The AI race is often described as a contest between chatbots, search engines and flashy consumer apps. In reality, it is a story of factories, patents, chip blueprints, cloud infrastructure and a handful of companies whose products most people will never see. Every AI prompt, image, video or line of code passes through a supply chain built over decades by businesses that solved very different problems long before artificial intelligence became the defining technology of the decade.

Understanding how these companies make money, and how deeply they depend on one another, explains far more than the latest product launch or earnings report. It reveals why Nvidia’s success is tied to TSMC, why TSMC cannot operate without ASML, why Microsoft, Google and Amazon are both Nvidia’s biggest customers and its future competitors, and why companies such as Arm and Broadcom quietly collect profits regardless of which AI model dominates. Together, they form the industrial backbone of the AI economy, a tightly connected ecosystem where one breakthrough or one disruption can ripple across global markets within hours.

For anyone trying to understand the AI industry, it can be difficult to keep track of what each company actually does. Although they often appear together in headlines, their business models, customers, and roles in the AIon ecosystem are remarkably different.

This report compares the business models of thirteen companies that sit at the center of the global chip supply chain. Some of these companies design chips. Some manufacture them. Some make the machines that manufacture them. Some license the blueprints. And some spend hundreds of billions of dollars buying all of the above to build the data centers that run today’s AI. Understanding how they make money, and how their fortunes depend on each other, explains most of what is driving markets in 2026.

All market capitalization and share price figures below are approximate as of July 28, 2026 and change by the hour once markets are open. They are included to show scale and relative size, not as investment guidance.

Chart 1. Market capitalization across the thirteen companies, approximate, late July 2026.

The value chain: How these companies fit together

It helps to think of these thirteen companies as sitting at different layers of a single stack, rather than as thirteen unrelated businesses competing at random. Very roughly, the layers run like this.

Design the machines that make chips	ASML
Manufacture the physical chips	TSMC, Samsung, Intel
License the chip blueprints	Arm
Design the chips themselves	NVIDIA, AMD, Apple, Qualcomm, Intel, Broadcom, and increasingly Google and Amazon
Buy the finished chips to build AI data centers	Microsoft, Google, Amazon, and to a lesser extent Apple

A single Blackwell AI chip from NVIDIA illustrates the whole chain. NVIDIA designs it using Arm licensed technology for some supporting components. TSMC manufactures it, using ASML’s lithography machines to etch the pattern onto silicon. Samsung or SK Hynix supplies the high bandwidth memory that sits next to it. The finished chip is bought by Microsoft, Google or Amazon and installed in a data center, where it trains and runs AI models that Apple, among others, quietly taps into for some of its own AI features. Every company in this report touches that one chip somewhere along its journey.

A second pattern worth noticing is that the biggest buyers of AI chips are also trying hardest to stop needing them. Microsoft, Google and Amazon each buy enormous numbers of NVIDIA and AMD chips, and each is also designing its own AI chips, Microsoft’s Maia, Google’s TPU and Amazon’s Trainium, specifically to reduce that dependence over time. Broadcom makes much of its AI revenue helping them do exactly that, designing the custom chips that Google and reportedly Meta and OpenAI use as alternatives to NVIDIA hardware.

Chart 2. Year founded for each company, spanning nearly six decades.

The age spread is a useful reminder that this is not a story of thirteen AI startups. Intel and AMD are both older than the moon landing. Microsoft and Apple were founded in the 1970s. TSMC, ASML and NVIDIA, which now dominate headlines about AI, were all built in the 1980s and 1990s to solve much older problems in chip design and manufacturing, long before anyone used the term artificial intelligence the way we do today.

Company profiles

Each profile below covers the company’s core business, the side businesses that diversify its revenue, and where it stands on artificial intelligence, followed by a widely cited quote from its chief executive.

NVIDIA

Founded	1993, by Jensen Huang, Chris Malachowsky and Curtis Priem
Headquarters	2788 San Tomas Expressway, Santa Clara, California, United States
CEO	Jensen Huang, co-founder and CEO since 1993
Market cap	approximately $4.8 trillion (late July 2026)
Share price	approximately $196 to $208 per share

NVIDIA designs graphics processing units, the chips that were originally built to render video game images fast but turned out to be extremely good at the kind of math that artificial intelligence needs. Today NVIDIA does not manufacture anything itself. It designs the chip, then TSMC builds it. Its main product line for AI is called the Blackwell and Hopper family of GPUs, sold mostly to big cloud companies and AI labs building giant data centers.

Beyond GPUs, NVIDIA sells networking equipment through its Mellanox acquisition, software subscriptions through its CUDA and AI Enterprise platforms, chips for self driving cars through its DRIVE unit, and simulation software for robotics and factories called Omniverse. It has also started designing its own central processing units, moving into territory that used to belong to Intel, AMD and Arm licensees.

NVIDIA is the most central company in the AI boom. Its GPUs train and run almost every major AI model in the world, from ChatGPT to Google’s Gemini. This has made it briefly the most valuable public company on earth, trading places with Apple and Alphabet for that title through 2026. Its dominance is also its biggest risk. Every major customer, including Microsoft, Amazon, Google and Meta, is trying to design its own chips so it depends less on NVIDIA.

“The more you buy, the more you save.”
  Jensen Huang, CEO of NVIDIA, describing how AI data center spending pays for itself through efficiency gains

TSMC (Taiwan Semiconductor Manufacturing Company)

Founded	1987, by Morris Chang
Headquarters	Hsinchu Science Park, Hsinchu, Taiwan
CEO	C. C. Wei, chairman and chief executive officer since 2018
Market cap	approximately $2.1 to $2.2 trillion (late July 2026)
Share price	approximately $300 to $400 per American depositary share

TSMC does not design any chips of its own. It builds chips that other companies design. This is called a foundry business, and TSMC invented the modern version of it. When NVIDIA, Apple, AMD, Qualcomm or Broadcom finish designing a chip, they send the blueprint to TSMC, which manufactures the physical silicon in enormous factories called fabs. TSMC makes roughly nine out of every ten of the most advanced chips in the world.

TSMC also offers advanced chip packaging, a process called CoWoS that is essential for stacking the memory and logic chips used in AI accelerators, plus mask making and design support services for its customers.

TSMC is the single most important company in AI hardware that most people have never heard of. Every advanced AI chip from every major designer, NVIDIA, AMD, Apple, Google and Amazon included, is physically built inside a TSMC factory. The company is spending tens of billions of dollars building new plants in Arizona, Japan and Germany to reduce the world’s reliance on its Taiwan factories, a concern given regional tensions with China.

“Trust with customers is what gives us our long term compass.”
Christophe Fouquet’s ASML counterpart at TSMC, CEO C. C. Wei, has repeated a similar sentiment across earnings calls, describing TSMC’s foundry model as built on decades of customer trust rather than competing with its own clients

AMD (Advanced Micro Devices)

Founded	1969, by Jerry Sanders and a group of colleagues who left Fairchild Semiconductor
Headquarters	2485 Augustine Drive, Santa Clara, California, United States
CEO	Lisa Su, chair, president and chief executive officer since 2014
Market cap	approximately $850 billion (late July 2026)
Share price	approximately $544 per share

AMD designs central processing units, the general purpose brains of computers, sold under the Ryzen brand for consumers and the EPYC brand for data centers. It also designs graphics processing units for gaming and, increasingly, for AI, sold under the Radeon and Instinct brands.

AMD acquired Xilinx in 2022 to add programmable chips called FPGAs used in networking and defense equipment, and it bought ZT Systems in 2025 to be able to sell whole AI server racks rather than just individual chips.

AMD is NVIDIA’s most direct challenger in AI accelerators. Its Instinct MI300 and MI350 chips are used by Microsoft, Meta and OpenAI as an alternative supply source so those companies are not entirely dependent on NVIDIA. Lisa Su has framed this as an industry wide need for more than one supplier, and under her leadership AMD has gone from near bankruptcy in 2014 to a company worth hundreds of billions of dollars.

“We’re seeing 30 times more performance with our Helios systems.”
  Lisa Su, CEO of AMD, on the performance gains of its newest AI server platform

Arm Holdings

Founded	1990, as a joint venture between Acorn Computers, Apple and VLSI Technology
Headquarters	110 Fulbourn Road, Cambridge, United Kingdom
CEO	Rene Haas, chief executive officer since 2022
Market cap	approximately $300 billion (late July 2026, though it has traded as low as $130 billion earlier in the year)
Share price	approximately $260 to $280 per share

Arm does not make chips at all. It designs the underlying instruction set and blueprints that other companies license to build their own chips. Almost every smartphone processor in the world, including Apple’s A series and M series chips, Qualcomm’s Snapdragon chips and Samsung’s Exynos chips, is built on Arm’s architecture. Arm earns money through licensing fees and small royalties on every chip built using its designs.

Arm is majority owned by the Japanese investment group SoftBank. It has recently begun designing its own complete chips rather than only licensing blueprints, starting with a new data center processor aimed at AI workloads.

Arm’s architecture is quietly inside almost every AI enabled device on the planet, from phones to data center servers. NVIDIA’s own Grace CPU, used alongside its GPUs in AI supercomputers, is built on Arm’s architecture rather than the older designs from Intel or AMD. This has made Arm newly important to the AI story even though it never manufactures a single chip.

“The AI boom is supply constrained with robust demand.”
Rene Haas, CEO of Arm Holdings

Intel

Founded	1968, by Robert Noyce and Gordon Moore
Headquarters	2200 Mission College Boulevard, Santa Clara, California, United States
CEO	Lip-Bu Tan, chief executive officer since March 2025
Market cap	approximately $460 billion (late July 2026)
Share price	approximately $91 to $100 per share

Intel both designs and manufactures its own chips, a model called integrated device manufacturing that TSMC and the fabless companies abandoned decades ago. Its main products are Core processors for personal computers and Xeon processors for servers, plus its own chip manufacturing plants that it is now trying to open up to outside customers under a new foundry business.

Intel previously owned Mobileye, a self driving car technology company, and Altera, a maker of programmable chips, though it has sold down its stakes in both to raise cash. It is investing heavily in becoming a contract manufacturer for other companies’ chip designs, directly competing with TSMC and Samsung.

Intel has struggled to keep pace with NVIDIA and AMD in AI accelerators and lost significant ground in the smartphone and mobile era to Arm based designs. Under Lip-Bu Tan, who took over in 2025, Intel is betting on custom AI chips built for specific customers and on its manufacturing turnaround, helped by a direct equity investment from the United States government intended to keep advanced chip production onshore.

“We are the only company that can design, manufacture, and build the entire range of computing solutions, from general purpose, traditional CPUs and GPUs, to more purpose built ASICs and CPUs optimized for agentic AI.”
Lip-Bu Tan, CEO of Intel, on the company’s second quarter 2026 earnings call

Samsung Electronics

Founded	1969, part of the wider Samsung Group founded in 1938 by Lee Byung-chul
Headquarters	129 Samsung-ro, Yeongtong-gu, Suwon, South Korea
CEO	Co-led by TM Roh, head of the Device eXperience division, and Young Hyun Jun, vice chairman and head of the Device Solutions division. Jay Y. Lee is executive chairman of the wider Samsung Electronics business.
Market cap	approximately $1.1 to $1.14 trillion (late July 2026)
Share price	varies by listing venue; Samsung is not listed on a United States exchange

Samsung Electronics is split into two very different businesses. Its Device Solutions division makes memory chips, the DRAM and NAND flash chips that store data in almost every computer and phone on earth, and it is the largest memory chip maker in the world. Its Device eXperience division makes finished consumer products, most famously the Galaxy line of smartphones, along with televisions and home appliances.

Samsung also runs a foundry business, similar to TSMC’s, that manufactures chips designed by other companies, though it holds a smaller share of that market than TSMC. It makes display panels, semiconductor manufacturing equipment for its own use and a wide range of consumer electronics.

Samsung’s memory chips, particularly a specialized type called high bandwidth memory, are a critical ingredient in NVIDIA’s and AMD’s AI accelerators, sitting right next to the processor to feed it data fast enough. On the consumer side, Samsung has built AI features directly into its Galaxy phones. Its foundry business is trying to win more AI chip manufacturing work away from TSMC, so far with limited success.

“We expect him to leverage his extensive experience and play a leading role in efforts to ensure swift, well coordinated responses to key investor and stakeholder engagements.”
Samsung’s board, describing vice chairman and co-CEO Young Hyun Jun’s mandate over the memory business at the heart of its AI hardware strategy

ASML

Founded	1984 in Veldhoven, as a joint venture between Philips and ASM International
Headquarters	De Run 6501, Veldhoven, Netherlands
CEO	Christophe Fouquet, chief executive officer since April 2024
Market cap	approximately $650 to $675 billion (late July 2026)
Share price	approximately $1,700 to $1,900 per share, or around €1,400 to €1,600 on the Amsterdam exchange

ASML makes the machines that make the chips. It is the only company in the world that builds extreme ultraviolet lithography systems, the enormously complex machines that use light to carve the microscopic patterns onto silicon that become transistors. Without ASML’s machines, none of the world’s most advanced chips, including every leading AI chip, could physically be manufactured.

ASML also sells older deep ultraviolet lithography systems for less advanced chips, along with metrology and inspection equipment that checks chips for defects during manufacturing, and it earns a large recurring revenue stream from servicing and upgrading machines already installed at customer factories.

ASML sits at the very base of the entire AI chip supply chain. TSMC, Samsung and Intel all depend entirely on ASML’s machines to manufacture the advanced chips that NVIDIA, AMD, Apple and others design. Export restrictions on selling ASML’s most advanced machines to China have become a major flashpoint in the broader technology rivalry between the United States and China.

“We understand the critical place we have, and if you are in a critical place, you need to be responsible for it.”
Christophe Fouquet, CEO of ASML

Apple

Founded	1976, by Steve Jobs, Steve Wozniak and Ronald Wayne
Headquarters	One Apple Park Way, Cupertino, California, United States
CEO	Tim Cook, chief executive officer since 2011, planning to step down on September 1, 2026 and become executive chairman, handing the role to John Ternus
Market cap	approximately $4.9 to $5 trillion (late July 2026)
Share price	approximately $339 to $343 per share

Apple designs and sells premium consumer hardware, above all the iPhone, alongside the Mac, iPad, Apple Watch and AirPods. It also runs a fast growing Services business built around the App Store, iCloud storage, Apple Music, Apple TV+ and other subscriptions, which now delivers some of the company’s highest profit margins.

Apple designs its own chips, the A series for iPhones and M series for Macs, though it has TSMC manufacture them. It offers financial products through Apple Card and Apple Pay, produces original film and television content through Apple TV+, and recently introduced a device leasing program called Upgrade that lets customers pay monthly for hardware rather than buying it outright.

Apple has taken a noticeably more cautious approach to AI spending than its rivals. Rather than building enormous AI data centers itself, it has kept capital expenditure comparatively low while building AI features, marketed as Apple Intelligence, directly into its devices and quietly using outside AI models, including from OpenAI and Google, for some of those features. Investors have rewarded this restraint, with Apple briefly overtaking NVIDIA as the world’s most valuable company in July 2026.

“Once criticized for not spending more on AI, they have been able to avoid some of those capex pitfalls.”
Jay Woods, chief market strategist at Freedom Capital Markets, describing how markets have come to view Apple’s restrained AI spending as a strength

Qualcomm

Founded	1985, by Irwin Jacobs and six co-founders
Headquarters	5775 Morehouse Drive, San Diego, California, United States
CEO	Cristiano Amon, president and chief executive officer since 2021
Market cap	approximately $170 to $180 billion (late July 2026)
Share price	approximately $170 per share

Qualcomm designs the Snapdragon family of chips that power most Android smartphones, combining a processor, modem and other components onto a single piece of silicon. It also earns a large share of its profit from licensing its wireless patents, charging phone makers a fee for every device that uses cellular technology Qualcomm helped invent.

Qualcomm has expanded into chips for cars, wearables, augmented and virtual reality headsets, and Windows laptops, where its Snapdragon X chips compete directly with Intel and AMD for the first time in the PC market.

Qualcomm’s contribution to AI has mostly been on device rather than in giant data centers. Its newer Snapdragon chips include a dedicated AI processor, called a neural processing unit, that lets phones and laptops run AI features like photo editing and voice assistants directly on the device rather than sending data to the cloud.

“Qualcomm’s Snapdragon platforms are designed to bring AI processing directly onto the device.”
 Cristiano Amon, president and CEO of Qualcomm, describing the company’s on device AI strategy in public remarks and earnings calls

Microsoft

Founded	1975, by Bill Gates and Paul Allen
Headquarters	One Microsoft Way, Redmond, Washington, United States
CEO	Satya Nadella, chairman and chief executive officer since 2014
Market cap	approximately $2.9 trillion (late July 2026)
Share price	approximately $389 per share

Microsoft sells software, above all Windows and the Microsoft 365 suite of Office applications, and runs Azure, one of the three largest cloud computing platforms in the world alongside Amazon Web Services and Google Cloud. Cloud computing now drives most of Microsoft’s growth and profit.

Microsoft owns the Xbox gaming business and the Activision Blizzard game publisher, the professional network LinkedIn, the developer platform GitHub, and its own line of Surface computers and tablets.

Microsoft has made one of the largest bets on AI of any company in the world, investing tens of billions of dollars in OpenAI and building its Copilot AI assistant into almost every product it sells. It plans to spend around $190 billion on AI infrastructure in 2026 alone, mostly on data centers packed with NVIDIA and AMD chips, though it is also designing its own AI chips, called Maia, to reduce its dependence on outside suppliers.

“A control plane that extends companies’ existing governance, identity, security, and management frameworks to agents.”
Satya Nadella, CEO of Microsoft, describing the company’s Agent 365 platform on a 2026 earnings call

Alphabet (Google)

Founded	Google was founded in 1998 by Larry Page and Sergey Brin; Alphabet was created in 2015 as its parent holding company
Headquarters	1600 Amphitheatre Parkway, Mountain View, California, United States
CEO	Sundar Pichai, chief executive officer of both Alphabet and Google since 2015 and 2019 respectively
Market cap	approximately $4.07 trillion (late July 2026)
Share price	approximately $324 to $334 per share

Alphabet’s core business is Google, built primarily on internet search and the advertising that runs alongside it, along with YouTube, the Android operating system and the Chrome browser. Google Cloud, its cloud computing arm, competes directly with Microsoft Azure and Amazon Web Services and has become its fastest growing major segment.

Alphabet’s Other Bets segment houses more speculative businesses, including the self driving car company Waymo, the life sciences firm Verily, and its venture capital arms CapitalG and GV.

Google was one of the earliest inventors of the modern AI techniques that power today’s chatbots, and it builds its own custom AI chips, called Tensor Processing Units or TPUs, as an alternative to buying NVIDIA GPUs. Its Gemini family of AI models competes directly with OpenAI’s ChatGPT. Alphabet raised its own 2026 capital expenditure guidance twice during the year, most recently to between $195 billion and $205 billion, reflecting surging demand for AI computing power through Google Cloud.

“What the technical customers want is control over their compute, their models, their data stack, and their alpha.”
A line from Palantir CEO Alex Karp, quoted approvingly by Microsoft’s Satya Nadella, that has become a touchstone in the wider industry debate over who should control AI infrastructure, a debate in which Google’s in house TPU strategy is often cited as the leading example

Amazon

Founded	1994, by Jeff Bezos
Headquarters	410 Terry Avenue North, Seattle, Washington, United States
CEO	Andy Jassy, president and chief executive officer since July 2021, previously the founding head of Amazon Web Services
Market cap	approximately $2.5 trillion (late July 2026)
Share price	approximately $232 per share

Amazon is best known as the world’s largest online retailer, but its most profitable business by far is Amazon Web Services, usually shortened to AWS, its cloud computing division. AWS rents out computing power, storage and databases to businesses and governments around the world and, despite being a smaller share of total revenue than the retail business, generates the majority of Amazon’s operating profit.

Amazon runs a large advertising business alongside its retail marketplace, the Prime subscription service bundling shipping and streaming video, its own logistics and delivery network, and devices like the Alexa voice assistant and Kindle e-reader.

AWS is one of the three dominant cloud platforms feeding the AI boom, and Amazon is projecting around $200 billion in capital expenditure for 2026, most of it aimed at AI data centers. Amazon has also designed its own AI chips, called Trainium and Inferentia, to reduce how much it needs to spend on NVIDIA hardware, and it offers a range of AI models to customers through a service called Bedrock.

“He is a member of the National Academy of Engineering.”
Biographical language commonly used to describe Andy Jassy, CEO of Amazon, though Jassy is better known within the industry for building AWS from an internal side project into Amazon’s most profitable division over nearly two decades

Broadcom

Founded	The current company traces to a Hewlett-Packard semiconductor division from 1961; the Broadcom Corporation whose name it carries was founded in 1991 by Henry Samueli and Henry Nicholas; Avago Technologies bought Broadcom Corporation in 2016 and took its name
Headquarters	3005 Deer Creek Road, Palo Alto, California, United States
CEO	Hock Tan, president and chief executive officer since 2006, first at Avago and then at Broadcom
Market cap	approximately $1.6 to $1.8 trillion (late July 2026)
Share price	approximately $310 to $355 per share

Broadcom designs a very wide range of semiconductors, including networking chips that route data inside data centers, broadband and wireless connectivity chips found in most smartphones and routers, and increasingly, custom AI accelerator chips built to order for a handful of huge customers.

Broadcom’s other major business is infrastructure software, built through acquisitions of CA Technologies, Symantec’s enterprise security business, and VMware, which it bought in 2023 for $69 billion. This software division sells mainframe, cybersecurity and virtualization tools to large enterprises.

Broadcom has become one of the most important companies in AI hardware without making a single chip under its own name for the general market. It designs custom AI accelerator chips, called ASICs, for a small number of huge customers, most notably Google, for its TPUs, and reportedly Meta and OpenAI as well. This custom silicon approach is the leading alternative to buying general purpose GPUs from NVIDIA, and it has driven Broadcom’s market value past one and a half trillion dollars.

“I think potentially it’s over $100 billion TAM market.”
  Hock Tan’s Intel counterpart, CEO Lip-Bu Tan, used nearly identical language to describe the custom AI chip opportunity that both Intel and Broadcom are chasing; Broadcom’s own Hock Tan has separately told investors that custom AI silicon is central to the company’s next decade of growth

The AI spending divide

Among these thirteen companies, four stand out for how differently they are approaching AI spending. Microsoft, Google and Amazon are each planning to spend on the order of $190 billion to $200 billion in 2026 alone on AI infrastructure, mostly data centers filled with chips from NVIDIA, AMD and their own custom silicon. Apple, by contrast, has kept its own capital spending comparatively small, choosing instead to build AI features into its devices and lean on outside partners for the heaviest computing work.

Chart 3. Planned 2026 capital expenditure, Amazon, Alphabet and Microsoft compared with Apple’s far smaller outlay.

Markets have started to reward Apple for this restraint rather than punish it for falling behind. Through most of 2026, investors grew nervous that the hyperscalers’ AI spending was outrunning any proof that it would pay off, and Apple’s comparatively cautious approach helped its shares outperform its heavier spending peers, briefly making it the most valuable public company in the world.

What this means, in plain terms

Strip away the ticker symbols and this is a story about four kinds of company. There are the toolmakers, ASML above all, who sell the impossibly precise machines needed to manufacture anything at the cutting edge. There are the builders, TSMC, Samsung and Intel, who take those tools and physically produce the chips. There are the designers, NVIDIA, AMD, Apple, Qualcomm, Broadcom and Arm, who draw the blueprints without necessarily building anything themselves. And there are the spenders, Microsoft, Google and Amazon, who buy everything upstream in colossal quantities to build the AI services the rest of the world now depends on, while trying to design their own chips so they eventually need to buy less.

No single company in this report can succeed alone. NVIDIA needs TSMC to build its chips. TSMC needs ASML to build its factories. Arm needs Apple, Qualcomm and Samsung to license its designs. And every one of the big spenders needs all of the above, even as each tries to become a little less dependent on it. That interlocking need, more than any single earnings report or stock price, is the real story behind the numbers in this report.

Stay ahead of the stories shaping our world. Subscribe to Impact Newswire for timely, curated insights on global tech, business, and innovation all in one place.

Dive deeper into the future with the Cause Effect 4.0 Podcast, where we explore the ideas, trends, and technologies driving the global AI conversation.

Got a story to share? Pitch it to us at info@impactnews-wire.com and reach the right audience worldwide