One of the overlooked businesses of Tesla that has enormous potential is Dojo.

While many focus upon Robotaxis, Semis, and a more affordable vehicle, Tesla could be asserting itself in the chip market.

This is something that, considering the AI era, is crucial. Right now, Nvidia is the dominant player. However, Tesla could be in position to garner market share (internally) simply by scaling Dojo. This is something that Musk says could happen towards the end of this year.

We are looking at D2 emerging. The forecast is D3 will be available in 2026.

So what is this overlooked business and what could it do for Tesla?

Tesla: Trillions In Market Cap With Dojo

Nvidia is a name that most know these days. That was not the case 5 years ago. The push into AI has put the company at the forefront of every discussion. You cannot have AI with chips. GPUs dominate the space and Nvidia really has no rivals.

Tesla is getting into the game, albeit with a different slant. The business model is more in line with Google than Nvidia.

Google powers their AI platforms using TPUs. This is a different technology than what Nvidia offers, one that is controlled by Google. For this reason, not only does the company have access to the chips it needs, it also does not feed into the profitability of Nvidia.

According to their filings, the GPUs have roughly a 75% profit margin for Nvidia.

With such a margin, Tesla doesn't need to meet the performance of Blackwell. If their performance is 1/3-1/2 that of Nvidia, the company will save billions. Instead of spending $40K on each Nvidia GPU, Tesla has a cost of $10K. On a cost per performance basis, this is a 50% savings if they achieve the 1/2 performance rate.

According to TSMC, D2 is already in production, fitting the timeline that Elon laid out.

Another advantage is the networking between the chips. With Dojo, Tesla has this down to microseconds. That allows a large cluster of chips to operate coherently.

Built-In Customer

It is likely we see integration between Grok and Tesla. This is the LLM put forth by xAI. The collaboration of the two companies means an AI structure that is designed specifically for the needs of each company.

There are reports that Dojo and xAI will have compatibility. Each is being built for the other. This is in addition to what is taking place with FSD. More of that system is being handled by Dojo, although it is still a small percentage.

xAI is also one of the largest customers for Nvidia. They have purchased more than 200K chips. Couple that with the 100K Tesla bought and we can see some big checks flowing into the coffers of Nvidia.

Jensen Huang, the CEO of Nvidia, loves selling to Tesla and xAI. The reason is not profit margin or personality. Instead, it is simply they get the data centers running quickly. This is to Nvidia's benefit.

There was a 100K cluster set up in Memphis by xAI that was done in under 100 days. In contrast, Meta took more than a year to get their cluster of H200 chips running.

That said, Nvidia doesn't necessarily need Elon's business. They will continue to sell him chips as he cannot meet all his needs. However, with the AI race taking place, I feel certain Huang will have a lot of people knocking on his door.

If D2 is indeed a 10x in performance of D1, that would put it on par with the first generation Blackwell chips. Simple math shows that a 75% reduction in cost yet similar performance would net out to billions of dollars saved.

It would also allow for the possibility of scaling at a rate faster than any other company.

The second half of the year will lead us to finding out more about D2. Naturally, until it is in the field under real conditions, the performance is nothing more than speculation.