The AI & digital infrastructure value chain: 62 U.S. stocks in 8 segments
Every U.S.-listed company Sectivia tracks in the AI & digital infrastructure value chain, grouped by segment from upstream suppliers to downstream customers, with 119 mapped supplier-to-customer links and today's numbers.
From chips and manufacturing equipment to cloud platforms, data centers and the software layer around AI. Companies are grouped by their role in the value chain; tags mark exposure that cuts across sectors. Click a card for the full profile.
Today
The AI chain in numbers, from the latest close.
Data as of September 17, 2026 · updates daily after U.S. close
Chips & accelerators15 companies
Designs and builds the processors, accelerators and memory that train and run AI.
Nvidia sells the accelerators most large AI models are trained on, from H100 through B200 to Rubin.
Custom silicon designed to order: Broadcom builds AI ASICs to hyperscalers' own specifications, Google's TPUs among them.
America's only large memory manufacturer, third in HBM, selling DRAM and NAND broadly into data centres.
The MI series is the closest thing Nvidia has to a direct rival in AI GPUs.
A turnaround with government money behind it.
Licensing, not manufacturing: the CPU architectures and instruction sets (Cortex, Neoverse) that nearly every mobile and server chip is built around come from Arm, which takes a royalty on each chip produced.
No one ships more analog chips and embedded processors: the parts that handle power and signal duties inside industrial equipment, cars and consumer electronics.
Data infrastructure only.
Snapdragon made Qualcomm the reference point in mobile silicon.
No one ships more hard drives.
Precision analog and mixed-signal chips, plus DSPs, for industrial equipment, communications, cars and healthcare.
Hard drives for data centres, where AI has pushed demand for cheap bulk capacity to records.
Everything here is power delivery: the chips and modules that feed processors in AI server racks, with cars and industrial equipment taking the same parts.
The microcontroller is the anchor product; analog and FPGA parts fill out the range for embedded designs in industry and cars.
Electric cars and factory automation are the end markets for its silicon carbide power semiconductors and image sensors.
Semiconductor equipment & IP7 companies
The machines and chip IP advanced semiconductor manufacturing cannot do without.
Etch and deposition tools, strongest where HBM memory and 3D stacking need them.
Applied Materials covers more steps of chipmaking than any other equipment vendor: deposition, etch, implantation, inspection, and the packaging tools that stack AI chips.
Finding a defect a few nanometres across is its own industry, and KLA dominates it.
Ranks second in EDA behind Synopsys: design and verification software, licensable IP blocks, plus simulation that reaches past the chip to the whole system.
Chips are designed and verified in Synopsys' EDA software, and much of the reusable silicon IP inside them is licensed from Synopsys too.
Two businesses: automated chip test systems, and the Universal Robots cobot line.
Filtration and advanced materials that keep contamination out of chip fabrication.
Cloud platforms5 companies
Rents out the compute, the 'power grid' of the AI era.
Few companies own the whole stack.
No software company is bigger, and few have as many places to put AI: Azure rents the infrastructure, Copilot sells the assistant inside Microsoft 365 and GitHub, Maia is a chip of Microsoft's own, and the OpenAI partnership sits underneath much of it.
AWS is still the biggest cloud platform, and Amazon increasingly fills it with its own Trainium and Inferentia silicon rather than buying everything in.
The database business paid for the pivot.
Rents out GPU capacity and nothing else.
Data centers14 companies
Servers, networking, optics and data center capacity that tie AI infrastructure together.
Switches and routers at a scale no other network vendor matches, plus Silicon One chips of its own design and the security business built around Splunk.
Dell's AI server backlog has grown quickly, sold to enterprises and cloud operators alike, on top of a long-standing PC and storage business.
Arista's 800G switches and the EOS software running them sit inside AI data centres, which makes hyperscaler spending the number to watch.
Interconnect hardware: connectors and cable, plus a sensor business.
Carrier-neutral by design.
Cray supercomputers and ProLiant AI servers on one side; on the other, a networking business that grew considerably with the Juniper acquisition.
Owns and runs data centres as a REIT, from wholesale halls let to hyperscalers down to smaller colocation space.
Makes the transceivers and lasers that turn data into light for the hop between AI servers.
Long-haul optical transmission, mostly between data centres.
Signal integrity becomes a real constraint at AI-server speeds.
Keeps enterprise data on flash arrays and moves it between on-premises kit and the cloud; AI workloads are the newer source of demand.
Inside the rack, GPUs and switches are joined by active electrical cables and SerDes chips.
Racks of GPU servers, liquid-cooled, put together quickly after each new GPU arrives.
All-flash arrays for data centers, sold as a subscription service.
Cooling, power & data center infrastructure5 companies
Power, cooling and physical equipment, AI's physical foundation.
An AI rack throws off enough heat that air alone stops working.
Mechanical contracting: HVAC and cooling, plus electrical and plumbing work.
Air conditioning and refrigeration, from heat pumps in buildings to cooling systems for data centers.
Builds the electrical and mechanical systems inside large facilities, data centers among them, then services what it installed.
Thermal systems were the old business; data centre cooling is the new one, sized for the heat AI servers give off.
AI software & productivity9 companies
Embeds AI models into the products and workflows we already use.
Facebook and Instagram pay for one of the largest AI build-outs anywhere: GPU clusters at enormous scale, the Llama models released openly, and advertising tools that are so far the clearest payoff.
Selling electric cars and energy storage pays the bills, but the share price increasingly rests on software: autonomous driving under the FSD and robotaxi banner, and the Optimus humanoid robot, both built on AI chips and data Tesla owns outright.
Sells more CRM software than any rival, and is now pushing Agentforce, AI agents that handle sales and service conversations on their own.
Large organizations run their internal workflows on the Now Platform, from IT tickets to HR and customer-service cases, with AI agents taking over the routine steps.
The Axon engine decides which ad a given app user sees, and its results are among the most cited evidence that applied AI pays off in advertising.
Photoshop and the rest of Creative Cloud are default tools in the design trade, now with Firefly generative AI built in.
TurboTax files tax returns, QuickBooks keeps small-business books, Credit Karma sits on top of consumer credit; AI assistants are being pushed into all of them.
Cloud HR and finance software for large employers, with AI built in.
Language lessons on a phone, with generative AI writing much of the course content and playing the conversation partner in the paid Max tier.
Data, analytics & enterprise AI4 companies
Makes enterprise data 'AI-ready' and powers operational AI.
Governments and large companies feed their data into Foundry and Gotham; AIP, the newer layer, turns AI models into tools that staff use to make actual decisions.
The warehouse comes before the models: companies pool and share their data in Snowflake's cloud, and Cortex AI runs on top of it.
Tracks the health of cloud systems, including the AI models and LLM applications running on them.
A document database for developers, run as the managed Atlas cloud service and extended with vector search for AI applications.
IT consultants & integrators3 companies
Help enterprises adopt and scale AI in practice.
An old hand in enterprise computing, selling the watsonx AI platform alongside Red Hat's hybrid cloud, with a consulting arm that builds the AI systems customers actually run.
When a large company decides to roll out AI or move to the cloud, Accenture is usually among the firms hired to do the work.
Runs IT operations for clients under outsourcing contracts; the AI and digital transformation side of the business is growing.
How the chain fits together
Sectivia has mapped 119 supplier-to-customer relationships between the 62 companies in this chain, 11 of them between companies in the same segment. The densest link runs from Chips & accelerators to Cloud platforms: 24 mapped relationships. The most connected company is Microsoft (MSFT), with 16 suppliers and 5 customers in the map.