Current rate from 1 AUD to CNY is ¥4.86231 at time 2026-05-26T00:00:00+0000, more info please check <https://www.tradingview.com/chart/MozVMfOY/?symbol=BITSTAMP%3ABTCUSD>
Across these six names, the cleanest conclusion is that **all six are genuinely relevant to AI data-centre buildout**, but they sit at very different points in the stack and deserve very different valuation frameworks. **Marvell** and **Micron** are the most semiconductor-levered beneficiaries. **Lumentum** and **Coherent** are the most direct photonics/optics component plays. **Corning** is the fibre-and-connectivity enabler rather than a chip company. **Nokia** is the systems-and-network-domain participant, with growing AI-cloud optical/IP relevance but a much broader telecom backdrop. citeturn11search11turn21view1turn13search4turn40search6turn38view3turn15search1
On competitive position, my ranking for **core technology defensibility in chips and optical communications** is: **Micron** (memory/HBM, scale, process roadmap), **Marvell** (custom AI silicon, switching, DSP/coherent interconnect), **Coherent** (broad photonics vertical integration), **Lumentum** (cloud/datacom laser and optical-component concentration), **Corning** (fibre/connectivity materials leadership, but less chip leverage), and **Nokia** (strong optical/IP systems and Bell Labs heritage, but lower pure-play AI exposure than the others). That ranking is about **technology exposure**, not expected stock returns. citeturn38view3turn43search14turn13search6turn21view1turn40search6turn15search0
On valuation, the market is already pricing in a large part of the AI buildout story. By current multiples, **Marvell, Lumentum, Coherent, Corning, and Nokia all look expensive versus their historical business quality and cyclicality**, while **Micron is the one name where valuation looks most defensible because forward earnings are rising much faster than near-term price/sales optics imply**. In other words, the current market is paying for AI scarcity and design-win optionality almost everywhere, but Micron has the strongest **earnings-backed** case among the group. citeturn44search1turn43search1turn43search2turn43search3turn41search5turn44search0
My research-only bottom line is: **Micron = best relative risk/reward**, **Marvell = high-quality but needs disciplined entry**, **Corning = reasonable strategic hold on pull-backs**, **Coherent = attractive business but valuation-rich**, **Lumentum = powerful momentum but the most demanding valuation**, **Nokia = improving strategically but probably ahead of itself after the rally**. These are research views only, not trading instructions. citeturn44search0turn44search1turn43search3turn43search2turn43search1turn42search23
## Comparative snapshot
The table below condenses the highest-confidence current picture. Valuation ratios are current market-data snapshots around 27 May 2026; revenues and margins come from the latest annual report or latest reported quarter, depending on availability. “Current profitability” uses the latest clearly disclosed GAAP metrics when available; where a metric was not cleanly disclosed in the accessible source set, I mark it as unspecified rather than guessing. citeturn44search1turn43search1turn43search2turn43search3turn44search0turn41search5
| Company | Core AI-interconnect role | Latest FY revenue | Current profitability snapshot | Cash / debt snapshot | Current valuation snapshot | Recent analyst consensus snapshot | Research view |
|---|---|---:|---|---|---|---|---|
| **Marvell (MRVL)** | Custom AI silicon, Ethernet switching, DSP/coherent optics, interconnect | **US$8.19bn FY2026** | FY2026 GAAP gross margin **51.0%**; GAAP net margin c. **32.6%**; GAAP operating margin not cleanly reproduced in accessible summary | FY2026 year-end cash **US$0.95bn**; debt c. **US$4.06bn** at FY-end | Mkt cap **US$182.1bn**; EV **US$184.3bn**; **P/E 67.8x**, **Fwd P/E 54.1x**, **P/S 22.2x**, **EV/EBITDA 70.1x** | WSJ average target **US$157.03** vs current **US$196.33**; consensus **Buy** | **Hold / accumulate only on pull-backs** |
| **Lumentum (LITE)** | EML/CW lasers, optical components, cloud/datacom subsystems, optical circuit switching | **US$1.65bn FY2025** | FY2025 GAAP gross margin **28.0%**; GAAP operating margin **-10.9%**; latest Q3 FY2026 revenue and profitability improved sharply | Q2 FY2026 cash + ST investments **US$1.16bn** | Mkt cap **US$70.9bn**; EV **US$71.0bn**; **P/E 173.2x**, **Fwd P/E 57.5x**, **P/S 28.5x** | Barron’s/WSJ-style target snapshot: average near **US$1,127**; Public/Benzinga-type sources also show **Buy** consensus | **Avoid chasing / tracking position only** |
| **Coherent (COHR)** | Lasers, photonic devices, SiPh/InP components, transceivers, optical materials | **US$5.81bn FY2025** | Latest Q3 FY2026 GAAP gross margin **37.7%**; GAAP net income positive; GAAP operating margin not cleanly reproduced in accessible summary | Balance-sheet cash/debt not fully reproduced in accessible summary; EV suggests modest net debt | Mkt cap **US$74.6bn**; EV **US$75.6bn**; **P/E 159.2x**, **Fwd P/E 51.2x**, **P/S 11.3x**, **EV/EBITDA 57.6x** | MarketWatch/Fintel-style average target about **US$384**; recommendation **Overweight/Buy** | **Wait for better entry** |
| **Micron (MU)** | HBM, DRAM, NAND, data-centre SSDs, memory subsystem scaling | **US$37.38bn FY2025** | Latest Q2 FY2026 GAAP gross margin **74.4%**; GAAP operating margin **67.6%**; GAAP net margin c. **57.8%** | Q2 FY2026 cash + investments **US$16.7bn**; FY2025 debt **US$14.4bn** | **P/E 42.1x**, **Fwd P/E 9.6x**, **P/S 17.4x**, **EV/EBITDA 27.3x** | WSJ median target **US$600**, average **US$684.32**; Reuters notes a high target of **US$1,625** after UBS raised estimates | **Best relative idea in the group** |
| **Nokia (NOK)** | Optical transport, coherent/line systems, IP routing, data-centre switching, AI-cloud networking | **€19.89bn FY2025** | FY2025 reported gross margin **43.5%**; operating margin **4.4%**; net margin c. **3.3%** | Q4 2025 net cash + interest-bearing financial investments **€3.38bn** | Mkt cap about **US$91.9bn**; **P/E 98.2x**, **Fwd P/E 38.9x**, **P/S 3.9x**, **EV/EBITDA 29.9x** | MarketWatch/Barron’s average target about **US$11.6**; Fintel average target ~**US$10.2** | **Trim / don’t chase** |
Sources: Marvell filings and market statistics citeturn11search0turn22view0turn44search1turn41search2; Lumentum releases and statistics citeturn16view2turn16view0turn43search1turn42search8; Coherent releases, presentation and statistics citeturn16view5turn16view4turn43search2turn42search13; Corning quarterly/annual releases and statistics citeturn40search3turn17view1turn43search3turn42search14; Micron releases, 10-K, statistics, and analyst pages/news citeturn35view2turn36view2turn39view1turn44search0turn41search3turn41news41; Nokia reports and valuation/target pages citeturn45view0turn16view6turn41search5turn42search23turn42search11turn44search4
## Technology and competitive positions
**Marvell** is the most complete AI-networking semiconductor platform in this group. Its disclosed stack spans **custom silicon, Ethernet switches, DSP/coherent interconnect, SerDes, DPUs and broader interconnect IP**. In FY2026, data-centre revenue reached **US$6.10bn**, or **74%** of total revenue, confirming that Marvell is no longer just a diversified infrastructure chip vendor but a data-centre-first supplier. That is strategically important: hyperscalers increasingly want merchant-plus-custom options, and Marvell is one of the few vendors positioned across the **custom-XPU adjacencies**, **switching fabric**, and **optical/electrical interconnect** layers. The trade-off is that scale-up optimism is already capitalised into the stock. citeturn43search14turn23view1turn29view1turn44search1
Marvell’s near-term AI interconnect roadmap is especially strong. Recent public announcements included **NVLink Fusion ecosystem participation with NVIDIA**, the **Polariton acquisition to push optical performance scaling to 3.2T and beyond**, and a February 2026 launch claiming **industry-first 1.6T ZR/ZR+ pluggables and 2nm coherent DSPs** for AI scale-across networks. Those releases materially strengthen Marvell’s competitive position in both **inside-the-rack electrical connectivity** and **between-cluster optical transport**. citeturn11search4turn11search12
**Lumentum** is a focused optical/photonics supplier rather than a semiconductor generalist. Its Cloud & Networking business is built around **optical and photonic chips, components, modules and subsystems** used in intra-data-centre links, DCI, metro, long-haul and submarine networks. That includes the most commercially relevant AI-era building blocks: **laser chips and laser assemblies, line subsystems, wavelength-management systems, cloud transceivers, optical circuit switches, and co-packaged-optics-enabling components**. The business mix has shifted decisively toward cloud/datacom, which is the right place to be, but concentration risk is correspondingly high. citeturn21view1turn19view0turn33view0
Lumentum’s own commentary is notable because management explicitly identified **optical circuit switches**, **co-packaged optics**, and **cloud transceivers** as its “next major growth engines”. Q1 FY2026 commentary also pointed to strong momentum across **data centre, DCI and long-haul** markets, while the FY2025 annual report tied Cloud & Networking growth to **higher unit sales from cloud and AI/ML customers** and the Cloud Light acquisition. That is the right strategic direction, but it also means the stock now depends on sustained hyperscaler capex and clean optical-component ramps. citeturn19view0turn21view1
**Coherent** has perhaps the broadest underlying photonics platform in the group. Its portfolio combines **engineered materials, lasers, optoelectronic devices, compound semiconductor capability, VCSELs, InP/SiPh-related assets, optical transceivers and subsystems**, and it is increasingly framing the company around **Datacenter & Communications** plus **Industrial**. The key competitive advantage is vertical breadth: Coherent can participate from active optical chips and laser sources through packaging and systems-level modules. That breadth matters as AI networks push toward higher lane speeds and more complex optical architectures. citeturn13search4turn16view3turn16view5
Coherent’s most helpful disclosure for investors is that **Datacenter & Communications represented 72% of revenue in Q2 FY2026**, up from 63% in FY25 Q2 on the investor presentation chart, while management also highlighted a **next-generation 2D VCSEL-array-based solution for AI scale-up networks**. In practice, that means Coherent is moving away from being “just” a diversified photonics company and toward being a datacom/AI-optics compounder — but with meaningful industrial cyclicality still attached. citeturn16view3turn13search11
**Corning** is often misunderstood in AI discussions because it is **not a chip company** in the usual sense. Its relevance is in **optical fibre, cable, connectivity hardware, and dense optical infrastructure**, which form the physical layer under AI data-centre and campus buildouts. Corning’s own optical-communications materials explicitly target carriers, data centres and enterprise networks, and management has repeatedly said that **Optical Communications’ enterprise sales are being lifted by new GenAI products**. The competitive moat is manufacturing/process leadership and installed-base relationships, not semiconductor IP. citeturn40search6turn10search16turn40search15
Corning’s “Springboard” plan makes the AI linkage concrete. Management upgraded the plan and is targeting **US$20bn annualised sales run-rate by end-2026**, with much of the optical upside tied to high-density connectivity and AI data-centre products. That is strategically attractive because fibre demand is real and persistent; however, Corning still carries exposure to displays, environmental technologies and life sciences, so the stock should not be valued like a pure AI semiconductor name. citeturn40search9turn40search15turn40search3
**Micron** sits outside optical communications but squarely inside the AI-capacity problem. Its four reportable segments — **Cloud Memory, Core Data Center, Mobile & Client, and Automotive & Embedded** — show how the business has been reorganised around AI-era demand pools. The key technologies are **HBM, DDR5/LPDDR, GDDR, NAND, data-centre SSDs, advanced packaging and process-node transitions such as 1-gamma DRAM and G9 NAND**. In AI systems, memory bandwidth and capacity are no longer secondary inputs; they are hard constraints. That makes Micron strategically critical. citeturn38view3turn38view2turn37view0
Micron’s FY2025 10-K explicitly ties CMBU to **large hyperscale cloud customers and HBM for all data-centre customers**, and CDBU to **data-centre memory and storage solutions**. It also identified products such as the **9550 series SSD**, **6500 ION**, and early **PCIe Gen6 SSD** activity, while the Q2 FY2026 release showed a dramatic acceleration in both cloud memory and core data-centre revenue. Among the six names here, Micron has the strongest evidence that AI demand is already translating into **very large, margin-accretive earnings**, not just roadmap optionality. citeturn38view3turn38view2turn36view2
**Nokia** is the least “pure” AI play here, but it has become more relevant than the market used to assume. Nokia’s AI-data-centre case runs through **Optical Networks, IP Networks, pluggables, line systems, data-centre switching and Bell Labs / routing software capability**, now strengthened by the **Infinera acquisition** and the company’s stated push into **AI & Cloud design wins**. The strategic problem for Nokia is that these opportunities coexist with more mature and more cyclical mobile-network businesses. citeturn15search0turn45view0
Still, Nokia’s recent disclosures were better than many investors appreciate. In Q1 2026, management said Optical Networks grew **20%**, that it won **AI & Cloud design wins and orders for both pluggables and line systems**, and that IP progress is helping it expand its presence “inside the data centre”. The Q4/full-year 2025 report also said demand from AI & Cloud customers was driving optical and IP book-to-bill above one. Strategically, Nokia is no longer just a telecom-equipment story; the question now is whether that transition is durable enough to justify the stock’s rerating. citeturn16view6turn45view0
## Financial performance and revenue mix
The clearest revenue-mix change in this group is at **Marvell**. Its last four reported quarters show a sustained increase in data-centre concentration: **Q1 FY2026 data-centre revenue US$1.441bn**, **Q2 US$1.491bn**, **Q3 US$1.518bn**, **Q4 US$1.651bn**. By Q4 FY2026, Marvell had also simplified reporting into **Data Center** and **Communications and other**, with Q4 revenue of **US$1.651bn** and **US$0.567bn** respectively. For FY2026, Marvell reported **US$6.10bn data-centre revenue** and **US$2.09bn communications and other**. Latest annual shipment geography was **China US$2.97bn**, **Taiwan US$1.66bn**, **United States US$1.17bn**, and **Other US$2.39bn**. This is a very AI-centric mix, but it also leaves the company highly sensitive to cloud/customer concentration and export-control complexity. citeturn28view3turn28view2turn29view0turn29view1turn23view1
**Lumentum** also delivered a significant revenue inflection, but with more concentrated product economics. The last four quarters by product type were: **Q4 FY2025 Components US$424.1m / Industrial Tech US$56.6m** under the old segment lens, then **Q1 FY2026 Components US$379.2m and Systems US$154.6m**, **Q2 FY2026 Components US$443.7m and Systems US$221.8m**, and **Q3 FY2026 Components US$533.3m and Systems US$275.1m**. FY2025 revenue by segment was **Cloud & Networking US$1.411bn** and **Industrial Tech US$234m**. FY2025 region mix was **Americas US$481m**, **Asia-Pacific US$1.001bn**, **EMEA US$164m**, with Hong Kong and Thailand especially important shipment locations. The takeaway is simple: the business is pivoting quickly toward cloud/datacom optical demand, and that is showing up in both growth and margin recovery. citeturn18view2turn19view0turn18view0turn18view3turn21view1turn21view2
For **Coherent**, the reporting change matters. The investor presentation showed a five-quarter progression in revenue distribution from **Datacenter & Communications US$905m / Industrial US$530m** in FY25 Q2 to **US$1,208m / US$478m** in FY26 Q2, and total revenue rose further to **US$1.81bn** in FY26 Q3. FY2025 full-year revenue was **US$5.81bn**. That mix shift is strategically bullish because Datacenter & Communications is structurally better aligned with AI than the industrial businesses, but it also means the stock is now partly pricing Coherent as a datacom growth company rather than as a mixed photonics manufacturer. Geography detail for the last four quarters was not cleanly reproduced in the accessible source set and is therefore left unspecified here. citeturn16view3turn16view4turn16view5
For **Corning**, the key financial point is less about segment-level decomposition and more about the profitability quality of the optical uplift. Q1 2026 delivered **US$4.144bn** GAAP sales, **36.9%** gross margin, and **15.4%** operating margin; full-year 2025 delivered **US$15.63bn** GAAP sales, **36.0%** gross margin and **14.6%** operating margin. Management also highlighted that Optical Communications enterprise sales grew **81% year on year** in Q2 2025 on demand for GenAI products. Full quarterly segment and geography schedules exist in Corning’s detailed financial statements, but they were not fully reproduced in the accessible summary sources I used here, so I am not inventing a more granular quarterly table. citeturn17view1turn40search3turn10search16turn40search2
For **Micron**, the last four reported quarters by business unit are unusually clear and very useful. **Q3 FY2025** revenue was **Cloud Memory US$3.386bn**, **Core Data Center US$1.530bn**, **Mobile & Client US$3.255bn**, **Automotive & Embedded US$1.127bn**. **Q4 FY2025** was **US$4.543bn / US$1.577bn / US$3.760bn / US$1.434bn**. **Q1 FY2026** was **US$5.284bn / US$2.379bn / US$4.255bn / US$1.720bn**. **Q2 FY2026** then surged to **US$7.749bn / US$5.687bn / US$7.711bn / US$2.708bn**. FY2025 revenue by business unit was **CMBU US$13.524bn**, **CDBU US$7.229bn**, **MCBU US$11.859bn**, **AEBU US$4.753bn**; geographic revenue by customer HQ was **US US$24.113bn**, **Taiwan US$5.672bn**, **Mainland China US$2.639bn**, **Other Asia Pacific US$1.913bn**, **Hong Kong US$1.138bn**, **Japan US$895m**, **Europe US$625m**. This is the strongest hard financial evidence in the group that AI demand is moving the entire P&L. citeturn35view2turn36view0turn36view2turn38view2turn39view1
For **Nokia**, the latest accessible detail shows FY2025 **reported net sales of €19.889bn** and Q4 2025 revenue of **€6.125bn**, with Q4 segment net sales of **Network Infrastructure €2.407bn**, **Mobile Networks €2.502bn**, **Cloud and Network Services €837m**, and **Nokia Technologies €384m**. In Q1 2026, management said group net sales grew **4%**, Network Infrastructure grew **6%**, and Optical Networks grew **20%**, with important AI-cloud design wins in pluggables and line systems. Because Nokia shifted its operating model entering 2026, a fully like-for-like last-four-quarters segment bridge is harder than it is for Micron or Marvell, but the strategic direction is still clear: AI-related optical/IP demand is now a real contributor. citeturn45view0turn16view6
## Valuation and fair-value framework
The market is currently pricing **Marvell** as a premium AI-infrastructure semiconductor franchise: roughly **22.2x sales**, **54.1x forward earnings**, and **70.1x EV/EBITDA** on current market-data sources. That is a very demanding multiple set even for a company with dominant AI-interconnect positioning. My fair-value framework therefore assumes Marvell deserves a premium, but not an unconstrained one: roughly **14x–18x forward sales** or an equity value band that translates to about **US$174–US$240 per share** on a medium-term base case. On that basis the stock is **not obviously cheap**; it is closer to **fair-to-rich**, with upside dependent on sustained custom-silicon wins and optical attach-rate expansion. citeturn44search1turn11search0turn23view1turn41search2
**Lumentum** is the most difficult valuation in the group to support on conventional metrics. Current market-data sources show roughly **28.5x sales** and **57.5x forward earnings**. That is extremely rich for a component supplier, even one riding a powerful datacom super-cycle. The fair-value question is not “is the business improving?” — it clearly is — but whether the market is paying too far in advance for margin normalisation, co-packaged optics, and optical-switching scale. My justified range is materially below the current market cap unless FY2027 revenue and margins move dramatically above current trajectory; in practice I would only call it attractive on a substantial de-rating or after several quarters of proof that revenue above the recent run-rate is sustainable. citeturn43search1turn19view0turn18view3turn42search8
**Coherent** looks somewhat less extreme than Lumentum in sales terms, but still rich. Current market-data sources show about **11.3x sales**, **51.2x forward earnings**, and **57.6x EV/EBITDA**. For a business that still has industrial exposure, integration complexity, and reporting-transition noise, that is a premium multiple set. My fair-value framework would be more comfortable closer to **7x–9x forward sales** or **18x–24x EBITDA**, which implies the market is currently assuming a very long datacom runway with limited cyclicality. That may eventually prove correct, but today I think the stock offers a less attractive entry point than the business quality alone would suggest. citeturn43search2turn16view3turn16view4turn42search13
**Corning** also screens expensive relative to how the market normally values high-quality industrial/materials businesses. The stockanalysis snapshot shows about **10.3x sales** and **58.2x forward earnings**. That is a rich rating for a company whose AI upside is real but indirect, and whose portfolio still includes slower or less-AI-linked businesses. My fair-value range is therefore more conservative than the market’s current enthusiasm: roughly **6x–8x sales** for a high-quality optical-infrastructure compounder with improving margins. That still supports a constructive long-term view on the franchise itself, but it argues for patience on entry. citeturn43search3turn40search3turn17view1turn42search14
**Micron** is the outlier. On trailing optics it still looks expensive — current sources show about **17.4x sales** and **27.3x EV/EBITDA** — but the forward-earnings lens is completely different, because the market is discounting a very large jump in memory profitability. The current **forward P/E of roughly 9.6x** is the lowest and most understandable valuation in the group once you accept the AI memory thesis. My fair-value range is therefore built from forward earnings rather than trailing sales: on a mid-cycle-to-super-cycle framework I think Micron is **closer to fair than overvalued**, and it is the easiest name here to justify fundamentally. citeturn44search0turn35view2turn36view2turn41news41
**Nokia** screens cheaper on sales than the U.S. names, but its re-rating has still been aggressive. Current market-data sources show around **3.9x sales**, **38.9x forward earnings**, and **29.9x EV/EBITDA**. For a business still carrying telco cyclicality and margin variability, that is not obviously cheap. The market is now paying for the AI optical/IP pivot and for Infinera-related strategic improvement. My fair-value framework ends up near the broad analyst-target range — roughly the low teens in ADR terms — not far above, which is why I see the shares as **strategically improved but valuation-stretched after the run**. citeturn41search5turn42search23turn42search11turn45view0
## Research-only investment views
**Marvell** is a high-quality strategic asset in AI networking, and I would be comfortable owning it over a **12–24 month** horizon — but only with valuation discipline. For research sizing, I would think in terms of a **2%–4% position** at most, accumulated on weakness rather than at today’s premium. My buy triggers would be: a material pull-back, evidence of another step-up in custom-silicon wins, and continued confirmation that optical/coherent products are broadening the data-centre moat. My trim triggers would be: hyperscaler digestion, export-control pressure on the China/Taiwan mix, or signs that custom-AI programmes are extending timelines. citeturn44search1turn23view1turn31view0
**Lumentum** is strategically well placed, but I would treat it as a **tracking position at most** until valuation becomes easier to defend. A sensible research allocation would be **0%–1.5%**, with the default choice being to wait. The purchase case would strengthen only if the company proves it can scale cloud transceivers, optical circuit switches and co-packaged-optics-related products while keeping margins in the improved range. My trim/avoid triggers are simpler: any sign that hyperscaler optical component demand is pausing, that cloud transceiver pricing is worsening, or that the systems mix becomes less favourable. citeturn19view0turn18view3turn43search1turn42search8
**Coherent** is one of the most interesting businesses in the set, because the technology base is broad and the datacom mix is improving, but I still prefer it as a **watch-list quality compounder rather than an aggressive entry today**. A **1%–2.5% research position** would be the upper end of what I would consider, and only after a better entry or more evidence that Datacenter & Communications can keep compounding above the industrial drag. I would buy more confidently if the company sustains revenue growth while delevering the income statement and simplifying the equity story. citeturn16view3turn16view4turn43search2turn42search13
**Corning** is the most acceptable “infrastructure derivative” of the group if you want AI-network exposure without taking direct silicon-cycle risk. The expected return is probably lower than Micron’s in a strong cycle, but so is the business-model volatility. For research purposes I would consider **1.5%–3%** on pull-backs, not on momentum spikes. The buy trigger is straightforward: proof that GenAI-related optical demand is translating into durable free cash flow and not just transient enterprise optical orders. The trim trigger is equally straightforward: if Optical Communications growth slows materially before the broader Springboard plan has delivered its margin targets, the market multiple could compress sharply. citeturn10search16turn40search15turn43search3turn42search14
**Micron** is my preferred name in the basket. It is still cyclical and therefore not “safe”, but it has the most attractive combination of **technology centrality, earnings leverage, and valuation support**. For research sizing I would be comfortable with **3%–5%** over a **12–24 month** horizon, subject to standard cyclicality discipline. My buy triggers are continued HBM/memory leadership, signed long-term agreements, and evidence that supply discipline is holding. My trim triggers are classic memory-cycle signals: aggressive industry capacity additions, price erosion, inventory rebuilds, or a weakening of data-centre demand intensity. citeturn44search0turn41news41turn38view3turn36view2
**Nokia** has improved strategically, but I do not think the post-rally risk/reward is as attractive as the narrative sounds. If I wanted exposure, I would keep it modest — **1%–2%** — and only on pull-backs closer to or below the current target-price cluster. The buy case depends on Optical and IP Networks sustaining AI-cloud momentum while the broader company avoids margin leakage. The trim case is that mobile/telecom softness reasserts itself before AI-cloud optical wins become large enough to dominate consolidated results. citeturn45view0turn16view6turn42search23turn42search11
## Timeline and representative IP appendix
The timeline below summarises the most relevant disclosed milestones for AI data-centre interconnects over the last three years and the next 12 months. The milestones are drawn from company filings, earnings releases, and official investor materials. citeturn11search4turn19view0turn13search6turn40search15turn38view3turn45view0
```mermaid
timeline
title Major AI-interconnect milestones
2023 : Lumentum adds Cloud Light to strengthen datacom modules
: Nokia begins integrating Infinera optical assets
2024 : Micron scales HBM and data-centre memory positioning
: Corning begins seeing stronger GenAI optical demand in enterprise
: Marvell acquires Polariton for 3.2T+ optical scaling
: Lumentum posts record Q3 FY2026 revenue on cloud/datacom demand
: Coherent Q3 FY2026 revenue reaches US$1.81bn
: Nokia Q1 2026 reports AI & Cloud design wins in pluggables and line systems
Next 12 months : Watch HBM4 / next-gen packaging at Micron
: Watch co-packaged optics and OCS commercialisation at Lumentum
: Watch 1.6T/3.2T coherent optical roadmaps at Marvell
: Watch DC optics attach-rate and industrial mix at Coherent
: Watch AI fibre/connectivity conversion at Corning
: Watch AI optical and IP growth versus mobile drag at Nokia
```
Because none of these companies publicly publish an issuer-certified ranking of their most valuable patent families, the IP list below should be read as **five representative patent-family themes/search links per company that are closely aligned with the AI interconnect thesis**, not as a legally definitive “top five by value” list. The underlying claim — that these firms have meaningful IP depth in the named areas — is supported by their annual reports and product disclosures. citeturn43search14turn21view1turn13search4turn40search6turn38view3turn15search1
A few parts of the request are constrained by what the companies explicitly disclose in public filings and what was reproducible from the accessible source set. The biggest limitations are these: some firms changed reporting structures mid-stream, so **last-four-quarter segment series are cleaner for Marvell, Lumentum, Micron and Coherent than for Corning and Nokia**; **quarterly geography is not always fully reproduced in the accessible summaries**, so where a complete quarterly geography bridge was not available I used the latest annual geography or marked it unspecified; and **no company publishes an official “top five patent families by economic value” list**, so the patent appendix is explicitly framed as representative thematic family coverage rather than issuer-ranked top five patents. Those limitations matter for precision, but they do **not** change the main conclusions on technology position, AI exposure, or relative valuation. citeturn23view1turn21view1turn38view2turn45view0turn40search3
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
hermes