Orbital compute opens a new theme for European semiconductors, while memory momentum enters a peak in the rate of change
AI summary card
Orbital compute opens a new theme for European semiconductors, while memory momentum enters a peak in the rate of change
Morgan Stanley believes orbital compute may gradually evolve from near-orbit edge AI to space AI infrastructure and prefers STMicroelectronics and Infineon; however, it remains more cautious on the relative performance of the memory segment.
- Orbital compute is modeled in three phases: short-term as in-orbit edge AI, mid-term as orbital cloud/distributed compute, and a long-term bull-case as space AI infrastructure.
- Orbital compute capex/W is modeled to fall from around US$60/W in 2030 to US$9/W by 2040, mainly driven by reusable launch, lower satellite hardware costs, and improved payload economics.
- STMicroelectronics is seen as a beneficiary of the European semiconductor space theme, with a price target of €78 and roughly 23% upside versus the July 6, 2026 close.
- In memory, DRAM and NAND prices are still rising year over year, but the rate of change may have peaked in 4Q26; DRAM earnings revision breadth is about 89%, close to historical highs.
Report interpretation
Overview
This report is from Morgan Stanley’s global technology webcast and covers SpaceX, orbital compute, the memory cycle, and related European semiconductor investment opportunities. The report defines orbital compute as a new frontier of AI infrastructure: satellites can process images, sensor data, and inference workloads on orbit, and only useful outputs are sent back to the ground. Over time, many satellites can form a distributed AI layer resembling a "space data center" through high-capacity optical links.
Core views
The key views are threefold: first, orbital compute may become a structural theme for European semiconductors, with STMicroelectronics and Infineon viewed as key enablers. Second, the economics of orbital compute are expected to improve as launch and satellite hardware costs decline, but radiation, thermal management, on-orbit maintenance, optical-network bandwidth, orbital debris, and security remain key constraints. Third, the memory cycle is not over, but the pace of price increase and earnings revisions may have peaked; relative performance of the segment should be treated more cautiously. The report favors DRAM and traditional memory over NAND and memory module makers.
Analysis framework
The report uses a combination of thematic investing and value-chain mapping: it first breaks down the phased evolution path of orbital compute, then analyzes technical components such as capex/W reduction, satellite compute stack, optical interconnect, power, thermal management, memory, and orchestration software, and finally maps those to European semiconductor names. For STMicroelectronics, valuation is done using a sum-of-the-parts framework, assigning different P/E multiples to core business versus optical and LEO businesses.
Methodology notes
Sum-of-the-parts valuation
The €78 target price for STMicroelectronics is based on sum-of-the-parts valuation: core business is valued at FY28e EPS of €2.11 on an 18x P/E multiple, while optical and LEO are valued at FY28e EPS of €1.34 on a 36x P/E multiple, with a 10% WACC discount rate.
Orbital compute technology stack mapping
The report decomposes orbital compute into layers such as AI compute payloads, server/payload modules, power systems, thermal systems, orbital and constellation management, optical networking, and memory and orchestration software to identify semiconductor and systems-supply-chain opportunities.
Earnings revision breadth
The report uses DRAM earnings revision breadth at around 89%, close to historical peak levels, to judge that memory momentum may be near a phase high and relative performance may weaken.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- STMicroelectronics NV (STMPA.PA)A beneficiary of the European semiconductor space theme, with an Overweight rating and a €78 target price.
- Strengths
- It can benefit from growth in optical and LEO revenues, SiC adoption, increasing design wins in automotive and industrial designs, and semiconductor demand related to space/orbital compute.
- Weaknesses
- The core business is still sensitive to automotive and industrial cycles, MCU pricing, and inventory normalization.
- Comparison
- The report prefers STMicroelectronics and Infineon as key enablers in the European orbital compute technology stack.
- Risks
- Weak global macro, weaker OEM demand in automotive and industrial sectors, softness in MCU pricing, DRAM shortages affecting automotive sales, and declines in critical smartphone sensor content or design-socket losses.
- Infineon Technologies AG (IFXGn.DE)Listed as one of the key enablers in the European orbital compute technology stack.
- Strengths
- Its role in power management, and in power and industrial semiconductor segments, may be relevant to orbital compute power-supply and system-control requirements.
- Weaknesses
- The report does not provide detailed company-level valuation decomposition.
- Comparison
- Preferred together with STMicroelectronics, but detailed valuation coverage in this report focuses mainly on STMicroelectronics.
- Risks
- Industry cycle, automotive and industrial demand, semiconductor cyclicality, and potential conflict-of-interest risks in related disclosures.
- DRAMOne of the relatively preferred names within the memory cycle.
- Strengths
- Demand remains robust under AI capex trends, and prices are still rising year over year.
- Weaknesses
- Earnings revision breadth is around 89%, near historical highs, and the pace may have peaked.
- Comparison
- The report favors DRAM and traditional memory over NAND and memory module makers.
- Risks
- The year-on-year price increase in 4Q26 may slow, and relative outperformance in the segment may weaken.
- NANDA relatively less preferred direction within the memory segment compared with DRAM.
- Strengths
- Prices are still increasing.
- Weaknesses
- The report believes investment flow and bottlenecks are more concentrated in DRAM and traditional memory.
- Comparison
- Assigned lower preference than DRAM.
- Risks
- Relative performance risk may be higher once the rate of change has peaked.
- Memory module makersListed as the least preferred memory-related asset.
- Strengths
- Raw materials did not provide clear advantages.
- Weaknesses
- At a point when memory momentum may be fading and earnings revisions are near historical peaks, relative attractiveness is lower.
- Comparison
- Lower than DRAM, traditional memory, and NAND.
- Risks
- Corporate earnings pressure and a slowdown in memory price momentum may weigh on shares.
Key data
- Report date2026-07-07The cover shows July 7, 2026 10:05AM GMT.
- STMicroelectronics target price€78Based on sum-of-the-parts valuation, implying roughly 23% upside versus the July 6, 2026 close.
- STMicroelectronics current price€63.36The disclosure table shows STMPA.PA price as of 07/06/2026.
- Orbital compute capex/W modelabout US$60/W in FY30 down to US$9/W by FY40Driven by reusable launch, reduced satellite hardware costs, and improved payload economics.
- LEO sales growthFY26 to FY28 CAGR 48%The report says this is mainly driven by end-customer terminal content.
- LEO sales upside scenarioover $3bn cumulative in FY26-FY28Listed as one of STMicroelectronics’s upside risks.
- DRAM earnings revision breadtharound 89%Near historical peak, signaling relative performance risk in memory.
- Industry viewEurope Industry View In-LineThe cover shows an In-Line view for European industry.
Impact & implications
From an investment perspective, orbital compute offers a long-duration and semiconductor value-chain-rich AI infrastructure theme, supporting European suppliers with capabilities in power management, MCU, sensors, optics, automotive/industrial, and space-related technology. At the same time, the memory segment remains supported by structural AI capex tailwinds, but short-term stock performance may shift from company commentary to earnings, and once the rate of change peaks, downside risk from chasing higher valuations increases.
Risks
- Orbital compute requires radiation-hardened chips, ECC, and shielding, with high hardware-reliability demands.
- There is no air or liquid cooling in space; heat must be removed through heat pipes and radiators, making thermal management a key engineering constraint.
- On-orbit maintenance is difficult, while AI hardware refresh cycles are short, raising repair and upgrade risks.
- AI-grade distributed computing requires inter-satellite optical-link bandwidth far above current satellite network levels.
- LEO satellites move at high speed, have limited resources, and have continuously changing links, which puts pressure on scheduling, routing, and dynamic resource management.
- Orbital debris, collision avoidance, cybersecurity, and physical hardening will become core operational requirements.
- Memory prices are still rising, but rate of change is peaking, and relative underperformance may appear after DRAM earnings revision breadth reaches near historical highs.
- Downside risks for STMicroelectronics include weak global macro conditions, weaker automotive and industrial OEM sales, softness in MCU prices, DRAM shortages impacting car sales, and declines in critical smartphone sensor content or design-socket losses.
What to watch
- Whether orbital compute capex/W declines from around US$60/W in 2030 toward US$9/W by 2040 as modeled.
- The pace of improvements in reusable launch costs, satellite hardware costs, and payload economics.
- Whether LEO sales can achieve around 48% CAGR from FY26 to FY28 and exceed $3bn cumulatively.
- Whether optical business sees unexpected expansion in 2H26/FY27.
- STMicroelectronics design-win momentum, SiC adoption, and sensor-content changes in automotive and industrial areas.
- Whether year-over-year price growth in DRAM and NAND slows in 4Q26 as the report expects.
- Whether DRAM earnings-revision breadth retreats from around 89%.
- The impact of hyperscaler capex and profit pressure on AI semiconductor and memory equities.