instead of durable earnings and cash. Dell still lacks granular AI-server margin, customer concentration, backlog terms, cancellation detail, and services attach. HPE must prove AI backlog conversion, Cloud & AI margin quality, Juniper integration, and debt reduction. Celestica has high customer concentration and a roughly $1B capex plan while FCF guidance stayed at $500M. Jabil remains a low-margin EMS model with heavy receivables-sale usage. Sanmina must prove ZT demand was not pulled forward and that customer concentration, debt, and GAAP versus non-GAAP margin stay controlled. Supermicro remains the highest-risk direct rack-scale route because Q3 FY2026 evidence shows large working-capital cash use, high inventory, debt and convertibles, ineffective controls, customer concentration, and export/compliance uncertainty.
Static setup labels and threshold metadata use completed weekly bars aggregated from discovery daily_ohlc through 2026-06-05. The local store now has daily rows through 2026-06-12, and the selected-security right rail should be treated as the fresher chart route. Fundamental claims route to the knowledge theme page, sector lane, linked security lanes, and bounded official company checks listed in Source Trail.
Basket
This basket is inherited from the AI Capex Cycle servers, racks, and EMS sleeve, then re-ranked by source-backed operating exposure, cash conversion, customer/program evidence, and disclosure quality before technical timing. Visible rows are DELL, HPE, CLS, JBL, SANM, and SMCI. SMCI remains visible because it is the highest-purity rack-scale server route, but it is option-ranked because cash conversion, controls, customer concentration, and export-review risk are still gating items. TSSI, PENG, and SNX stay option or watch routes until local security lanes exist. FLEX is a watch route because its CPI exposure overlaps the power/cooling node and depends on spin, capex, and customer disclosure.
Scale AI-server OEM and enterprise infrastructure integrator with the clearest order, backlog, revenue, and cash-flow evidence.
Market cap$169.3B
Next earningsLate Aug 2026 est.
Latest qtr revenue$43.8B
Role in stack
Dell sells branded servers, storage, PCs, services, and financing to hyperscalers, enterprises, governments, and channel buyers. AI capex converts when AI server orders and backlog become ISG revenue, ISG margin, operating cash flow, adjusted FCF, and storage or services attach.
Revenue mix
The local lane shows FY2026 ISG revenue of $60.8B versus CSG revenue of $51.0B. Q1 FY2027 evidence shifted the mix further toward ISG, with AI-optimized servers alone reported at $16.1B.
Latest qtr revenue
Dell's May 28, 2026 official release reported Q1 FY2027 revenue of $43.8B for the quarter ended May 1, 2026, including $29.0B of ISG revenue and $16.1B of AI-optimized server revenue.
Enterprise AI systems, server, networking, storage, and financing route with fresh Cloud & AI and FCF evidence.
Market cap$41.6B
Next earningsNot confirmed
Latest qtr revenue$10.7B
Role in stack
HPE sells AI systems, servers, storage, networking, private-cloud infrastructure, and financing to enterprises, sovereign buyers, cloud providers, and private-cloud users. Conversion depends on AI Systems orders and backlog becoming Cloud & AI revenue, Networking mix, free cash flow, and post-Juniper debt reduction.
Revenue mix
Q2 FY2026 Cloud & AI revenue was $7.7B, including $5.5B of server revenue in the local lane. Networking revenue was $2.7B and carried higher operating-margin mix after Juniper.
Latest qtr revenue
HPE's June 1, 2026 official release reported Q2 FY2026 revenue of $10.7B for the quarter ended April 30, 2026, with $0.9B of free cash flow and raised fiscal 2026 guidance.
EMS and hardware-platform manufacturer with high data-center exposure through CCS and HPS programs.
Market cap$43.2B
Next earningsNot confirmed
Latest qtr revenue$4.05B
Role in stack
Celestica builds EMS and hardware-platform programs for cloud and AI customers. The mechanism is compute, storage, networking, optical, and platform manufacturing revenue converting into CCS margin, utilization, FCF, and manageable capex.
Revenue mix
Q1 2026 CCS revenue was $3.24B, about 80% of revenue, and HPS revenue was about $1.7B, or 42% of revenue. ATS diversifies the company but does not drive this node ranking.
Latest qtr revenue
Celestica's April 27, 2026 official release reported Q1 2026 revenue of $4.047B for the quarter ended March 31, 2026, with CCS revenue of $3.24B and HPS revenue of about $1.7B.
Outsourced engineering, manufacturing, supply-chain, and product-management route into AI infrastructure programs.
Market cap$37.5B
Next earningsJun 17, 2026 est.
Latest qtr revenue$8.28B
Role in stack
Jabil builds and manages customer programs across cloud and data-center infrastructure, networking, communications, and capital equipment. Theme pressure converts if Intelligent Infrastructure growth lifts margin and adjusted FCF without heavier receivables-sale or working-capital dependence.
Revenue mix
Q2 FY2026 Intelligent Infrastructure revenue was $4.028B, about 49% of revenue and up 52%. Regulated Industries and Connected Living and Digital Commerce diversify the base but are not the main AI hardware route.
Latest qtr revenue
Jabil's March 18, 2026 official release reported Q2 FY2026 net revenue of $8.282B for the quarter ended February 28, 2026. Kiplinger listed the next weekly calendar item for JBL as before the open on June 17, 2026, but the calendar note says dates are tentative.
Post-ZT Systems EMS and integrated manufacturing route into cloud and AI infrastructure.
Market cap$13.3B
Next earningsNot confirmed
Latest qtr revenue$4.01B
Role in stack
Sanmina manufactures complex electronics and integrated systems for OEM and cloud infrastructure programs. The node route now runs through ZT-backed accelerated compute and rack manufacturing, with conversion gated by demand durability, GAAP margin, working capital, debt, and customer concentration.
Revenue mix
The local lane puts Q2 FY2026 communications networks and cloud/AI infrastructure revenue near $2.77B. IMS generated $3.57B of Q2 revenue after ZT, while CPS adds components, products, and services.
Latest qtr revenue
Sanmina's April 27, 2026 official release reported Q2 FY2026 revenue of $4.013B for the quarter ended March 28, 2026, with $342M of free cash flow and Q3 revenue guidance of $3.2B to $3.5B.
Pure rack-scale AI server route with direct exposure but the highest cash, controls, financing, and customer-risk burden.
Market cap$21.2B
Next earningsNot confirmed
Latest qtr revenue$10.24B
Role in stack
Supermicro sells AI servers, GPU/Super Rack systems, storage, liquid-cooled systems, and total data-center infrastructure. It converts only when shipped and accepted systems produce gross margin, receivables collection, inventory release, operating cash flow, and clean control and export-review outcomes.
Revenue mix
The business is reported as one segment. The local lane says FY2025 server and storage systems were 97% of revenue, making SMCI the purest visible rack-scale route but also the least diversified.
Latest qtr revenue
Supermicro's May 5, 2026 official release reported Q3 FY2026 net sales of $10.243B for the quarter ended March 31, 2026. The same release guided Q4 FY2026 net sales to $11.0B to $12.5B.
Market caps use local discovery instruments.market_cap queried on 2026-06-13: DELL $169.3B, HPE $41.6B, CLS $43.2B, JBL $37.5B, SANM $13.3B, and SMCI $21.2B. Next-earnings fields use bounded web checks on 2026-06-13; only JBL had a current calendar source in the searched set and it is marked estimated because Kiplinger states its tables are tentative. DELL uses the local knowledge-lane late-August trigger as an estimate; HPE, CLS, SANM, and SMCI were left as not confirmed.
What Confirms Or Weakens
AreaWhat confirmsWhat weakens or invalidatesWatch next
01Node thesis
Demand becomes systems
What confirms
AI server, rack, HPS, Cloud & AI, Intelligent Infrastructure, and ZT manufacturing revenue converts into segment margin, operating cash flow, FCF, and customer-backed backlog or program visibility.
What weakens or invalidates
Revenue rises while gross margin, receivables, inventory, factoring, capex, controls, customer concentration, or debt absorb the economics.
Watch next
DELL AI-server orders
HPE AI backlog
CLS HPS conversion
JBL AI programs
SANM ZT durability
SMCI cash release
02Economics mechanism
Margin and cash
What confirms
Gross margin and segment margin hold while operating cash flow improves after inventory, receivables, capex, payables, financing, and customer-acceptance timing are included.
What weakens or invalidates
AI systems carry pass-through economics, capex runs ahead of utilization, receivables sales rise, non-GAAP addbacks widen, or inventory and AR consume cash faster than revenue converts.
Watch next
Dell ISG margin
HPE FCF/debt
CLS capex vs FCF
JBL receivables sales
SANM GAAP margin
SMCI AR and inventory
03Customer and backlog quality
Program proof
What confirms
Orders come from multiple customers or contracts with visible acceptance, pricing, cancellation protections, customer payment terms, and deployment timing.
What weakens or invalidates
One or a few customers dominate revenue, accounts receivable, capacity plans, program timing, or backlog quality, especially where terms are undisclosed.
Watch next
Dell backlog customer mix
HPE sovereign/enterprise mix
CLS top-10 concentration
JBL customer detail
SANM top-customer mix
SMCI AR concentration
04Funding and balance sheet
Working-capital gate
What confirms
Higher AI activity is financed by customer advances, operating cash flow, disciplined capex, ordinary payables, and manageable debt service rather than dilutive or emergency funding.
What weakens or invalidates
Equity issuance, expensive debt, customer-payment gaps, inventory prebuilds, receivables sales, or covenant pressure become necessary to fund the build cycle.
Watch next
SMCI financing updates
HPE H3C proceeds
Dell DFS funding
CLS capex plan
SANM debt and cash
05Policy and controls
Filing reliability
What confirms
Filings are current, controls remediation progresses, export-review scope stays contained, acquisition integration is transparent, and tariff pass-through does not erase margin.
What weakens or invalidates
Filing delays, restatement risk, unremediated controls, export restrictions, acquisition-accounting opacity, tariff leakage, or supplier restrictions change delivery timing or economics.
Watch next
SMCI controls/export
HPE Juniper integration
Dell supplier commitments
EMS tariff language
Memory and power routes
06Operating and supply constraint
Build readiness
What confirms
GPU, memory, networking, storage, power, cooling, factory capacity, and site-readiness constraints ease enough for orders to ship without margin concessions.
What weakens or invalidates
HBM, GPU, networking, liquid-cooling, rack-power, or site-delivery bottlenecks delay accepted shipments, inflate bill-of-materials cost, or force lower-margin substitutions.
Watch next
GPU allocation
HBM pricing
Networking supply
Rack power/cooling
Customer site readiness
07Stale condition and source gap
Refresh trigger
What confirms
Discovery daily_ohlc stays current, linked lanes still cover the latest filings and guidance, and the selected-security API remains newer than static setup levels.
What weakens or invalidates
A new completed trading week, earnings release, 10-Q, customer disclosure, export update, source-lane refresh, or guidance update arrives before this page is updated.
Watch next
Refresh setup levels
Check Q2/Q3 filings
Reconcile lineage
Add lanes for TSSI/PENG/SNX
Source Trail
Canonical Thesis
AI Capex Cycle defines the downstream hardware conversion test: AI demand and hyperscaler budgets have to become systems revenue, margin, and FCF.
Information Technology supports the broad AI infrastructure route across servers, storage, networking, data-center systems, semiconductors, and implementation services.
Security Lanes
DELL supports Q1 FY2027 AI server revenue, AI orders, backlog, FY2027 AI server guide, ISG margin, cash conversion, and backlog-quality caveats.
HPE supports Cloud & AI revenue, server revenue, AI Systems orders and backlog, Networking margin, Juniper integration, FCF guidance, and debt-reduction proof.
SANM supports ZT Systems integration, cloud/AI infrastructure manufacturing, Q2 FY2026 cash generation, customer concentration, debt, and GAAP versus non-GAAP margin caveats.
SMCI supports AI rack-scale exposure, Q3 FY26 revenue and gross margin, operating-cash-use risk, debt and cash position, customer concentration, controls, and export-review risk.
FLEX supports the adjacent CPI route through data-center power, cooling, and compute integration, but it is treated as watch-only here because the exposure cross-routes to the power/cooling node.
No local security lanes were found for TSSI, PENG, or SNX during this refresh. They remain option or watch routes until local lane work catches up with the official sources.
Companion Routes
Memory And Component Scarcity owns HBM, DRAM, NAND, GPU allocation, component-cost, and bill-of-material pressure that can change system margin and shipment timing.
Power Scarcity And Grid Load owns power, cooling, interconnection, and site-delivery constraints that determine whether delivered systems can become usable data-center capacity.
Market Fragility Amplification should be reopened when beta, short interest, or AI-hardware crowding changes the risk of a fast unwind.
Bounded Official Checks
Dell Q1 FY2027 results release was used to check Q1 revenue, AI server revenue, AI orders, AI backlog, ISG margin, operating cash flow, adjusted FCF, Q2 guidance, and FY2027 AI server revenue guidance.
HPE Q2 FY2026 release and Q2 presentation were used to check revenue, Cloud & AI revenue and margin, Networking revenue and margin, AI Systems orders, AI backlog, FCF, and FY2026 guidance.
Supermicro Q3 FY26 release and Q3 FY26 Form 10-Q were used to check Q3 sales, gross margin, operating cash use, cash, debt plus convertibles, customer concentration, and Q4/FY26 revenue guidance.
Celestica Q1 2026 release was used to check Q1 revenue, adjusted operating margin, CCS revenue and margin, HPS revenue, FCF guide, and the raised 2026 outlook.
Jabil Q2 FY26 release was used as the official release route for Q2 results; detailed AI revenue, receivables-sale, and segment commentary route through the local JBL security lane and its raw sources.
Sanmina Q2 FY2026 release was used to check Q2 revenue, free cash flow, ZT commentary, Q3 guide, FY2026 guide, and ZT integration risk language.
Flex FY2026 release was used only for watch-route context around CPI, data-center power/cooling/compute integration, FCF, FY2027 guidance, and the planned CPI spin.
The chart API reads daily_ohlc from the local discovery store and returns weekly packages using first open, maximum high, minimum low, final close, and summed volume.
Read-only DuckDB coverage checks on 2026-06-13 showed local daily_ohlc rows through 2026-06-12 for DELL, HPE, CLS, JBL, SANM, and SMCI, with rows starting on 2021-05-12 for each ranked ticker.
Static setup labels and thresholds in this node's sidecar metadata still use completed weekly bars through 2026-06-05. The selected-security right rail calls /api/securities/TICKER/chart?frequency=weekly&window=3y&as_of=latest, so refresh static setup levels before using them as current trading evidence.
Ticker-level lineage still shows provider freshness around 2026-05-08 while the canonical daily_ohlc table has rows through 2026-06-12. Reconcile lineage before using it as live provider freshness.
Right-rail charts are expected to show weekly OHLC, 20W EMA, 100W EMA, weekly volume, and 20W average volume over a three-year visible horizon. Chart evidence is price and volume context only; it does not infer AI-capex flows.
Earnings Date And Revenue Checks
DELL: next earnings shown as Late Aug 2026 est. from the local DELL knowledge-lane trigger list; no official IR event date was found in bounded web checks on 2026-06-13. Latest-quarter revenue uses Dell's May 28, 2026 Q1 FY2027 release for the quarter ended May 1, 2026.
HPE: next earnings shown as Not confirmed; bounded web checks on 2026-06-13 found the June 1, 2026 Q2 FY2026 release but no credible next-event date. Latest-quarter revenue uses HPE's June 1, 2026 release for the quarter ended April 30, 2026.
CLS: next earnings shown as Not confirmed; bounded web checks found the April 27, 2026 Q1 release and no credible next-event date. Latest-quarter revenue uses Celestica's Q1 2026 release for the quarter ended March 31, 2026.
JBL: next earnings shown as Jun 17, 2026 est. from Kiplinger's June 12, 2026 earnings calendar; Kiplinger states calendar-table dates are tentative. Latest-quarter revenue uses Jabil's March 18, 2026 Q2 FY2026 release for the quarter ended February 28, 2026.
SANM: next earnings shown as Not confirmed; bounded web checks found the April 27, 2026 Q2 release and no credible next-event date. Latest-quarter revenue uses Sanmina's Q2 FY2026 release for the quarter ended March 28, 2026.
SMCI: next earnings shown as Not confirmed; bounded web checks found the May 5, 2026 Q3 FY2026 release and no credible next-event date. Latest-quarter revenue uses Supermicro's Q3 FY2026 release for the quarter ended March 31, 2026.
Market-cap metrics use local discovery instruments.market_cap queried read-only on 2026-06-13, not a live quote feed.
Known Gaps
Dell still needs AI server gross margin, storage/services attach, customer concentration, backlog terms, cancellation protections, DFS funding quality, and the June 25, 2026 redomestication outcome.
HPE needs Q3 guide delivery, AI Systems backlog conversion, Cloud & AI margin quality, Juniper integration, H3C proceeds and debt reduction, and GAAP/non-GAAP comparability.
Celestica needs Q2 2026 results, 2027 HPS/CPO visibility, customer concentration updates, capex versus utilization, and FCF versus the $500M guide.
Sanmina needs Q3 durability after Q2 pull-forward, standalone ZT margin, customer concentration, GAAP versus non-GAAP margin bridge, working capital, debt, and covenant detail.
Supermicro needs Q4 revenue conversion, operating-cash-flow recovery, AR and inventory release, gross margin, debt/convertibles, customer concentration, controls remediation, and export-control updates.
TSSI, PENG, and SNX have official evidence and local price coverage, but no local security lanes. They should stay option or watch routes until source lanes are created.
FLEX is relevant through CPI, data-center power, cooling, and compute integration, but it cross-routes to the power/cooling node. Do not overstate it as a pure server or EMS name until SpinCo/Form 10 and CPI customer/margin evidence are available.