AI chips stocks

NVIDIA is a primary accelerator supplier whose revenue is directly linked to AI data-center deployment.

NVIDIA is a direct AI accelerator designer and a major consumer of the advanced design infrastructure whose demand Synopsys highlighted.

NVIDIA is a direct AI-accelerator supplier whose products consume TSMC advanced-node wafers and CoWoS packaging.

NVIDIA Corporation stands as a prominent provider of advanced graphics, computational, and networking solutions, operating across the United States, Taiwan, China, and numerous international markets. Its Graphics division encompasses GeForce GPUs, central to PC gaming and personal computing experiences, along with the GeForce NOW cloud gaming service and its supporting infrastructure, as well as dedicated solutions for various gaming platforms. For professional visualization, it provides Quadro and NVIDIA RTX GPUs for enterprise workstations, further offering vGPU software designed for cloud-centric visual and virtual computing, automotive platforms for in-vehicle infotainment, and the Omniverse software suite, facilitating 3D design and virtual world creation. The Compute & Networking segment is a cornerstone for AI, high-performance computing (HPC), and accelerated data center platforms. It integrates Mellanox networking and interconnect solutions, delivers automotive AI Cockpit technologies, fosters autonomous driving development through strategic agreements, and offers comprehensive autonomous vehicle solutions. This segment also manufactures cryptocurrency mining processors, supplies Jetson platforms for robotics and other embedded applications, and offers enterprise AI software, including NVIDIA AI Enterprise. These diverse offerings find widespread application across the gaming, professional visualization, data center, and automotive sectors. NVIDIA distributes its portfolio through a broad ecosystem, engaging original equipment and device manufacturers, system integrators, add-in board makers, retail channels, software vendors, internet and cloud service providers, automotive companies (both manufacturers and tier-1 suppliers), mapping firms, nascent technology ventures, and other industry stakeholders. A notable strategic partnership exists with Kroger Co. Founded in 1993, NVIDIA Corporation maintains its corporate headquarters in Santa Clara, California.

Broadcom has expanding HBM requirements associated with custom AI accelerators.

Broadcom's custom AI silicon and ASIC activities benefit from stronger hyperscaler design programs, although it captures downstream chip economics rather than EDA software revenue.

70% market share in custom AI accelerators for the largest hyperscalers.

Dominates high-end data center switching silicon required to connect massive GPU clusters for token efficiency.

SK Hynix is a leading HBM supplier and can capture demand and pricing from additional AI accelerator output.

The expanded AI accelerator opportunity increases demand for HBM, a critical memory input for AMD, NVIDIA, and other accelerators. SK Hynix is a leading HBM supplier with constrained committed output, giving it strong volume and pricing leverage.

TSMC captures fabrication and packaging revenue regardless of whether the accelerator architecture is merchant GPU or custom ASIC.

TSMC supplies the foundry and CoWoS packaging layers required for AI accelerator deployment.

Additional AI workloads can increase demand for accelerator wafers and advanced packaging across NVIDIA, AMD, and hyperscaler-designed silicon.

TSMC supplies advanced-node fabrication and CoWoS packaging for accelerator systems.

Additional AI infrastructure spending increases demand for advanced-node wafers and CoWoS packaging used by GPU, accelerator, and custom-silicon suppliers. Packaging constraints may delay shipments but support utilization and pricing.

AI accelerator demand and HBM integration increase utilization of advanced foundry and packaging capacity. TSMC's fully booked CoWoS platform indicates that continued AI infrastructure investment can support packaging-related revenue, although the direct exposure is to the accelerator ecosystem rather than conventional DRAM pricing.

AMD generates a large portion of its revenue from CPUs and GPUs for data centers and AI.

Expanded AI investment would enlarge the market for alternative accelerator suppliers. AMD is explicitly included among GPU and accelerator suppliers, giving it potential demand and share-capture upside, but its products depend on the same constrained HBM and advanced-packaging ecosystem.

Shanghai Iluvatar CoreX Semiconductor Co., Ltd. engages in the research, design, and deployment of advanced GPGPU products and AI computing solutions in China. It offers various GPGPU chips and accelerators, along with its proprietary and full-stack software development kits, including performance-optimized libraries, AI training frameworks and inference engines, enabling powerful AI applications across training and inference scenarios. It primarily develops and sell GPGPU accelerators, the accelerators integrate GPGPU chips with memory modules, PCB, and other critical components, forming complete computing accelerators that are ready for direct application in the end computing systems. TG Series, a training-focused product line; and ZK series, a specialized product line designed for cloud and edge inference applications. The AI Computing Solutions provides GPGPU Servers and GPGPU Clusters. The company sells its products through direct sales force and distributors. Shanghai Iluvatar CoreX Semiconductor Co., Ltd. was founded in 2015 and is based in Shanghai, China.

ASML supplies EUV lithography systems used to manufacture advanced semiconductors.

Synopsys benefits from recurring EDA, verification, and IP activity for every new custom accelerator generation.

A stronger custom-silicon pipeline creates recurring demand for electronic-design-automation tools used in design, verification, and implementation. The effect is second-order and depends on hyperscalers funding additional accelerator generations rather than merely reallocating existing workloads.

A multi-generation custom-accelerator program requires repeated architecture, implementation, verification, and interface-IP work. Synopsys can benefit from the design-tool intensity of custom silicon even though it does not capture the server-chip revenue directly.

Synopsys, Inc. is a leading provider of electronic design automation (EDA) software, instrumental in the creation and validation of integrated circuits. The company offers a robust portfolio of platforms, including the Fusion Design Platform for digital implementation, and the comprehensive Verification Continuum Platform. The latter provides a suite of solutions such as virtual prototyping, static and formal verification, simulation, emulation, and FPGA-based prototyping with integrated debugging tools. Synopsys also develops FPGA design products, which enable programming for specific functions. A significant part of its business involves intellectual property (IP) solutions. This encompasses IP for widely adopted communication standards like USB, PCI Express, DDR, Ethernet, SATA, MIPI, HDMI, and Bluetooth Low Energy. They also furnish analog IP, including data converters and audio codecs, alongside system-on-chip (SoC) infrastructure IP, specialized datapath and building block IP, and verification IP products. Furthermore, their offerings include mathematical and floating-point components, as well as Arm AMBA interconnect fabric and peripherals. Beyond core IP, Synopsys supplies logic libraries, embedded memories, and configurable processor cores, complemented by application-specific instruction-set processor tools for embedded applications. They also develop IP subsystems for audio, sensor, and data fusion functionalities, in addition to specialized security IP solutions. For advanced system design and analysis, Synopsys provides Platform Architect solutions for SoC architecture optimization, along with general virtual prototyping tools and HAPS FPGA-based prototyping systems. Their expertise extends to tools designed for optical systems and photonic devices. In addition to software and IP, the company delivers a range of services, including security testing, managed services, professional programs, and training. These services are designed to help customers identify and resolve security vulnerabilities and defects across the software development lifecycle. Manufacturing solutions are also part of their extensive offerings. Synopsys serves a diverse global client base, with applications across electronics, financial services, automotive, medicine, energy, and various industrial sectors. The company was founded in 1986 and is headquartered in Mountain View, California.

Cadence is a direct EDA workflow provider exposed to AI-driven chip-design automation, verification, and optimization demand.

Custom AI-chip programs require EDA software for design, physical implementation, verification, and tape-out. Cadence is part of the shared Synopsys/Cadence toolchain for advanced-node ASIC programs, creating a modest indirect revenue opportunity without a disclosed project-specific order.

Cadence Design Systems, Inc. operates globally, delivering a comprehensive suite of solutions that includes software, specialized hardware, professional services, and pre-designed integrated circuit (IC) building blocks. The company's robust functional verification offerings feature dedicated emulation and prototyping hardware. This portfolio encompasses JasperGold for formal verification, Xcelium for parallel logic simulation, Palladium as an enterprise-grade emulation platform, and Protium, a prototyping platform designed for thorough chip verification. Cadence additionally provides extensive products for digital IC design and final sign-off. These include the Genus logic synthesis solution, Joules for RTL power analysis, and the Modus software solution, which streamlines design-for-test (DFT) processes for systems-on-chip. Their tools further extend to physical implementation, covering place and route, optimization, and multiple patterning preparation, alongside essential sign-off products for validating designs prior to silicon manufacturing. Moreover, the company furnishes custom IC design and simulation tools capable of generating both schematic and physical representations of circuits down to the transistor level. These tools support a wide array of applications, such as analog, mixed-signal, custom digital, memory, and radio frequency (RF) designs. Cadence also offers system design and analysis products crucial for developing printed circuit boards (PCBs) and integrated circuit packages, along with capabilities to analyze complex multi-physics effects like electromagnetic and electro-thermal interactions. Furthermore, Cadence supplies intellectual property (IP) products, consisting of pre-verified, customizable functional blocks ready for integration into customer ICs. This includes verification IP and memory models used to simulate and characterize the behavior and interactions of standard industry system interface protocols. Complementing its product portfolio, Cadence provides a range of services, including methodology consulting, educational programs, hosted design solutions, and ongoing technical support and maintenance. Its diverse clientele spans sectors such as 5G communications, aerospace and defense, automotive, industrial and healthcare, mobile technology, consumer electronics, and hyperscale computing. Established in 1987, Cadence Design Systems, Inc. is headquartered in San Jose, California.

AI-critical chip demand currently exceeds near-term supply, supporting positive volume and pricing leverage for HBM and related memory products.

The rapid volume scaling of AI accelerators—even at lower margins for the designers—drives massive demand for High Bandwidth Memory (HBM). As custom silicon volume surges (e.g., Google TPU and Meta accelerators), Micron benefits from the structural supply shortage and pricing power in the memory market.

Micron supplies high-bandwidth memory used in AI accelerators and server platforms. Increased production of AI compute systems should support memory demand and improve the utilization opportunity for HBM suppliers, but the Nevada campuses do not disclose a Micron procurement relationship.

Samsung is one of the principal HBM suppliers and can capture memory demand from expanded AI accelerator production.

Samsung competes in HBM and advanced packaging. The broader growth in interposer-based AI packages supports its HBM and I-Cube alternatives, but TSMC's additional CoWoS supply may strengthen TSMC's competitive position against Samsung's packaging platform.

Samsung participates in the concentrated HBM supplier group serving AI accelerators. Stronger expected accelerator production supports HBM demand and pricing, while capacity constraints may delay conversion of demand into shipped systems.

Supplier of optical interconnects and networking silicon for cluster connectivity.

AI supercomputers require high-bandwidth optical interconnects and digital-signal-processing components. Marvell is identified as a DSP supplier used with optical modules in NVIDIA AI clusters, so additional cluster deployments can support interconnect-related demand.

Marvell benefits from AI data-center networking and custom-silicon programs. The NVIDIA result reinforces the scale of AI infrastructure spending, while custom ASIC growth creates design-service and accelerator revenue opportunities that are complementary to, but partly substitutive for, GPUs.

Innolight is a direct supplier of optical transceivers needed when processor clusters scale to thousands of nodes.

The target requires networked accelerator clusters rather than isolated servers, increasing demand for 800G, 1.6T, and future co-packaged optical links. Zhongji Innolight is identified as a leading optical-transceiver supplier with 1.6T production and CPO capability, providing direct exposure to AI-networking demand.

Unimicron supplies advanced package substrates that are required alongside silicon interposers in AI accelerator packages.

Unimicron is one of the concentrated suppliers of advanced package substrates used beneath interposer-based AI packages. More CoWoS/interposer output increases the need for qualified high-density substrates, though the announcement does not identify a direct Unimicron order.

Ibiden is a concentrated ABF-substrate supplier used in advanced accelerator packages.

AI-chip packaging requires qualified advanced substrates in addition to silicon interposers. Ibiden is one of the concentrated global suppliers of high-layer-count ABF substrates, so additional AI packaging capacity and chiplet demand can support substrate utilization, pricing and revenue, although the Singapore announcement does not identify a direct Ibiden contract.

ASM International supplies ALD and epitaxy process tools used in leading-edge semiconductor manufacturing.

ASM International is a focused beneficiary of advanced-node capital expenditure because GAA logic and advanced memory require atomic-layer deposition and epitaxy tools. Its narrower product portfolio makes the exposure more concentrated but also more directly linked to process-intensity increases.

Applied Materials provides deposition, CMP, materials-engineering, and process-control equipment used in semiconductor and advanced-packaging capacity expansion. The impact is positive but less immediate than for dedicated bonding-equipment vendors because customer capex and qualification precede revenue.

Applied Materials is an indirect equipment beneficiary. If stronger AI design activity leads to additional advanced-node wafer and packaging capacity, semiconductor manufacturers will require more deposition, etch, inspection, and packaging equipment. The pathway is delayed and capex-dependent: Synopsys's guidance concerns design activity, whereas Applied Materials is paid when fabs fund and build manufacturing capacity.

Lam benefits indirectly from process-equipment expansion needed to manufacture custom accelerators and HBM.

Incremental demand for HBM and AI accelerators can drive additional memory and advanced-node capacity investment, supporting Lam Research etch and deposition equipment demand. Equipment orders would lag the immediate chip purchase.

Commercialization of European AI accelerators would require additional etch and deposition capacity for advanced logic, memory, and packaging. Lam Research is identified as a core equipment supplier in the AI semiconductor bottleneck, creating an indirect capex opportunity without a disclosed order.

Incremental accelerator and HBM demand can eventually require additional wafer-fabrication, memory, and advanced-packaging capacity, supporting etch and deposition equipment demand. The pathway is indirect and capex-dependent.

KLA supplies inspection, metrology, and process-control equipment for expanded accelerator and HBM production.

Expanded AI-accelerator and HBM production requires inspection, metrology, and process-control equipment. KLA could benefit if the new compute demand induces funded semiconductor and packaging capacity expansions, but qualification timing makes the impact indirect.

ASE/SPIL is identified as an advanced-packaging capacity provider for AI accelerator production.

ASE captures outsourced backend spending when AI accelerator and HBM producers require additional assembly, packaging, and testing capacity.

ASE has an outsourced assembly and advanced-packaging exposure to incremental accelerator production.

A funded accelerator startup that reaches production will require outsourced assembly, testing, and advanced packaging. ASE is explicitly identified as an OSAT provider expanding advanced-packaging capacity and receiving outsourced CoWoS work, creating a potential beneficiary of packaging overflow and new accelerator volume.

Custom AI accelerators require outsourced assembly, advanced packaging, and testing. ASE can capture overflow and alternative backend work as CoWoS capacity remains constrained, supporting utilization and potentially pricing for advanced packaging services.

Amkor is an OSAT expanding 2.5D and 3D advanced-packaging lines for AI hardware demand.

Amkor benefits from backend packaging demand whether the accelerator is a GPU or custom ASIC.

A larger production run of AI accelerator systems could create overflow or supplemental outsourced assembly, testing, and advanced-packaging demand if captive CoWoS capacity is constrained. Conversion depends on qualification and procurement choices that have not been disclosed.

Amkor is a potential overflow beneficiary when AI accelerator customers cannot obtain sufficient captive advanced-packaging and testing capacity. Stronger AI, HBM, and packaging demand expands the addressable outsourced-backend market, but customer qualification and substitute-capacity constraints make the impact less certain than for primary equipment vendors.

Advantest captures test-equipment demand as HBM-equipped accelerator qualification and production complexity increase.

Advantest directly captures automated test-equipment spending as HBM and AI accelerators enter qualification and production.

Advantest supplies automated test equipment used as AI accelerators and HBM products move into higher-volume qualification and production. Record back-end semiconductor orders therefore imply incremental tester demand, although customer capital approvals and installation timing create a lag.

Advanced packaging is a recognized bottleneck in AI accelerators, supporting demand for qualified OSAT capacity.

JCET can support domestic localization through advanced packaging, but its role is manufacturing services rather than accelerator design.

Domestic accelerator expansion increases demand for outsourced assembly, advanced packaging, and testing. JCET can benefit from backend capacity requirements, particularly because advanced packaging is a bottleneck for HBM-equipped AI accelerators.

AMEC supplies domestic etch and deposition equipment that can become more valuable as China expands localized accelerator and memory manufacturing.

Scaling domestic accelerator fabrication increases demand for Chinese etch and deposition tools as foreign equipment access is restricted. AMEC can capture localization-driven equipment procurement, although process qualification remains an execution risk.

Direct supplier of critical fabrication tools that CXMT will purchase using the IPO proceeds for expansion.

BESI supplies packaging equipment that benefits from capacity expansion, but it does not receive accelerator revenue directly.

The custom-accelerator production ramp increases the need for advanced packaging capacity and related bonding equipment. BESI is an equipment supplier positioned to benefit from packaging-capacity expansion and higher pricing, although its exposure is indirect and depends on equipment orders.

BE Semiconductor Industries N.V. (BESI) specializes in the creation, manufacturing, global distribution, and servicing of advanced machinery vital for semiconductor assembly within the electronics and chip manufacturing industries. Its extensive product portfolio features sophisticated die attach equipment, encompassing systems for single-chip, multi-chip, multi-module, flip-chip, thermal compression bonding, fan-out wafer level packaging, hybrid and embedded bridge die bonding, and die sorting. Additionally, BESI provides a variety of packaging solutions, including conventional, ultra-thin, and wafer-level molding, alongside trim and form, and singulation systems. The company also offers plating equipment for tin, copper, precious metals, and solar applications, complete with their corresponding process chemicals. Supplemental to its core machinery, BESI supplies critical tooling, conversion kits, spare components, and various support services. With well-known brand names like Datacon, Esec, Fico, and Meco, the company primarily serves major multinational chip producers, assembly outsourcing specialists, and diverse electronics and industrial enterprises. Established in 1995, BE Semiconductor Industries N.V. maintains its headquarters in Duiven, Netherlands.

Teradyne captures downstream semiconductor test demand if custom AI accelerators progress from design to volume production.

Teradyne captures additional semiconductor test intensity as complex accelerator production expands.

Teradyne is an indirect beneficiary of the AI design-demand signal because successful complex AI chip programs eventually require automated test equipment. Larger and more intricate accelerators generally increase test intensity and backend production requirements, supporting incremental demand for Teradyne’s logic-testing systems.

The result supports broader demand for signal-integrity and scale-up connectivity products in AI servers. Astera's PCIe/CXL connectivity and retimer modules address the same physical expansion of GPU clusters, although Credo's AEC and retimer momentum also represents competitive pressure.

Astera Labs, Inc. develops, produces, and markets semiconductor-based connectivity solutions for cloud computing and artificial intelligence infrastructure. Its core offering, the Intelligent Connectivity Platform, comprises a comprehensive portfolio of data, network, and memory connectivity products. These are unified by a software-defined architecture, which empowers customers to seamlessly deploy and operate high-performance cloud and AI systems at scale. Established in 2017, the company is headquartered in Santa Clara, California.

Eoptolink participates in constrained 800G and 1.6T optical-module supply supporting AI infrastructure.

Large AI clusters need high-bandwidth, low-latency optical links between accelerators, switches, and data centers. Eoptolink is described as a high-speed optical-transceiver specialist with 400G and 800G shipments, positioning it to benefit from China’s computing-network expansion.

Applied Optoelectronics manufactures the 800G and 1.6T transceivers required to connect dense AI racks.

Higher custom-accelerator deployment requires additional optical modules and transceivers for cluster networking. Applied Optoelectronics is a named optical supplier, though qualification and production capacity may delay revenue conversion.

Coherent supplies optical-networking components that can be required by high-speed GPU clusters.

A 420,000-GPU-class hosting expansion requires high-speed optical connectivity between GPUs, switches and storage. Coherent is identified as a supplier of 800G and 1.6T transceivers required by AI clusters, creating an indirect networking-volume opportunity.

Production deployment of custom accelerator clusters increases demand for optical transceivers, lasers, and photonic components connecting racks and data-center fabrics. Coherent can benefit from higher 800G and 1.6T networking volumes, without a disclosed project-specific order.

Lumentum’s optical components and lasers are complementary inputs to the high-bandwidth interconnect required by distributed GPU clusters.

Large multi-rack GPU clusters require optical interconnects, lasers, and high-speed photonic components. Lumentum’s networking-optics exposure gives it potential incremental demand, although revenue depends on the final network architecture and qualification outcomes.

Lumentum Holdings Inc., established in 2015 and headquartered in San Jose, California, is a global leader in optical and photonic product manufacturing. The company's reach extends across the Americas, Asia-Pacific, Europe, the Middle East, and Africa. Lumentum operates through two principal business units: Optical Communications (OpComms) and Commercial Lasers. The OpComms segment develops and supplies components, modules, and subsystems essential for transmitting video, audio, and data across high-capacity fiber optic networks. Its offerings encompass a wide range of products, including tunable transponders, transceivers, and transmitter modules, alongside various tunable lasers, receivers, and modulators. Furthermore, this segment provides critical transport products such as reconfigurable optical add/drop multiplexers (ROADMs), optical amplifiers, and optical channel monitors. A comprehensive suite of components is also available, featuring 980nm, multi-mode, and Raman pumps; switches; attenuators; photodetectors; gain flattening filters; isolators; wavelength-division multiplexing (WDM) filters; arrayed waveguide gratings (AWGs); multiplex/de-multiplexers; and integrated passive modules. Distinctive innovations from this segment include the Super Transport Blade, which integrates multiple optical transport functions into a single unit; vertical-cavity surface-emitting lasers (VCSELs); directly modulated and electro-absorption modulated lasers; and advanced laser illumination sources for 3D sensing systems. OpComms serves key clientele in the telecommunications, data communications, consumer, and industrial markets. Conversely, the Commercial Lasers segment produces an array of laser technologies, such as diode-pumped solid-state, fiber, diode, direct-diode, and gas lasers. These are primarily utilized in original equipment manufacturer (OEM) applications. This segment's laser solutions cater to a broad spectrum of markets and uses, including sheet metal processing, general manufacturing, biotechnology, graphics and imaging, remote sensing, and precision machining.

High-bandwidth fiber and optical interconnect are necessary for distributed GPU clusters; HFCL's 800G and 1.6T-oriented portfolio can mitigate this networking bottleneck.

Cisco supplies the networking infrastructure and 800G optics used to connect the AI clusters.

Large GPU clusters need high-bandwidth networking and campus connectivity. Cisco is identified in the retrieved Saudi infrastructure context as a networking supplier, including Silicon One and 800G optics, but the specific 84 MW NVIDIA deployment does not disclose a Cisco award.

New GPU clusters require high-throughput networking and optical interconnect equipment.

Large GPU clusters require high-bandwidth Ethernet or InfiniBand-adjacent networking, switches and optical interfaces. Arista is identified as an enterprise networking provider serving high-performance AI networks, creating indirect demand exposure rather than direct hosting-contract revenue.

More custom accelerators and AI clusters require high-throughput Ethernet switching and networking fabrics. Arista can capture incremental data-center switching and interconnect revenue as hyperscalers expand AI infrastructure, but its exposure is downstream of chip availability, facility commissioning, and customer procurement decisions.

ASMPT is exposed to incremental HBM and advanced-packaging equipment spending. Persistent scarcity of leading-edge tools increases the incentive to expand packaging capacity, while its thermocompression-bonding equipment supports HBM assembly and accelerator production.

ASMPT is an advanced-packaging equipment supplier positioned to benefit as TSMC and other manufacturers add packaging capacity to address AI accelerator constraints. Demand is positive but timing depends on tool qualification and the pace of packaging-line construction.

Atlas Copco's Edwards and Leybold businesses supply vacuum pumps and abatement equipment used in EUV systems. Faster scanner production and fab expansion provide incremental subsystem demand, but the company is diversified and the retrieved evidence does not disclose a specific ASML contract.