NYSE · AI-infra ODM · 1.6T switches & full racks
Celestica
An ODM of networking and AI systems for hyperscalers. It designs and builds 1.6T switch programs (a second hyperscaler), liquid-cooled AI racks and full-rack AI systems (HPS). 2026 revenue outlook $16B (+31%). With white-box switches and AI/ML servers, an ODM axis distinct from Cisco and Dell.
CLS Updated Jul 12, 2026 4 AI consumers
Per the valuechain.wiki graph, Celestica is directly linked to 3 companies (1 supply, 2 revenue relationships) and reaches 51 companies within two hops; the links are backed by 3 dated sources, 1 of them primary filings (as of 2026-07-12).
Financials & Segments
- Total revenue · 2026 outlook $16B (+31%)
- Customer concentration · 1.6T switches; full-rack AI systems (HPS).
- Segments · Connectivity & Cloud (CCS) · AI/ML servers, 1.6T switches · Advanced Tech
Revenue from
- Hyperscalers · neoclouds (networking · AI racks) High confidence
Supplies 1.6T switches and full-rack AI systems to hyperscalers and neoclouds — the second-hyperscaler 1.6T win signals customer diversification. - Amazon Medium confidence
Networking and AI systems to AWS — a large-customer axis tied to hyperscaler capacity build-out. - enterprise · industrial (EMS) Medium confidence
Enterprise and industrial EMS (CCS/ATS) — a revenue base beyond hyperscalers that softens cycle swings. - Meta Platforms Medium confidence
1.6T switches and AI racks to Meta — a channel directly benefiting from Meta's expanding in-house AI infrastructure.
Pays to
- NVIDIA High confidence
Builds AI/ML servers and full racks (HPS) with NVIDIA GPUs — GPU allocation is effectively the ceiling on AI-system shipments, a core cost and supply dependency. - components · memory · manufacturing (undisclosed) Low confidence
Components, memory and precision assembly — a large slice of white-box ODM cost where sourcing drives margin (undisclosed). - cooling · capacity expansion (undisclosed) Low confidence
Liquid-cooling systems and capacity investment — the bottleneck for delivering dense AI racks, where cooling capacity caps full-rack (HPS) shipments.
Value-chain ripple (2 tiers)
Following node-to-node links up to 2 tiers; each firm is shown once, at its nearest tier.
| Company | Chain | '21 | '22 | '23 | '24 | '25 | YTD |
|---|---|---|---|---|---|---|---|
| Datadog | lead2 | +80.9 | -58.7 | +65.1 | +17.7 | -4.8 | +89.4 |
| Palo Alto Networks | lead2 | +56.7 | -24.8 | +111.3 | +23.4 | +1.2 | +76.9 |
| CrowdStrike | lead2 | -3.3 | -48.6 | +142.5 | +34 | +37 | +59.7 |
| Illumina | lead2 | +2.8 | -46.9 | -31.1 | -1.3 | -1.8 | +45 |
| Snowflake | lead2 | +20.4 | -57.6 | +38.6 | -22.4 | +42.1 | +19.2 |
| Airbnb Inc | lead2 | · | -48.6 | +59.2 | -3.5 | +3.3 | +9.5 |
| Procter & Gamble | lead2 | +20.5 | -5 | -0.9 | +17.3 | -12.3 | +4.1 |
| Schrodinger | lead2 | -56 | -46.3 | +91.5 | -46.1 | -7.3 | -7.8 |
| MercadoLibre, Inc. | lead2 | -19.5 | -37.2 | +85.7 | +8.2 | +18.5 | -8 |
| Uber | lead2 | -17.8 | -41 | +149 | -2 | +35.5 | -8.8 |
| The Walt Disney Company | lead2 | -14.5 | -43.9 | +4.3 | +24.4 | +3.3 | -15.3 |
| Coupang | lead2 | · | -49.9 | +10.1 | +35.8 | +7.3 | -20.3 |
| Netflix | lead2 | +11.4 | -51.1 | +65.1 | +83.1 | +5.2 | -21.7 |
| Amazon | lead1 | +2.4 | -49.6 | +80.9 | +44.4 | +5.2 | +6.3 |
| Meta Platforms | lead1 | +23.1 | -64.2 | +194.1 | +66 | +13.1 | +1.6 |
| Celestica | node | +37.9 | +1.3 | +159.8 | +215.2 | +220.3 | +21.7 |
| NVIDIA | benef1 | +125.5 | -50.3 | +239 | +171.2 | +38.9 | +13.3 |
| Kioxia | benef2 | · | · | · | · | +1082.1 | +260.5 |
| Micron Technology | benef2 | +24.2 | -45.9 | +71.9 | -1 | +240.2 | +243.3 |
| Murata Manufacturing | benef2 | -14 | -11.7 | +24.7 | -16.5 | +29.9 | +217.4 |
| Nan Ya PCB | benef2 | +110.6 | -46.7 | -0.5 | -38.3 | +188.2 | +213.4 |
| ASE Technology | benef2 | +38.3 | -12.8 | +60.9 | +10.1 | +65.6 | +167.6 |
| Nebius Group | benef2 | · | · | · | · | +202.2 | +162.4 |
| Astera Labs | benef2 | · | · | · | · | +25.6 | +148.2 |
| Ibiden | benef2 | +30.4 | -19.6 | +51.6 | -38.6 | +261 | +147.1 |
| SK hynix | benef2 | -0.4 | -25.6 | +53.9 | +48.6 | +361.1 | +140.3 |
| Unimicron | benef2 | +133.5 | -30.2 | +35.2 | -19.7 | +175.7 | +131.6 |
| Corning | benef2 | +5.9 | -11.6 | -1.2 | +60.6 | +87.8 | +118.8 |
| Lumentum | benef2 | +11.6 | -50.7 | +0.5 | +60.1 | +339.1 | +117.6 |
| Samsung Electronics | benef2 | -9 | -14.2 | +20.9 | -26.4 | +212.9 | +77.9 |
| onsemi | benef2 | +107.5 | -8.2 | +33.9 | -24.5 | -14.1 | +77.2 |
| Coherent | benef2 | -10 | -48.6 | +24 | +117.6 | +94.8 | +75.8 |
| Eoptolink | benef2 | -10.1 | -27.7 | +136.6 | +197 | +368.8 | +74.9 |
| InnoLight | benef2 | -30.7 | -18.3 | +253.7 | +12 | +468.1 | +68.8 |
| Keysight Technologies | benef2 | +56.3 | -17.2 | -7 | +1 | +26.5 | +58.5 |
| Delta Electronics | benef2 | -3.4 | +10.4 | -0.3 | +59.2 | +183.8 | +54.9 |
| Monolithic Power | benef2 | +35.5 | -27.8 | +79.8 | -5.6 | +54.5 | +49.8 |
| Analog Devices | benef2 | +21 | -4.9 | +23.4 | +8.8 | +29.8 | +46.7 |
| TSMC | benef2 | +12.1 | -36.7 | +42.3 | +92.6 | +55.9 | +43.6 |
| ABB | benef2 | +23.9 | +7.3 | +18.9 | +39 | +36.1 | +27.2 |
| Pegatron | benef2 | -5.9 | +0.6 | +35.3 | +21.2 | -23.8 | +24 |
| Luxshare Precision | benef2 | -11.9 | -30.2 | -21.2 | +58.8 | +29.3 | +20.6 |
| Digital Realty | benef2 | +30.7 | -41 | +39.9 | +35.9 | -10.1 | +18.2 |
| Amphenol | benef2 | +34.9 | -12 | +31.5 | +41.3 | +96.1 | +18.1 |
| Wistron | benef2 | +7.9 | +9.7 | +272.7 | -1.6 | +22.7 | +14.8 |
| King Yuan Electronics (KYEC) | benef2 | +26.7 | -7.7 | +132.4 | +38.6 | +171 | +4.2 |
| BlackRock | benef2 | +29.4 | -20.4 | +17.9 | +29.3 | +6.6 | -2.1 |
| TE Connectivity | benef2 | +35.1 | -27.7 | +24.6 | +3.5 | +61.1 | -11.3 |
| Recursion Pharmaceuticals | benef2 | · | -55 | +27.9 | -31.4 | -39.5 | -13 |
| BYD Electronics | benef2 | -56.1 | +16 | -0.4 | +60.6 | -19.4 | -34.4 |
YTD as of 2026-07-10 · source: Yahoo Finance
Celestica's thesis is a re-rating from EMS subcontractor to hyperscaler AI-infrastructure ODM. By designing and building white-box switches, AI/ML servers and full-rack systems (HPS) itself, it sits on a different axis from Cisco and Dell, and the 2026 revenue outlook of $16B (+31%) is driven by 1.6T switches and full-rack AI systems in the CCS (Connectivity & Cloud) segment. The core driver is hyperscaler capacity build-out — with Meta and AWS as customers and a second hyperscaler's 1.6T switch program won, it has broadened its base. On the cost side it depends on NVIDIA GPUs and liquid-cooling/component sourcing, so GPU allocation and cooling capacity govern lead times. Re-rating triggers are the 1.6T switch ramp, full-rack (HPS) attach, and the durability of ODM margin, with concentration on a few hyperscalers a two-way risk.
Analysis generated from returns as of 2026-07-09Analyst ratings
Consensus: Buy 5 · Hold 0 · Sell 0 · Avg. target $454 · Price $359.85 (07-10) · Upside +26%
| Broker | Rating | Target | Published | |
|---|---|---|---|---|
| Wolfe Research · George Notter | Buy | $425 | Jun 15, 2026 | Report ↗ |
| RBC Capital · Paul Treiber | Buy | $440 | Jun 2, 2026 | Report ↗ |
| Citi · Nilesh Bhaiya | Buy | $475 | May 27, 2026 | Report ↗ |
| Barclays · Tim Long | Buy | $441 | May 8, 2026 | Report ↗ |
| Bank of America Securities · Ruplu Bhattacharya | Buy | $490 | Apr 30, 2026 | Report ↗ |
Buy at $425, positive on the 400G/800G networking transition and differentiated custom-design capability versus commodity ODMs.