Morgan Stanley: Meetings on ALAB, MRVL, and INTC were overall positive, with AI interconnect and Intel foundry as the main themes

Overview

This report is a North American semiconductor company meeting note published by Morgan Stanley on June 15, 2026, primarily covering Astera Labs Inc, Marvell Technology Group Ltd, and Intel Corporation. The report says the public meetings with the CEOs or management teams of the three companies were generally positive, focusing on how AI infrastructure is evolving from scale-out to scale-up and scale-across, and the resulting changes in demand for PCIe, CXL, UALink, NVLink Fusion, optical interconnect, custom chips, CPUs, and advanced packaging.

Core views

The core views include: first, ALAB has strong growth visibility in the near-term Amazon Trainium 3 scale-up ramp and long-term UALink/NVLink Fusion/optical interconnect opportunities, though valuation demands are higher after the stock’s rise; second, MRVL’s competitive edge comes from its full-stack interconnect and custom chip coverage, and its AI connectivity business still has high growth, but its valuation has become expensive relative to NVDA; third, under Lip-Bu Tan’s leadership, Intel is strengthening its CPU roadmap, x86 ecosystem, and foundry investment, and 14A, EMIB, and advanced packaging may provide option value, but market expectations for it to regain share versus AMD may be premature.

Analysis framework

The report is mainly based on public management meetings during the bus tour, summarizing company strategy statements, product roadmaps, customer demand, technology nodes, capacity planning, and valuation risks. The analysis focuses on management confidence, changes in the AI infrastructure interconnect value chain, the timing of key technology commercialization, and relative positioning versus competitors.

Methodology notes

• Rating systemMorgan Stanley relative rating system

  Overweight, Equal-weight, Not-Rated, Underweight

  Morgan Stanley uses a relative rating system rather than simple buy, hold, and sell; Overweight indicates expected risk-adjusted total return over the next 12-18 months above the average level of the analyst’s covered industry.

• Industry viewAttractive industry view

  North America Industry View Attractive

  Attractive indicates the analyst expects this covered industry to be attractive relative to the relevant broad market benchmark over the next 12-18 months.

• Valuation methodEV/Sales/Growth and EPS multiples

  Growth semiconductor valuation multiples

  The excerpt shows that ALAB is valued using 0.53x CY27 EV/sales/growth, about 21x EV/sales, assuming 40% revenue CAGR; INTC uses about 42x CY2027 EPS of $1.73; MRVL uses about 40x CY27e base case MW EPS of $4.98.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Astera Labs Inc (ALAB.O)

  A beneficiary tied to AI scale-up interconnect, PCIe/CXL, UALink, NVLink Fusion, NPO/CPO, and optical engine opportunities.

  Strengths

  Management is highly confident in both near-term and long-term opportunities; PCIe is expected to keep growing over the coming years, especially remaining important in add-in-card AI systems in China; the company has capabilities in copper links, Cosmos software, standardized product portfolios, electrical ICs, PICs, and aiXscale.

  Weaknesses

  CXL previously fell short of expectations, and UAL and optical interconnect ramps still require time; after the sharp stock increase, the investment expectation hurdle is high.

  Comparison

  Management described large PCIe switch products as the open-ecosystem “NVSwitch”; it also emphasized that unlike companies that “invest in everything,” ALAB is more focused on areas where it believes it can win.

  Risks

  Intensifying competition leading to share loss; a pause in AI and data center investment; meaningful delays in CXL servers or 1.6T port speeds; slower-than-expected NPO/CPO migration.

• Marvell Technology Group Ltd (MRVL.O)

  A name related to AI interconnect, custom silicon, scale-out/scale-up switching, optics, SerDes, CXL, NIC, and data-center connectivity.

  Strengths

  The company is positioned as one of the few suppliers spanning DCI, die-to-die IP, scale-out switching, scale-up switching, optics, SerDes, CXL, NIC, and custom chips; management believes customer preference for alternatives to AVGO benefits MRVL.

  Weaknesses

  The report believes its long-term and short-term confidence is clear and well-founded, but valuation is high, especially with its premium multiple versus NVDA not easy to digest.

  Comparison

  The report notes MRVL trades at about 2.5x NVDA’s multiple, while the company has not outgrown NVDA for a long time; compared with AVGO, customers may want to retain an alternative supplier in software, silicon, and optics integration scenarios.

  Risks

  AI opportunity smaller than expected; Storage and Networking weaker than expected; Enterprise DC and Networking continue to drag performance; supply-chain signals are over-extrapolated by the market.

• Intel Corporation (INTC.O)

  A name tied to CPU, x86 ecosystem, foundry, 14A, EMIB, and advanced packaging.

  Strengths

  New CEO Lip-Bu Tan is driving a shift in culture and growth execution; the company is placing more emphasis on the CPU roadmap, low-power competitiveness, Arm-related experience, x86 customization, and coordination between internal products and external foundry; 14A is viewed as a key foundry node.

  Weaknesses

  The analyst remains concerned about the product roadmap; Intel needs to prove that incremental capital spending can translate into customer wins, mass production, scale, and margin recovery; the company does not have an NVDA-style balance-sheet advantage in securing DRAM resources.

  Comparison

  Relative to AMD, the report believes market expectations for Intel to regain server and desktop share may be premature, because AMD says it can secure more wafers for its leading Venice product; relative to TSMC, Intel hopes to maintain external manufacturing options and keep the 14A timeline broadly comparable.

  Risks

  Intensifying AMD competition causing continued processor share loss and ASP pressure; limited foundry success leading to an inflated cost structure; 14A yield, PDK, customer onboarding, EMIB reliability, and advanced packaging scale-up execution falling short of expectations.

Key data

• Report date2026-06-15 10:19 AM GMTTime disclosed on the report’s front page.
• Covered companiesALAB.O; MRVL.O; INTC.OThe meeting notes focus on Astera Labs, Marvell, and Intel.
• Industry viewNorth America Industry View AttractiveThe North America semiconductors industry view is Attractive.
• ALAB content value per XPUcurrently about $1,000, and expected to continue growingManagement believes increasing interconnect intensity in AI infrastructure will drive higher content value.
• UAL commercialization timingfirst deployment expected next year, with larger-scale ramp in 2028ALAB management believes the UAL switching opportunity will not be winner-take-all.
• CXL revenue timingmanagement expects revenue generation in 2027Inference KV-cache offload, server memory life extension, and DDR4 reuse are demand drivers.
• NPO/CPO timingNPO is expected to be deployed first in 2027, while scale-up CPO is more likely at least by 2028If NPO proves effective, it could delay adoption of CPO by some customers.
• ALAB long-term TAM statementabout $10bn UAL TAM; about $10bn PCIe TAM and about $10bn Ethernet TAM by 2030; broader market > $100bnManagement did not quantify the specific size of NVL and ICI within the broader market.
• MRVL business model statementcustom rising from about $4bn to $10bn; legacy about $2.5bn and growing with GDP; connectivity about $10bn and growing about 70%Management said there has been no recent slowdown in connectivity.
• Intel 14A node metricsabout 0.5 defect density, about 40% yield, with a target of reaching 0.1-0.2 defect density by Q1 next year14A 0.5 PDK has been released, 0.9 is expected in October, risk production around 2028, and mass production around 2029.
• Intel capacity and packagingOregon and Arizona can still expand, Ohio will accelerate, Germany has been shut downManagement believes Intel could face insufficient capacity if demand continues to improve; EMIB was called a “secret weapon” but still has complexity and reliability issues.

Impact & implications

For investors, the report reinforces a multi-year chain of AI infrastructure capital spending expanding from GPU compute into interconnect, switching, optics, CXL, CPUs, and advanced packaging. ALAB and MRVL are more direct beneficiaries of rising AI interconnect value, but valuations already reflect considerable optimism; Intel has option value in foundry, advanced packaging, and CPU demand recovery, but the investment case is highly dependent on 14A execution, customer wins, returns on capacity investment, and the competitive outcome versus AMD/TSMC.

Risks

• After ALAB’s strong share-price increase, expectations are elevated; if AI spending, CXL, 1.6T, NPO/CPO, or UAL deployments are delayed, valuation could come under pressure.
• Although MRVL has a strong AI interconnect narrative, the report explicitly points out that it trades at about 2.5x NVDA’s multiple, and the company has not consistently outgrown NVDA recently.
• Uncertainty remains around Intel’s CPU roadmap and foundry execution, and market expectations for it to regain share from AMD may be premature.
• Intel’s 14A, EMIB, advanced packaging, and added capex need to be validated through customer wins, mass production, and margin recovery, otherwise they may create a high-cost structure.
• The AI infrastructure supply chain still faces risks from memory shortages, advanced packaging bottlenecks, substrate constraints, misread supply-chain signals, and shifts in the competitive landscape.
• Morgan Stanley discloses that it has or seeks investment-banking and other service relationships with several covered companies, and investors should treat this research as only one factor in decision-making.

What to watch

• ALAB’s actual content value and share performance in the second-half scale-up ramp of Amazon Trainium 3.
• The deployment pace of UALink, NVLink Fusion, NPO, and CPO in 2027-2028, as well as changes in per-link value as copper links migrate to optical links.
• Continued growth of PCIe in China AI systems and open ecosystems, and CXL revenue realization driven by KV-cache offload and memory pooling.
• Whether MRVL’s custom business can advance from about $4bn toward about $10bn, and whether connectivity growth of about 70% can continue.
• Customer selection and share for MRVL in merchant scale-up optics, scale-up switching, UAL, ESUN, and NVL Fusion.
• Intel 14A defect density, yield, PDK 0.9 release timing, internal test chip in 2027, risk production in 2028, and mass production progress in 2029.
• Whether Intel can stop CPU share losses within about nine months, and the impact of AMD Venice wafer supply on the competitive landscape.
• Intel’s Ohio capacity expansion, Oregon/Arizona expansion, TerraFab potential, EMIB reliability, and advanced-packaging customer wins.