Semiconductors & Manufacturing

Designs the GPUs and accelerated computing platforms used for much of modern AI training and inference, and sells the networking that ties them together.

The hardware is only half of it. CUDA, the software libraries built on top of it, and a very large developer base make NVIDIA systems the default choice for teams who want to ship rather than port.

Custom accelerators built by its own largest customers, a stronger AMD, revenue concentrated in a handful of buyers, export restrictions, and model architectures that shift the balance of demand.

Designs CPUs and the Instinct family of AI accelerators, sold as an alternative to NVIDIA at the data centre scale.

The credible second source. Buyers who dislike depending on one supplier have somewhere to go, and that alone shapes pricing conversations across the industry.

Software maturity remains the harder problem than silicon. Winning benchmark results does not automatically move workloads that were written against a rival's tooling.

Designs and manufactures CPUs, and is attempting to build a contract manufacturing business for other companies' chip designs.

The only Western firm seriously attempting leading edge logic manufacturing at scale, which makes it strategically interesting to governments quite apart from its commercial results.

Foundry economics are brutal. The turnaround requires capital, process execution and external customers arriving at roughly the same time.

Manufactures chips designed by other companies, including most leading edge AI accelerators, and provides the advanced packaging that binds logic dies to memory.

Capacity at the leading node is scarce and allocated years ahead. Packaging capacity has at times been a tighter constraint than the transistors themselves.

Geographic concentration in Taiwan, the cost of overseas fabs, and the cyclical nature of committing tens of billions to capacity ahead of demand.

Builds the lithography systems used to print circuit patterns onto silicon wafers, including extreme ultraviolet machines.

ASML is currently the world's only supplier of production EUV lithography systems, which are required for manufacturing many of the most advanced chips.

Export controls limit which customers it can serve, orders are lumpy, and demand ultimately tracks the capital spending decisions of a small number of chipmakers.

Manufactures memory, including high bandwidth memory stacks that sit alongside AI accelerators.

An accelerator can only compute as fast as it can be fed. HBM supply has repeatedly been the gating factor on how many AI systems can actually be built in a given year.

Memory is historically cyclical. Capacity added into a boom has a habit of arriving into a downturn.

Manufactures DRAM, NAND and high bandwidth memory used in AI servers.

One of a very small group able to produce HBM at volume, which turns a commodity business into a partly allocated one while demand runs hot.

Pricing power fades quickly once supply catches up, and HBM qualification cycles with accelerator vendors are demanding.

Manufactures memory including HBM, runs a contract chip manufacturing business, and builds consumer devices.

The third HBM supplier, and the swing factor in whether memory stays tight. Its foundry is also the main alternative to TSMC at advanced nodes.

Qualification delays on new memory generations and a foundry that has struggled to match its main rival on yield.

Co-designs custom AI accelerators for hyperscalers and supplies much of the switching and connectivity silicon inside data centres.

It sits on both sides of the custom silicon story. When a cloud provider builds its own chip to reduce dependence on merchant GPUs, Broadcom is often the partner making that possible.

Custom programmes are concentrated in a few customers who can change direction, and the software business carries its own integration risk.

Builds Gemini models through Google DeepMind, designs TPU accelerators, operates Google Cloud, and distributes AI through Search, Workspace and Android.

It is one of the few organisations that designs the chip, runs the data centre, trains the model and owns the consumer surface it ships on.

AI answers can cannibalise the advertising business that funds everything else, and antitrust proceedings could reshape distribution.

Runs AWS, designs Trainium and Inferentia accelerators, hosts third party models through Bedrock, and invests in Anthropic.

The largest cloud business is trying to own its own accelerator roadmap rather than rent it, which is the clearest example of the custom silicon threat to merchant GPUs.

Custom silicon needs software adoption to matter, and cloud growth is now compared against very large numbers.

Runs Azure, distributes AI through Microsoft 365 and Copilot, partners commercially with OpenAI, and designs its own Maia accelerators.

Few businesses touch as many parts of the ecosystem at once. It owns infrastructure, a commercial route to frontier models, and the enterprise distribution to sell the result.

Very heavy capital spending against uncertain payback, dependence on partner model roadmaps, and buyers questioning per-seat AI pricing.

Develops the Llama model family and deploys AI across its own products, while designing MTIA accelerators for internal workloads.

Meta's Llama is an open-weight model family. Its weights can be downloaded and run independently, although Meta's licensing terms mean it is not considered open source under the strict Open Source Definition. Releasing capable weights commoditises a layer that rivals sell.

Very large capital spending with returns that arrive through advertising rather than direct AI revenue, and licensing scrutiny.

Designs data infrastructure silicon including optical interconnect, custom compute and storage controllers.

Moving a signal off a chip, across a rack and between buildings is its own engineering discipline. Marvell sells into the parts of that path most people never think about.

Custom programme timing is lumpy and dependent on a handful of hyperscaler decisions.