Nvidia and the Chipmakers Powering the AI Boom

Samsung Electronics · $72.5B From Memory and Mobile

Samsung Electronics earns $72.5 billion from a chip business built on volume rather than novelty: DRAM and NAND flash memory that ships inside nearly every laptop, smartphone and server sold worldwide. Alongside memory, Samsung runs its own foundry and designs the Exynos processors that power select Galaxy phones, making it one of the only companies that both manufactures memory at scale and competes with TSMC for foundry contracts.

Samsung's memory chips are largely invisible to consumers — nobody buys a laptop for its DRAM brand — but the company's manufacturing scale gives it real pricing power over memory cycles that affect every other device maker's costs, from PC builders to smartphone assemblers.

SK Hynix · $61.0B Riding the HBM Memory Boom

SK Hynix's $61.0 billion in revenue tells the story of how AI reshaped the memory business. The South Korean company is the leading producer of High Bandwidth Memory (HBM), the specialized memory stacked directly next to AI GPUs like Nvidia's H100 and H200 to feed them data fast enough to keep up with AI training workloads. Without HBM, Nvidia's chips would be bottlenecked no matter how powerful their processing cores are.

That positioning has made SK Hynix one of Nvidia's most important suppliers rather than just a competitor in the broader memory market. Standard DRAM and NAND still make up a large share of its business, but HBM's higher margins and AI-driven demand are why SK Hynix's memory revenue has climbed past longtime rival Samsung in some product categories.

Intel · $53.0B Rebuilding Around Its Own Foundry

Intel's $53.0 billion in revenue comes from the company that has powered the majority of the world's laptops and servers for decades through its Core and Xeon CPU lines. That dominance has narrowed as AMD and Arm-based chips gained ground, but Intel remains the default processor in most Windows PCs and a major force in datacenter CPUs, even as its share of AI-specific compute lags far behind Nvidia's.

Intel's bigger bet is Intel Foundry, its push to manufacture chips for other companies the way TSMC does, turning its own fabs into a second revenue stream. It's also selling Gaudi AI accelerators as a lower-cost alternative to Nvidia's GPUs, though neither effort has yet closed the gap with the AI chip leader.

Broadcom · $50B+ From Networking Chips and Custom AI Silicon

Broadcom's $50 billion-plus in revenue comes from two very different businesses under one roof: networking chips — the switches and routers that move data inside AI datacenters — and custom AI silicon designed for specific hyperscale customers rather than sold as off-the-shelf GPUs. That custom-ASIC business has reportedly included work with Google on its in-house AI chips, positioning Broadcom as a path into AI hardware that doesn't run through Nvidia at all.

Broadcom also makes the wireless and RF chips found in iPhones and other smartphones, giving it a foothold in mobile hardware most people never see mentioned. That mix — invisible infrastructure chips plus custom AI silicon — is why Broadcom's revenue keeps climbing without a single consumer-facing product to its name.

Qualcomm · $40B+ From Snapdragon and Cellular Modems

Qualcomm's $40 billion-plus in revenue comes almost entirely from chips that never carry the Qualcomm name to consumers. Its Snapdragon processors power the majority of Android flagship phones, and its modem chips handle cellular connectivity in devices from multiple brands — including, for years, the iPhone itself. That makes Qualcomm the closest thing the smartphone chip market has to a default supplier.

The company is now pushing into new territory with Snapdragon X Elite chips built for Windows laptops, aiming to challenge Intel and AMD on their home turf, plus automotive chips for connected and self-driving vehicle systems — a bet that mobile-style efficiency matters as much outside phones as inside them.

Micron · $20B+ Chasing the AI Memory Market

Micron Technology's $20 billion-plus in revenue makes it the only major memory chipmaker headquartered in the United States, competing directly with Samsung and SK Hynix in DRAM and NAND flash. Its chips ship inside PCs, smartphones, servers and increasingly AI systems, where Micron has been racing to catch up in HBM3E memory to avoid ceding the entire AI memory market to its South Korean rivals.

Memory pricing is notoriously cyclical, and Micron's revenue swings harder with those cycles than most companies on this list — a reminder that even companies supplying the AI boom aren't insulated from the boom-and-bust economics that have defined the memory business for decades.

AMD · $20B+ Building Nvidia's Closest AI Rival

AMD's $20 billion-plus in revenue spans three product lines: Ryzen CPUs for consumer PCs, EPYC processors that have taken meaningful server market share from Intel, and Radeon graphics cards for gamers. All of it is manufactured by TSMC rather than AMD's own fabs, since AMD sold its manufacturing arm — now GlobalFoundries — back in 2009 to focus purely on chip design.

AMD's MI300 series is its answer to Nvidia in AI accelerators, and it's the closest thing to a credible alternative on the market — but Nvidia's CUDA software ecosystem still gives it a head start AMD hasn't fully closed, which is the main reason AMD's AI chip revenue remains a fraction of Nvidia's.

Texas Instruments · $15-18B From Chips Nobody Notices

Texas Instruments reports $15 to 18 billion in revenue from chips most people have never heard of and will never buy directly: analog and embedded processors that manage power, sense temperature and pressure, and control basic electronics inside cars, industrial equipment and appliances. It's the least AI-driven business on this list, and one of the most stable — TI's chips don't chase hype cycles.

That steadiness is the point: TI's products go into things that need to work reliably for a decade or more, from factory sensors to automotive safety systems, making it less exciting to hardware enthusiasts but arguably more essential to the physical world running smoothly.

For most people buying hardware specifically for AI work — training models, running local LLMs, or heavy machine-learning workloads — Nvidia is still the safe default, purely because of CUDA's software support; almost every AI framework and tutorial assumes an Nvidia GPU first. AMD's Radeon and MI300 chips are a legitimate, cheaper alternative if you're comfortable working around occasional software gaps, and Apple Silicon's unified memory makes it a strong pick for running smaller models locally on a laptop. Intel and Qualcomm remain better bets for general-purpose computing than for AI-specific acceleration, at least for now.