NVDA: What's a good brain worth?

What does NVDA sell? It sells brains. It sells brains that go into machines. One example is a car. But a car is a subset of an autonomous vehicle. A car is thought of as a vehicle that transports humans around on the ground. An autonomous vehicle can be anything that can move around on its own, not just a a box that moves human around.

So what is the brain of an autonomous car worth? Jensen Hwang might know a thing or two have the value of a brain that enables an autonomous car. Here’s what he said during yesterday’s Q&S session of the earnings call:

I think for autonomous vehicle that still have drivers passenger cars, branded cars, ASP anywhere from $500 to $1,000 makes sense. For robot taxies where they are driverless, they are not autonomous vehicles they are actually drivers less vehicles, the ASP will be several $1,000. And in terms of timing, I think that you’re going to see a larger and larger deployment starting this year, and I’m going through next year for sure especially with robust taxies. And then with autonomous vehicles, cars that have autonomous driving capability, automotive driving capability start late 2019, you could see a lot more in 2020 and just almost every creating car by 2022 will have autonomous automatic driving capabilities.

So how much money can NVDA make off of autonomous cars? There are 77 million cars sold be year. If all cars were autonomous then we wouldn’t need as many cars. Let’s cut the annual number in half. Eventually all cars will be fully autonomous. NVDA is working with EVERYBODY…320+ partnerships in the automotive sector. OK is NVDA received $2000 for the brain of each autonomous car that’s 38.5 million brains times $2000 or $77 billion PER YEAR in revenue. That’s the potential addressable market. The real number would be some fraction of that but even so we’re only talking about autonomous cars. There’s still autonomous everything else. Forklifts, all kinds of drones, maybe autonomous pollinators for agriculture, autonomous vehicles to mine asteroids, and who knows what else…Anything that moves and needs its own brain. The potential is enormous. NVDA’s annual revenue is now about $10B.

So at 60%+ gross margins and potential annual revenue of $77B or almost 8x their current revenue, what the potential valuation of NVDA? Can NVDA become the most valuable company in the world? In 10 years it may well be there. So what is a good brain worth?



The driver less use brings in the higher end “several 1000s” and the more mass deployed will probably be in the “$500-$1000” category. But regardless you make a compelling case for huge revenue just from the car segment. Nvidia is absolutely dominating this market right now. Nearly everyone who is in this space is using Nvidia and those that aren’t “partnered” will still need to build data centers to train their networks. There’s only one place to go if you want that done.

Good point about it’s not just cars. Millions of other vehicles will include this technology. The knowns for Nvidia’s TAM are vast but the unknown markets where AI and DL will penetrate or create entire industries could be even more significant.

Data is this centuries oil. A thing that exists all around us and has so many uses that will improve humans lives. When they first discovered oil did the people realize all the uses that would come out of it. It’s not just fuel for a light or gas for a car. And for every business or industry Nvidia will be a player as to find the usefulness of that data.

There’s some political speak about antitrust in this article but it has some good info.


Hi Chris, Somehow I get the impression that you like Nvidia as an investment. :grinning: You guys have convinced me. I add a little to my position steadily, and it’s up to a little over 8% now. Thanks for your great write-ups.


Somehow I get the impression that you like Nvidia as an investment.

I have put my money where my mouth is.

TTM adj EPS is $4.92
Stock is at $230 so the TTM P/E is 46.7.

TTM 1 yr growth is 60% so the 1YRPEG is 0.78.

1 year revenue growth is 41%.

Gross margins are expanding with next quarter’s non-GAAP GM guidance at 63%.

Data center growth was 105% this past quarter.

Automotive revenue is still all the infotainment and autonomous development. Autonomous manufacturing revenue will dwarf the autonomous development revenue and it won’t kick in until a couple of years from now.

AI revenue is almost all based on centralized AI. The big opportunity will be distributed AI (what some call AI at the edge). This is why I own NVDA. This is where the big bucks will be in the future. They will sell brains. A brain is the most valuable thing there is. What would each of us be without a brain? Nothing.



^^ What Chris said. NVDA is my largest holding for a reason, even if I did “lose faith” a little that one time in 2016 and sell the 4-bagger. The potential opportunity space is huge, and their next nearest competitors (the ones we know about today anyway) don’t seem to be able to catch up… and if a small start-up starts doing something awesome-er, they can just buy them.


Thx for the additional insights and confirmation Chris… now 9% of my portfolio.

By the way… I think Starfish don’t have a brain or a heart for that matter… autonomous bottom feeders though and very pretty:-)))

1 Like

Great numbers regarding the potential TAM. The potential really does look like it could be massive and growing!

So their nearest competition, AMD, is years behind them, right?

For datacenters and anything centralised, they will always be wanting the best. As data usage, bandwidth and inter-connectivity increases, so will the need for faster processing power. Particularly if all this power is in big data centers and the cloud.

But this ‘edge’ market. Autonomous vehicles, drones, vacuum cleaners. Doesn’t there come a point where the 2nd best is good enough? Drawing comparisions with ambarella’s chips for cameras, and invensense’s accelerometers. They had the best chips, winning prizes all the time, but others caught up and became ‘good enough’. End users didn’t require the best, they just required good enough at a reasonable price.

I know this is years and years and years away from even possibly happening but will the time come where GPU’s are good enough for edge use? Will NVIDIA’s moat remain with its CUDA language? By the time good enough happens, will the CUDA language be so entrenched in the world that no-one involved with AI considers learning or using any other language? Will it be good enough to stave off competition?


By the time good enough happens, will the CUDA language be so entrenched in the world that no-one involved with AI considers learning or using any other language? Will it be good enough to stave off competition?

Very good question…and I wonder if NVDA becomes so dominant with their pioneering of CUDA over the past 8+ (maybe 12+) years, whether governments in the future might end up trying to force a “tower of babble” situation to try to have AI languages that could compete with CUDA just to keep NVDA from being the far and away single most important company on the planet.

With today’s rise and other stocks being down in the past 2 weeks, NVDA now makes up over 28% of my portfolio.

1 Like

One question one might ask is, has there ever in the history of software been a language which completely dominated a problem space so that it effectively had no competition?

Even with something like real time apps which are heavily dominated by model-based design, there are at least three languages in common use and there are a number of differing modeling systems.

Expecting CUDA to dominate forever may be optimistic.

1 Like

Expecting CUDA to dominate forever may be optimistic.

And may eventually bring up anti-trust concerns…but probably not for a decent while longer.

has there ever in the history of software been a language which completely dominated a problem space so that it effectively had no competition?

That is a very interesting question. I think so and it the language of life: DNA. All life uses the same language and it is completely dominating the planet, changing it more than anything else. With Elon Musk’s efforts that language will soon leave the planet and dominate other planets. Will DNA control CUDA or will CUDA replace DNA? AI will undoubtedly become self-perpetuating. AIs will eventually be able to take resources to build machines to gather more resources to build more machines. Meantime, humans are learning to edit the language of DNA to correct defects and eventually add new functions. Will DNA integrate CUDA or will the remain separate? Or will there be another AI language?



Or will there be another AI language?

Now that to me is the 64 million dollar question. Yes!! There will always be another something. Our job is to find it early.



A lot of discussion/arguments will probably ensue from your last post. Some will argue about he vast difference between DNA and programing languages such as CUDA. that is getting in the weeds on a brilliant macro level thought.

1 Like

CUDA is no DOS/Windows, neither is it x86. CUDA is sticky, it creates switching costs, but it can be emulated. Google has its own language, and it can be used. There is an open source language for GPUs, that by all accounts is pretty dang good, that shocking! NVDA maintains for the world. But NVDA seems to make sure it is always a year behind, and it is not as good, but it can get the job done.

If something materially better than NVDA GPUs is found, there are a host of languages that can be put forward to become a new standard. But that is the thing, JUST AS GOOD, OR SOMEWHAT BETTER, is not going to do it. It has to be much better to knock aside all the infrastructure, knowledge, habit, standardization, supply chains, knowledge basis, etc.

It can be quickly overcome sometimes by surprise, but almost every analyst indicates that NVDA has at least a 1 year lead, I believe NVDA says at least a 2 year lead. AMD proved that it is at least a product cycle or more behind NVDA and will need to make a great leap if their next generation product is to catch up. Doubt it, AMD is unlikely to be able to keep up with product cycles.

I believe that for machine learning, other than what Google might try to do with its next generation tensor, that no one is trying to take on NVDA on the machine learning side of things. Where the battle is going on is the inference level of things. For that I have no information as to marketshare.

Google uses its tensor, NVDA uses its tensor and GPUs, Intel is putting out improved CPUs and FPGAs…another huge market to be fought over. Will it be an equal Gorilla market?

Quite easy to buy your GPUs for machine learning, and then take the volume discount and buy more NVDA product for inferencing. We shall see.



<<<Google reportedly has been trying to interest China’s tech companies in its TensorFlow AI software tools, which make it easier to develop apps. But it’s not clear if those talks have involved its AI chips as well. Google’s cloud-computing service rents access to its TPU chips optimized for TensorFlow.>>>

An example of Google trying to get in on the CUDA train with its own standard. I imagine it is to set itself apart from the other cloud titans for renting AI space, and Google’s TensorFlow AI software is unique to Google. Not sure what any advantages would be there other than it is Google.

<<<Venture-capital investment in AI chip startups is soaring, says CBI Insights. While Nvidia’s GPU chips have grabbed an early lead, startup AI companies are focused on developing chips designed from scratch to crunch artificial-intelligence software. They include U.K.-based Graphcore, KnuEdge and Cerebras Systems.>>.

This is the sales pitch for new AI start ups. We can build it better because we are building from scratch. Which I think ignores the fact that so has Nvidia. Nvidia did not just spend billions of dollars tweaking its historical graphic intended GPUs. And Nvidia will continue to make them better. This might also explain why, despite the hype and money so far, no one has seen any of these new AI chips in action, anywhere that I can find. Given, I have not looked very hard, but a thing like this would normally come to one’s attention and not be a secret.

We shall see.



With watching Falcon Heavy videos on YouTube and having previously watched some NVDA-related videos, I got the following Jensen Huang video from 2014 as a suggestion. In it, he shows how they modeled the photo of Buzz Aldrin descending from the lunar lander Eagle in July 1969.


It doesn’t directly relate to this thread specifically, but could give a glimpse of Jensen Huang and his presentation skills from 2014.