The Layers of the AI Chip Trade
When a promo says “AI Super Chip,” it is usually pointing at one small corner of a much bigger stack. Before you evaluate any single name, it helps to see the whole map, because the money in AI chips flows through several distinct layers that behave very differently from one another.
The first layer is the designers, the companies that draw the actual silicon. The second is the foundries, the manufacturers that turn those drawings into physical chips. The third is a layer most retail pitches skip entirely: the electronic design automation, or EDA, companies that build the software the designers use to draw the chips in the first place. The fourth is the challengers, the newer architectures trying to leapfrog the incumbents on a single headline benchmark.
Each layer has its own economics and its own distinct risk profile. The AI Super Chip teardown walks through the specific claims behind this promo, and it is worth reading before you try to map any individual stock onto the story.
Designers and Foundries: The Obvious Layer
This is where the headlines live. Nvidia sits here as the dominant designer of the graphics processing units, or GPUs, that power most model training today, and its stock is the easiest shorthand for the entire AI hardware trade. The Nvidia GPUs piece explains why its architecture became the default and how the training workload shaped that outcome over several years.
Foundries such as TSMC sit one step downstream, turning finished designs into physical chips. Their margins depend on manufacturing scale, equipment utilization, and process leadership, not on any single design win, which makes them a very different kind of bet than the designers.
The problem for most investors is that this layer is crowded and fully priced. The obvious names carry obvious multiples, which is precisely why so many promos spend their energy pushing readers one layer deeper, where the names are smaller and the story feels fresher.
EDA: The Tools That Build the Chips
This is the layer the promo actually hands out. Synopsys (SNPS) is one of the two dominant electronic design automation companies, the firms whose software is used to design nearly every advanced chip on the market today.
EDA is a deceptively strong business. Chip designers simply cannot build a modern processor without these tools, and the switching costs are enormous once a design flow is locked in. That gives the incumbents steady, high-margin, subscription-like revenue that sits upstream of every AI accelerator, whether it is Nvidia’s GPUs or Cerebras’ wafer-scale engine.
Think of EDA as the picks and shovels for the picks and shovels. You do not have to guess which chip architecture wins the next cycle; you get paid on the total volume of chip design work, and that volume only rises as AI keeps demanding more and more silicon.
The Wafer-Scale Challengers
The fourth layer is the challengers, and this is where the marketing gets loud. Cerebras builds a single enormous chip across an entire silicon wafer rather than many smaller chips wired together, and it is the company behind the “50X faster than Nvidia” claim that anchors this promo.
The honest read is that those figures are benchmark-specific inference numbers, not a measure of general capability. The architecture itself is genuinely different, but a narrow benchmark win does not make a challenger the default buyer of EDA tools, and it does not unseat the incumbent designers. It mostly demonstrates that the chip trade has more than one lane.
Where Synopsys Fits in the Map
That is the real value of the map. Synopsys is not a chip designer betting on a single architecture. It is the toolmaker that gets paid no matter which design wins, because every design, incumbent or challenger, still has to be drawn in EDA software before it can be manufactured.
For a reader deciding whether the “AI Super Chip” pitch makes sense, the map is more useful than any single ticker. The AI inference vs training piece explains the workload split that sits underneath the whole story, and it is the right next read if you want to understand why inference claims and training demand are really two different businesses.
What to Watch Across the Layers
If you want to follow the map rather than any single pitch, the signals sit in different places for each layer. For the designers, watch data center spending and the cadence of new chip generations. For the foundries, watch capacity additions and utilization. For EDA, watch the volume of new chip designs, because every tape-out means revenue regardless of whose chip it is.
The challengers are the hardest to read, because their numbers move on announcements and benchmark claims rather than steady fundamentals. That is where the marketing lives, and where a skeptical reader should demand the most proof.
The EDA layer is the calmest of the four. Its revenue follows design activity, which rises with the entire AI buildout, and it does not require you to pick a winner. That is the real reason a toolmaker like Synopsys keeps showing up in pitches like this one. It is a durable, high-margin business disguised as an exciting chip story.
Ready to see the research? Click here to access the Brownstone Research report.
NewsletterVetter is an independent publication. We receive compensation from some of the services we review through affiliate links. Nothing on this site is investment advice. Always do your own research.