The big swallow all competition.

Earlier in the week, it was announced that Nvidia was taking over Groq. It is doing so with the intention of eliminating a competitor. Unfortunately, this is not a true purchase, subject to scrutiny. Instead, Nvidia is basically taking the talent that it wants (like the CEO/founder) along with patents. What is leftover in Groq is basically a startup.

This is becoming a commonly used technique within the technology sector. Microsoft applied the same mechanism in 2024 when "acquiring" a company.

At present, there might not be a more important industry than semiconductor. If I had to point to one, perhaps energy since the two are closely linked. However, with the age of AI rapidly coming upon us, we are looking at the central infrastructure to the future.

Nvidia is the early leader. Obviously, it intends to remain there.

In this article I will look at some of what is taking place and how this does affect the future.

Nvidia Buys Groq

Groq is a specialized player in the semiconductor sector. Instead of focusing upon general purpose chips, which is dominated by Nvidia, it preferred to take a more specialized approach.

The companies main focus, through its LPU chips, is inference. Through the architecture, Groq's chips focus upon efficiency. It is able to produce better output while using less energy as compared to the Nvidia GPUs.

This results in faster throughput speeds while saving money on inference operations. As the use of AI grows, the cost of inference is something major corporations are concerning themselves with.

With the asset purchase, Nvidia is now placing this under its umbrella. All patents are now part of that company, allowing Jensen Huang free reign over the technology. Does he shelve it? That is up to him.

Whatever the decisions, we can see how a competitor was eliminated from the landscape. While Grow might never have rivaled Nvidia in terms of market cap, it did have the potential to pull some business (read traffic) away.

That is no longer the case.

Centralization

It is hard to argue the centralization of the AI world. We are talking about a big money industry. Nothing is priced in millions. Instead, we are looking at all numbers starting with a "B". Companies like Tesla, xAI, and OpenAI drop purchases in the billions.

Smaller players are almost non-existent in the West. China is taking a different approach since it is, for the most part, precluded from purchasing GPUs.

Big Tech is dominating the language model market. This might not be the end of the world since we are rapidly heading towards commoditization in that realm. Smaller models are growing in capability, enabling them to compete against the larger counterparts, at least within specialized realms.

However, when we consider the basics of AI, we see how the bigger companies dominate.

AI = compute + algorithms + data

The latter is something we discussed at length. Data is the ultimate in flywheels. As more is available, it can grow at a greater pace than those without. That is why each prompt is crucial to AI systems. It really becomes a numbers game.

Algorithms can be an equalizer. This is where the Chinese are focused. They create lighter stacks, mostly driven by the compute hindrance. The algorithms are designed to run more efficiently than the Western counterparts.

Compute is still the difference maker. Smaller entities simply do not have access to the resources (money) to load up on compute. Even if they were able to stitch the dollars together, it would be difficult. Nvidia is moving 100% of it produces. This company will take care of its larger customers (read Elon Musk and Mark Zuckerberg). They are the ones who keep returning with multi-billion dollar checks.

Nvidia taking over Groq is just another example of the consolidation that is taking place. It is why each individual has to look at what he or she is doing and ensure that steps are taking to decentralize as much as possible.