$KEYS KEY READ-THROUGHS FROM KEYSIGHT TECHNOLOGIES Q2 FY2026 EARNINGS CALL
Source material: Keysight Technologies Q2 FY2026 earnings call transcript.
Keysight’s Q2 FY2026 call is a high-signal cross-sector read-through because the company sits at the validation, emulation, manufacturing test, and systems assurance layer for AI data center infrastructure, high-speed Ethernet and optical networking, advanced semiconductor manufacturing, aerospace/defense electronics, high-performance PCBs, non-terrestrial networks, 6G research, software-defined vehicles, and quantum research. The quarter was not merely a company-specific beat. Q2 orders were $2.051B, up 56% reported and 48% core; revenue was up 31% reported and 24% core; ex-tariff revenue was up 35%; and FY2026 revenue expectations were raised to high-20s % growth. The highest-conviction broader-market message is that AI infrastructure demand is accelerating, broadening across the supply chain, and increasingly constrained by validation complexity and product-ramp execution rather than by end-demand weakness. The second major message is that defense, semiconductor capital intensity, and high-performance electronics are providing additional legs to the cycle, reducing the probability that AI optical/networking demand is an isolated or short-lived inventory event. The most important negative read-throughs are that hyperscaler capex intensity remains structurally elevated, proprietary networking moats face pressure from open Ethernet and heterogeneous architectures, legacy pluggable-optics models face longer-duration architectural risk from silicon photonics and co-packaged optics, and AI infrastructure suppliers may face near-term shipment timing risk from NPI ramp and validation capacity constraints.
AI DATA CENTER NETWORKING, OPTICS, AND ACCELERATORS
AI OPTICAL AND NETWORKING DEMAND IS ACCELERATING AND DOES NOT LOOK LIKE A SIMPLE PULL-FORWARD (READ-THROUGH 1)
Call support: Keysight said “wireline delivered record orders again this quarter with robust demand for both R&D and manufacturing solutions.” Management also disclosed that the AI-related business finished H1 FY2026 in the “$500 million to $600 million range,” almost equal to all of FY2025, and stated that there were “no pull forwards” discernible from the data. The demand signal was broad across the AI stack, with Keysight servicing computing, networking, transceivers, interconnects, and hyperscalers across R&D, early design, validation, conformance, compliance testing, emulation, and cluster deployment.
Affected companies and impact: Broadcom (AVGO: US) positive, high magnitude; Marvell Technology (MRVL: US) positive, high magnitude; Arista Networks (ANET: US) positive, high magnitude; NVIDIA (NVDA: US) positive, moderate-to-high magnitude; Coherent (COHR: US) positive, high magnitude; Lumentum (LITE: US) positive, moderate-to-high magnitude; Fabrinet (FN: US) positive, high magnitude; Zhongji Innolight (300308: China) positive, high magnitude; Eoptolink (300502: China) positive, high magnitude; Accelink Technologies (000988: China) positive, moderate-to-high magnitude.
Transmission mechanism: Keysight’s order acceleration is a leading indicator for customers preparing to validate, qualify, and manufacture next-generation AI networking systems. Higher test and validation demand generally precedes or accompanies production ramps in switch silicon, optical modules, transceiver manufacturing, high-speed interconnects, and AI cluster deployment. The fact that both R&D and manufacturing demand are strong reduces the risk that the cycle is merely a near-term inventory rebuild; it indicates simultaneous current production scaling and next-generation product development.
Timing distinction: The near-term trading catalyst is positive for AI networking and optical-exposure equities because Keysight’s bookings signal supports continued 800G deployment and 1.6T adoption. The longer-duration fundamental shift is that AI infrastructure test intensity appears to be structurally rising as cluster scale, vendor heterogeneity, speed transitions, and post-deployment failure costs increase.
OPEN ETHERNET AND HETEROGENEOUS AI FABRICS ARE GAINING SHARE OF MIND VERSUS CLOSED, PROPRIETARY NETWORKING ARCHITECTURES (READ-THROUGH 2)
Call support: Keysight highlighted “continued 800 gig deployments, accelerating adoption of 1.6 terabit architectures, and increased R&D activity around 3.2 terabit technologies.” The company also said it collaborated with Broadcom on the “first public interoperability demonstration” of Ultra Ethernet Consortium specifications. In Q&A, management said AI infrastructure is becoming more heterogeneous and that the relevant architecture debate is no longer optical versus electrical, open versus closed, or pluggable versus integrated; the answer is increasingly “both.”
Affected companies and impact: Broadcom (AVGO: US) positive, high magnitude; Arista Networks (ANET: US) positive, high magnitude; Marvell Technology (MRVL: US) positive, high magnitude; Credo Technology (CRDO: US) positive, high magnitude; Astera Labs (ALAB: US) positive, moderate-to-high magnitude; AMD (AMD: US) positive, moderate magnitude; NVIDIA (NVDA: US) mixed, with positive demand for GPU/networking scale but moderate negative implications for the durability of proprietary networking lock-in.
Transmission mechanism: Open Ethernet, UEC-aligned fabrics, and heterogeneous accelerator architectures raise the importance of interoperability testing, conformance, system emulation, and merchant-silicon ecosystem validation. This directly benefits switch silicon, Ethernet switching, retimers, active electrical cables, DSPs, optical interconnect silicon, and multi-vendor AI fabric vendors. It also lowers the relative advantage of closed single-vendor networking stacks, even though NVIDIA remains a major beneficiary of overall AI cluster growth.
Timing distinction: The near-term catalyst is sentiment-positive for Ethernet AI fabric suppliers and high-speed connectivity names because Keysight’s comments validate production-readiness progress around 1.6T and UEC. The longer-duration shift is that AI infrastructure may evolve toward a more multi-vendor fabric model, supporting broader wallet share for merchant silicon and Ethernet-system vendors while diluting the scarcity premium attached to proprietary interconnect ecosystems.
HYPERSCALER AI CAPEX INTENSITY IS CONFIRMED AS DURABLE, BUT THE READ-THROUGH IS MIXED FOR HYPERSCALER EQUITIES (READ-THROUGH 3)
Call support: Keysight said hyperscalers are adopting AI workload emulation solutions to improve GPU and power utilization while dealing with system and security complexity. Management also stated that “a few hundred gigawatts of capacity” is expected to come online through 2030 and that “multi-hundred billion” spend commitments take time to move through the supply chain and into data centers.
Affected companies and impact: Microsoft (MSFT: US) mixed, moderate magnitude; Amazon (AMZN: US) mixed, moderate magnitude; Alphabet (GOOGL: US) mixed, moderate magnitude; Meta Platforms (META: US) mixed, moderate magnitude; Oracle (ORCL: US) mixed-to-positive, moderate-to-high magnitude; NVIDIA (NVDA: US), Broadcom (AVGO: US), Arista Networks (ANET: US), Vertiv (VRT: US), Eaton (ETN: US), Schneider Electric (SU: France), and Quanta Computer (2382: Taiwan) positive, high magnitude as suppliers.
Transmission mechanism: Keysight’s demand confirms that hyperscalers are not simply announcing AI capex; they are actively validating and deploying complex AI clusters that require workload emulation, interoperability testing, security validation, and utilization optimization. For hyperscalers, this supports long-term AI capability but reinforces elevated capital intensity and potential free cash flow pressure. For suppliers, it validates a multi-year infrastructure spend funnel across networking, power, thermal, servers, accelerators, optical, and test.
Timing distinction: The near-term equity read-through is positive for AI infrastructure suppliers and mixed for hyperscalers, particularly where investors are sensitive to free cash flow conversion and capex discipline. The longer-duration shift is that AI infrastructure buildout remains in an early-to-mid deployment phase, with demand flowing through multiple layers of the supply chain over several years rather than peaking in a single procurement wave.
AI INFRASTRUCTURE SUPPLIERS FACE EXECUTION AND RAMP BOTTLENECK RISK EVEN THOUGH END-DEMAND REMAINS STRONG (READ-THROUGH 4)
Call support: Management said Q3 revenue conversion is influenced by “the mix and some timings of some new product introductions and how quickly we can ramp them.” Keysight also said it has higher AI backlog, a strong pipeline of systems wins, and “unprecedented ramp” requirements for AI-enabling NPIs. Capex guidance was raised from $160M to $200M, with most of the incremental investment directed toward supporting that ramp.
Affected companies and impact: Coherent (COHR: US) mixed, moderate magnitude; Lumentum (LITE: US) mixed, moderate magnitude; Fabrinet (FN: US) mixed, moderate magnitude; Zhongji Innolight (300308: China) mixed, moderate magnitude; Eoptolink (300502: China) mixed, moderate magnitude; Broadcom (AVGO: US) mixed-to-positive, moderate magnitude; Marvell Technology (MRVL: US) mixed-to-positive, moderate magnitude; Arista Networks (ANET: US) mixed-to-positive, moderate magnitude.
Transmission mechanism: Demand for AI infrastructure hardware is strong, but validation, manufacturing test capacity, and NPI ramp speed are increasingly important constraints. If suppliers cannot secure enough validation tools, test capacity, optical characterization equipment, or production-readiness infrastructure, revenue recognition may lag orders. This is not a negative demand signal; it is a potential shipment-timing and execution-risk signal.
Timing distinction: Near-term, this creates risk of quarterly volatility for optical, transceiver, networking, and high-speed interconnect suppliers even when end-demand remains strong. Longer term, it is positive for suppliers with stronger process engineering, test integration, and manufacturing scale because high validation complexity can become a barrier to entry.
SEMICONDUCTORS, SEMICAP, FOUNDRY, MEMORY, AND PHOTONICS
ADVANCED NODE, MEMORY, LITHOGRAPHY, AND SILICON PHOTONICS CAPEX ARE BROADENING THE AI SEMICAP CYCLE (READ-THROUGH 5)
Call support: Keysight said the semiconductor market saw “continued momentum” as the industry races to “scale capacity through 2030.” Management specifically cited AI ecosystem demand accelerating across “advanced node, memory, and silicon photonics,” wafer test wins across Asia, the US, and Europe, and strong growth in solutions for key lithography customers.
Affected companies and impact: ASML Holding (ASML: Netherlands) positive, high magnitude; KLA (KLAC: US) positive, high magnitude; Applied Materials (AMAT: US) positive, moderate-to-high magnitude; Lam Research (LRCX: US) positive, moderate-to-high magnitude; Tokyo Electron (8035: Japan) positive, moderate-to-high magnitude; Advantest (6857: Japan) positive, moderate magnitude; Teradyne (TER: US) positive, moderate magnitude; Taiwan Semiconductor Manufacturing (2330: Taiwan) positive, high magnitude; Samsung Electronics (005930: South Korea) positive, moderate-to-high magnitude; SK hynix (000660: South Korea) positive, high magnitude; Micron Technology (MU: US) positive, high magnitude.
Transmission mechanism: Keysight’s semiconductor test and validation demand provides evidence that AI-related semiconductor investment is not confined to GPUs or HBM. The demand is extending into advanced nodes, wafer-level testing, silicon photonics, lithography-adjacent workflows, advanced packaging, and memory capacity. For semicap names, this supports a broader and longer WFE cycle. For foundries and memory suppliers, it confirms that customers are preparing for multi-year AI compute, memory bandwidth, and photonics requirements.
Timing distinction: Near-term, the read-through is positive for semicap order expectations and memory-capex sentiment. Longer duration, the more important implication is that AI compute scaling through 2030 requires a sustained semiconductor manufacturing and inspection ecosystem buildout, not just a one-time accelerator procurement cycle.
SILICON PHOTONICS AND CO-PACKAGED OPTICS CREATE A LONG-DURATION MIX SHIFT, WITH POSITIVE READ-THROUGHS FOR INTEGRATED PHOTONICS BUT STRUCTURAL RISK FOR PLUGGABLE-ONLY MODELS (READ-THROUGH 6)
Call support: Management said “activity in silicon photonics and co-packaged optics continues to expand” and that early co-packaged optics engagements position Keysight well as the industry transitions to these architectures. Keysight also cited strong demand from next-generation optical component and transceiver development and deployment, plus new tools for advanced transceiver, photonics, chiplet, and 3D stack design.
Affected companies and impact: Broadcom (AVGO: US) positive, high magnitude; Taiwan Semiconductor Manufacturing (2330: Taiwan) positive, moderate-to-high magnitude; Intel (INTC: US) positive, moderate magnitude; Marvell Technology (MRVL: US) positive, moderate-to-high magnitude; Coherent (COHR: US) mixed, moderate-to-high magnitude; Lumentum (LITE: US) mixed, moderate magnitude; Fabrinet (FN: US) mixed, moderate magnitude; Zhongji Innolight (300308: China) mixed, moderate-to-high magnitude; Eoptolink (300502: China) mixed, moderate-to-high magnitude.
Transmission mechanism: Near-term, the same transition is positive for optical component and transceiver suppliers because more bandwidth, higher speeds, and more complex modules require more development, qualification, and production. Longer term, co-packaged optics and silicon photonics can shift value toward integrated optical engines, advanced packaging, and silicon-platform vendors, potentially pressuring suppliers overly dependent on discrete pluggable form factors if architectural migration accelerates.
Timing distinction: The near-term catalyst remains positive for optical supply chain demand at 800G and 1.6T. The longer-duration fundamental shift is mixed: suppliers with integrated photonics, advanced packaging, and co-design capabilities benefit, while pluggable-only suppliers face share and margin risk as architectures evolve.
EDA AND DESIGN-SOFTWARE COMPLEXITY ARE STRUCTURALLY SUPPORTED BY CHIPLET, PHOTONICS, AND 3D ARCHITECTURES (READ-THROUGH 7)
Call support: Keysight referenced chiplet and photonic design solutions, a new 3D interconnect designer, and the growing complexity of next-generation 3D stack chip architectures. Management also emphasized that leading foundry collaborations from R&D to production are enabling faster development and commercial ramp timelines for complex chip architectures and packaging.
Affected companies and impact: Synopsys (SNPS: US) positive, moderate-to-high magnitude; Cadence Design Systems (CDNS: US) positive, moderate-to-high magnitude; Siemens (SIE: Germany) positive, moderate magnitude; Keysight Technologies (KEYS: US) positive, high magnitude.
Transmission mechanism: As AI chips, optical interconnects, chiplets, advanced packaging, and 3D stacks become harder to design and validate, the design workflow requires tighter integration between simulation, EDA, measurement, validation, and production test. This increases demand for premium EDA, multiphysics simulation, signal integrity, electromagnetic analysis, photonics design, and hardware/software co-validation. The read-through is especially positive for vendors with toolchains spanning pre-silicon design and post-silicon validation.
Timing distinction: Near-term, this supports resilient EDA demand despite broader semiconductor cyclicality. Long term, it reinforces the structural pricing power and mission-critical status of advanced design-software vendors as physical design, photonics, and packaging complexity increase.
ELECTRONICS MANUFACTURING, PCBS, AND SUPPLY CHAIN GEOGRAPHY
AI SPENDING IS SPREADING INTO HIGH-PERFORMANCE PCBS AND ADVANCED ELECTRONICS MANUFACTURING (READ-THROUGH 8)
Call support: Keysight said general electronics delivered double-digit order and revenue growth driven by AI-related innovation and infrastructure investment. Management specifically said “customer capacity investment for high-performance PCBs was again strong” and that greater complexity, higher-density interconnects, multilayer architectures, and higher speeds are driving higher PCB test intensity.
Affected companies and impact: TTM Technologies (TTMI: US) positive, high magnitude; Celestica (CLS: Canada) positive, high magnitude; Jabil (JBL: US) positive, moderate-to-high magnitude; Sanmina (SANM: US) positive, moderate magnitude; Flex (FLEX: US) positive, moderate magnitude; Unimicron Technology (3037: Taiwan) positive, high magnitude; Compeq Manufacturing (2313: Taiwan) positive, moderate-to-high magnitude; Tripod Technology (3044: Taiwan) positive, moderate-to-high magnitude; Shennan Circuit (002916: China) positive, moderate-to-high magnitude.
Transmission mechanism: AI data center buildouts require not only GPUs and optics but also high-layer-count PCBs, dense backplanes, high-speed signal integrity, advanced substrates, and production test. Strong PCB capacity investment implies AI infrastructure spending is moving deeper into electronics manufacturing and board-level infrastructure. Companies with high-speed networking, AI server, switching, and advanced PCB exposure should benefit from mix upgrade and capacity utilization.
Timing distinction: Near-term, this is positive for PCB and EMS names with AI server or networking exposure because capacity investment and test demand point to stronger orders. Longer duration, the shift toward denser, faster, multilayer PCB architectures raises structural barriers to entry and can support higher margins for technically capable suppliers.
SOUTHEAST ASIA SUPPLY CHAIN RECONFIGURATION IS A POSITIVE GEOGRAPHIC SIGNAL FOR DIVERSIFIED EMS/ODM PLAYERS AND A NEGATIVE SIGNAL FOR CHINA-CONCENTRATED LOW-END ELECTRONICS EXPOSURE (READ-THROUGH 9)
Call support: Keysight said supply chain rebalancing and reshoring were among the global dynamics identified earlier, and that go-to-market investments enabled the company to grow its Southeast Asia business “significantly as the supply chains get reconfigured.”
Affected companies and impact: Flex (FLEX: US) positive, moderate-to-high magnitude; Jabil (JBL: US) positive, moderate-to-high magnitude; Celestica (CLS: Canada) positive, moderate magnitude; Hon Hai Precision/Foxconn (2317: Taiwan) positive, moderate magnitude; Quanta Computer (2382: Taiwan) positive, moderate magnitude; Wistron (3231: Taiwan) positive, moderate magnitude; Venture Corporation (V03: Singapore) positive, moderate magnitude. China-concentrated lower-end EMS and electronics manufacturers are negative, moderate magnitude, where diversification is limited and customers require geopolitical redundancy.
Transmission mechanism: Test and measurement equipment demand in Southeast Asia is a practical signal that manufacturing capacity, validation labs, and electronics production infrastructure are shifting geographically. The beneficiaries are companies with established or expanding Malaysia, Vietnam, Thailand, Singapore, and broader Southeast Asia footprints. The losers are suppliers dependent on single-country manufacturing footprints, especially where customers are pushing redundancy, tariff mitigation, export-control resilience, and regionalized capacity.
Timing distinction: Near-term, this supports capex and utilization for diversified EMS and ODM suppliers. Longer term, it suggests Southeast Asia is becoming a more permanent node for AI infrastructure, electronics manufacturing, and semiconductor-adjacent supply chains rather than merely a temporary tariff workaround.
AEROSPACE, DEFENSE, RADAR, ELECTRONIC WARFARE, AND SPACE
RADAR, ELECTROMAGNETIC SPECTRUM OPERATIONS, AND EUROPEAN DEFENSE MODERNIZATION REMAIN HIGH-CONVICTION POSITIVE READ-THROUGHS (READ-THROUGH 10)
Call support: Keysight said aerospace, defense, and government demand showed “broad-based global momentum led by Europe,” supported by strength in the Americas. Demand was strongest in radar and electromagnetic spectrum operations, with governments and prime contractors expanding capacity to support evolving operational requirements. Management also cited “budget stability both in US and in Europe” and a key U.S. Air Force win for next-generation operational flight line testing.
Affected companies and impact: L3Harris Technologies (LHX: US) positive, high magnitude; RTX (RTX: US) positive, high magnitude; Northrop Grumman (NOC: US) positive, moderate-to-high magnitude; Lockheed Martin (LMT: US) positive, moderate magnitude; BAE Systems (BA.: United Kingdom) positive, high magnitude; Thales (HO: France) positive, high magnitude; Leonardo (LDO: Italy) positive, high magnitude; Saab (SAAB B: Sweden) positive, high magnitude; Rheinmetall (RHM: Germany) positive, moderate magnitude.
Transmission mechanism: Keysight’s strength in radar target generation, high-fidelity emulation, signal analysis, PNT, RF validation, and mission-readiness services indicates that defense customers are investing in the enabling infrastructure required before and during deployment of advanced radar, EW, autonomous, and spectrum operations systems. This is an upstream validation signal for defense primes and mission-electronics suppliers. Higher services attach rates also point to more sustained lifecycle spending rather than only one-time equipment procurement.
Timing distinction: Near-term, this is positive for European defense and U.S. mission-electronics sentiment because bookings strength is tied to active program spend and budget stability. Longer term, the fundamental shift is toward increasingly software-defined, spectrum-intensive, contested-environment defense architectures that require continuous simulation, test, validation, and modernization.
LEO, NON-TERRESTRIAL NETWORKS, AND DIRECT-TO-CELL ARE BECOMING A REAL ECOSYSTEM, BUT CURRENT REVENUE SCALE REMAINS SMALL (READ-THROUGH 11)
Call support: Keysight said NTN activity contributed to robust wireless orders and that direct-to-cell deployments are targeted “in the next few quarters.” Management named Amazon, Globalstar, SpaceX, and AST SpaceMobile as part of a widening ecosystem, while also saying NTN remains sub-1% of Keysight’s total revenue today.
Affected companies and impact: AST SpaceMobile (ASTS: US) positive, high magnitude due to pure-play sensitivity; Globalstar (GSAT: US) positive, high magnitude due to direct satellite-to-device sensitivity; Amazon (AMZN: US) positive, low-to-moderate magnitude through Kuiper optionality; Qualcomm (QCOM: US) positive, moderate magnitude through modem/RF/device ecosystem exposure; Samsung Electronics (005930: South Korea) positive, moderate magnitude through devices and AI-RAN/NTN workflows; Iridium Communications (IRDM: US) mixed-to-positive, moderate magnitude; Apple (AAPL: US) positive, low magnitude through satellite-device feature optionality.
Transmission mechanism: Increasing demand for orbit emulation, PNT, RF validation, device emulation, network emulation, and core-network validation indicates that satellite-to-device and LEO constellation architectures are moving from concept to deployment preparation. This supports companies exposed to satellite connectivity, direct-to-cell devices, RF front-end integration, modem standards, constellation deployment, and ground/network validation. The call also limits the magnitude: sub-1% of Keysight revenue means the ecosystem is real but still early.
Timing distinction: Near-term, the read-through is most actionable for high-beta satellite connectivity equities and select device/RF suppliers because deployments are approaching. Longer term, it supports a broader direct-to-device satellite communications market, but commercialization timing, spectrum, handset integration, constellation economics, and regulatory execution remain key risk factors.
