The producers, toolmakers, and infrastructure suppliers at the centre of the AI buildout — each profiled against the chokepoints they touch.
Absolics operates what is described as the world's first dedicated glass substrate manufacturing facility in the United States, positioning it as a potential sole commercial supplier of a packaging material that could replace silicon interposers in next-generation AI accelerators.
Alchip is a pure-play turnkey ASIC design and back-end house operating exclusively at leading-edge process nodes, making it a direct and concentrated proxy for the hyperscaler shift from merchant GPUs to custom silicon for AI infrastructure.
As the only U.S.-based OSAT of scale building a domestic CoWoS-capable campus next door to TSMC's Arizona fabs, Amkor is the linchpin of any credible end-to-end AI chip supply chain on American soil.
Applied Materials sits at the broadest cross-section of AI hardware supply constraints, supplying deposition, etch, and materials engineering tools that are indispensable across wafer fab equipment, leading-edge foundry logic, and CoWoS advanced packaging.
AOI manufactures optical transceivers and the laser chips inside them, making it a direct participant in one of the hardest-to-scale links in the AI data center buildout: photonic interconnect capacity.
As the world's largest OSAT provider and a key overflow partner in TSMC's CoWoS ecosystem, ASE Technology sits at the center of the advanced packaging bottleneck that constrains AI chip production at scale.
ASM International is the dominant pure-play in atomic layer deposition equipment, the deposition technology without which advanced AI chips at gate-all-around nodes cannot be manufactured, making it a structural chokepoint in the wafer-fab-equipment layer of the AI supply chain.
ASML is the sole manufacturer of extreme ultraviolet (EUV) lithography scanners, which are the only tools capable of producing the sub-7nm chips that power AI accelerators, making every leading-edge foundry and wafer fab equipment supply chain dependent on a single Dutch company.
AT&S is the only European-headquartered producer of high-end ABF-based IC substrates at scale, making it a rare Western node in a supply chain otherwise dominated by Asian manufacturers and a critical enabler of AI processor packaging.
BESI is the market leader in hybrid bonding equipment, the precision bonding technology at the core of CoWoS-class advanced packaging for AI accelerators, making it a direct and high-exposure chokepoint in the AI hardware supply chain.
Bloom Energy's solid oxide fuel cells allow AI data centers to generate power on-site and bypass strained utility grids, making it a direct alternative to both traditional gas turbines and grid-dependent power delivery at a moment when interconnection queues and transformer shortages are slowing buildout.
Boyd (now Boyd Thermal, operating within Eaton) is one of the highest-volume producers of liquid cold plates and direct-to-chip cooling systems globally, making it a critical hardware node in the AI buildout's liquid-cooling chokepoint.
Broadcom is the dominant co-designer of custom AI accelerators (XPUs) for hyperscalers and a leading developer of co-packaged optics, making it a direct chokepoint in both the custom-silicon and optical-transceiver layers of the AI supply chain.
Carl Zeiss SMT is the sole supplier of precision optical systems to ASML, meaning every EUV lithography machine used to produce AI-era chips depends entirely on optics made in Oberkochen, Germany.
Caterpillar is one of the world's leading suppliers of diesel and gas generator sets used for backup and prime power at data centers, placing it directly inside the generation chokepoint of the AI infrastructure buildout.
Coherent is one of the few vertically integrated manufacturers of EML lasers and high-speed optical transceivers that are essential for moving data inside AI data centers, giving it direct exposure to the most congested link in the AI buildout supply chain.
Constellation operates the largest nuclear fleet in the United States and is the primary utility counterparty for long-duration, carbon-free baseload PPAs that hyperscalers are relying on to power AI data centers at scale.
CoolIT Systems is one of the world's leading pure-play manufacturers of direct-to-chip cold plates and coolant distribution units, placing it at the center of the liquid-cooling chokepoint that is gating the deployment of every high-density AI GPU cluster.
Corning supplies glass materials and Glass Core substrates that are emerging as a replacement for silicon interposers in AI accelerator packaging, positioning it as an upstream material chokepoint as the semiconductor industry shifts toward glass-based advanced packaging.
Eaton sits at two of the sharpest chokepoints in the AI buildout: grid-level switchgear and electrical distribution, where its backlog and order momentum directly reflect how constrained power infrastructure has become for hyperscale data centers.
Eoptolink is one of the highest-volume suppliers of 800G optical transceivers globally, making it a direct and high-exposure chokepoint in the physical layer that connects every GPU cluster, AI training pod, and hyperscale data center.
GE Vernova sits at three simultaneous chokepoints in the AI power buildout: it holds the largest gas-turbine backlog of any manufacturer, supplies the grid transformers that connect data centers to the grid, and is scaling both businesses into a demand environment that has outrun available capacity.
As the world's third-largest silicon wafer supplier and the only producer of advanced 300mm wafers currently expanding inside the United States, GlobalWafers controls a chokepoint that sits upstream of every leading-edge AI chip fab in the country.
As the world's largest transformer manufacturer and a pioneer of HVDC technology, Hitachi Energy sits at the center of the most acute hardware chokepoint blocking AI data center expansion: the global shortage of large power transformers.
Ibiden is the dominant producer of high-density ABF package substrates used in AI and HPC processors, making it one of the most direct chokepoints in the physical AI supply chain.
As the world's largest optical transceiver module maker and the supplier of more than half of NVIDIA's 800G module procurement, InnoLight sits at the single most congested bandwidth chokepoint in the AI data-center buildout.
Intel Foundry is the only U.S.-based operator attempting to reach leading-edge logic production at scale, making its 18A node a critical variable in whether Western AI chip builders can source advanced wafers from a domestic, non-TSMC supplier.
KLA holds a commanding position in semiconductor process control and inspection, a chokepoint that becomes more critical at every advanced node required for AI chips, making its tools a mandatory step in the AI supply chain regardless of which chipmaker or chip designer wins market share.
Lam Research is the global leader in plasma etch and a top-tier deposition equipment supplier, making it a direct chokepoint across wafer-fab equipment broadly, HBM memory manufacturing, and the leading-edge foundry transition to gate-all-around transistors that underpin AI chips and accelerators.
LiquidStack is a full-service provider of direct-to-chip and immersion cooling systems whose products sit at a hard physical constraint in the AI buildout: without adequate heat rejection, accelerator-dense racks cannot run at rated power.
Lumentum is the only supplier shipping 200G-per-lane EML laser chips at volume, making it the gating component for every 1.6T optical transceiver deployed in AI data centers today.
Marvell is one of the two dominant chip-shepherding partners for hyperscale custom AI silicon, with confirmed programs for Amazon Trainium, Microsoft Maia, Meta, and Google, making it a direct chokepoint in the custom-ASIC layer of the AI buildout.
MediaTek is transitioning from smartphone SoC leader to a custom AI ASIC provider for hyperscalers, with roughly $2 billion in AI ASIC revenue expected in Q4 2026 alone and a growing role inside Google's TPU roadmap that makes it a direct participant in the AI infrastructure buildout.
As the only U.S.-based manufacturer of DRAM and HBM, Micron is a structurally irreplaceable node in the AI memory supply chain, with its entire 2026 HBM output already committed and its consumer business shuttered to free capacity for AI customers.
As one of only three manufacturers worldwide capable of producing large-scale gas turbines, Mitsubishi Heavy sits at the center of the power-generation bottleneck throttling AI data center construction, with its order book sold out into 2028 and a capacity-doubling program underway.
Nan Ya PCB is one of Taiwan's leading ABF substrate manufacturers, making it a direct bottleneck in the packaging supply chain that connects TSMC's advanced nodes to finished AI chips for GPUs, ASICs, and high-speed networking silicon.
As the dominant supplier of AI accelerators and the single largest consumer of silicon interposers, Nvidia sits at the center of the most constrained chokepoint in the AI hardware supply chain, where packaging capacity directly limits how fast the global AI buildout can proceed.
A growing field of companies deploying glass substrates and large-format panel processes are positioning themselves as structural alternatives to TSMC's silicon-interposer-based CoWoS, directly targeting one of the hardest capacity chokepoints in the AI accelerator supply chain.
As the dominant U.S. contractor for electric grid construction and high-voltage infrastructure, Quanta Services sits at the physical chokepoint where AI-driven power demand must be translated into new transmission lines, substations, and transformer installations.
Samsung is the only company simultaneously supplying DRAM and HBM at scale, operating leading-edge EUV fabs for both logic and memory, and developing next-generation glass-interposer packaging, making it one of the widest single points of leverage across the AI hardware supply chain.
Schneider Electric supplies the power distribution, UPS, and liquid-cooling infrastructure that sits between the grid and every GPU in a hyperscale AI data center, making it a foundational chokepoint vendor across both the power and cooling layers of the AI buildout.
Shin-Etsu Chemical is the world's largest producer of semiconductor-grade silicon wafers, the foundational substrate on which every AI chip is fabricated, giving it direct and irreplaceable exposure to the AI semiconductor buildout.
Shinko Electric is one of a small handful of Japanese manufacturers that produce the advanced ABF substrates on which high-performance AI and HPC processors depend, making it a direct chokepoint in the AI chip packaging supply chain.
Siemens Energy sits at two of the hardest chokepoints in the AI power buildout simultaneously: it is one of only three global makers of large gas turbines, and it is a leading supplier of the grid transformers and HVDC systems that connect generation to data centers.
As Europe's primary volume producer of 300mm silicon wafers and one of five companies controlling roughly 85% of global 300mm capacity, Siltronic sits at a structural chokepoint in the substrate supply chain that underlies every advanced AI chip.
SK Hynix is the dominant supplier of High Bandwidth Memory to Nvidia and other AI accelerator makers, holding roughly 62% of HBM shipments and an estimated 70% of Nvidia's next-generation HBM4 allocation for the Rubin platform, making it the single most consequential choke point in AI memory supply.
As the world's fifth-largest silicon wafer maker and South Korea's sole semiconductor wafer manufacturer, SK Siltron sits inside the tightly concentrated oligopoly that supplies every AI chip fab with the substrate on which logic and memory are built.
As the world's second-largest silicon wafer supplier, SUMCO is actively exiting 200mm production to concentrate capacity on the AI-grade 300mm wafers that every advanced logic and high-bandwidth memory chip requires.
Super Micro is one of the primary rack-scale integrators for GPU-dense AI clusters, and its proprietary direct liquid cooling technology places it at the center of the thermal management chokepoint that determines how fast next-generation AI infrastructure can be deployed at scale.
Tokyo Electron holds a near-monopoly in EUV coater/developer tracks and ranks among the top three global wafer-fab-equipment suppliers by revenue, making it a direct, unavoidable chokepoint for every leading-edge logic and memory chip powering the AI buildout.
TSMC is the single indispensable node in the AI hardware supply chain, controlling leading-edge logic fabrication, the dominant share of CoWoS advanced packaging, silicon interposer production, and the foundry base-die process for next-generation HBM4 memory.
As the world's largest ABF substrate maker with roughly 22% global market share and direct positions in CoWoS and EMIB advanced packaging supply chains, Unimicron sits at the single most concentrated chokepoint between chip design and AI accelerator production.
Vertiv sits at two of the AI buildout's most acute chokepoints simultaneously: it supplies the power management and thermal systems that keep hyperscale and AI data centers running, and it holds the leading position in the liquid-cooling market that GPU-dense racks increasingly demand.