Now if you’re looking for a bit more details, let’s dive deeper.
With the AI Boom, Nvidia’s stock is through the roof, because Nvidia is the main producer of hardware that enables AI model training. Fun fact, there are rumors that over 50% of Nvidia’s employees are worth millions 🤑, and the median average is about 50 Million per employee 🤑. If you were an Nvidia employee before their stock went through the stratosphere 🚀, I envy you.
Anyways, since money is thrown at Nvidia faster than they can spend them, they capitalize on the gains, and work real hard to deliver hardware that can feed the hunger of the AI software. So naturally they make a ton of new things. 🛠️
At CES 2025 (Consumer Electronic Show 2025) in Las Vegas, they showed-off on stage Project DIGITS, which is a small computer that can fit on anybody’s desk, that can comfortably run AI Models up to 200 Billion Parameters, currently that would be llama 70b which is 70 Billion params, or flux image processing model (12 billion params), this one runs on texti.app.
Note: You can connect 2 of these machines to double your powers.
Project DIGITS comes with a Blackwell GPU, which is an extremely crucial part of the build. We have previously discussed the benefits of it in this post. But in a nutshell the main benefit of this graphics card is that it facilitates building virtual environments to train models for specific needs. It could be meant for robots 🤖, that are going to be doing activities in the real world, and instead of training in the real world where breaking things can be expensive, all you have to do is reconstruct an isolated environment specific to your needs in the virtual world, train a model with the physics, and then later just install the model into the brain 🧠 of your robot and voila 💁, it knows how to act and what to do in the real world.
For example: you want a robot that knows how to assemble a car, you can train it what to do, where to put it, what to do in case something is not exactly as specified and then launch it into a supervisor robot that will handle all the exceptional use-cases. Magic 🪄
At the starting price of 3000 USD 🤑 this is not exactly expensive, given that with this hardware, anybody can now become an AI researcher 🧑🔬, train and tune models as per one’s needs and finally release a product into the wild. I honestly think this is brilliant tech, which will propel us into the future at the speed of sound.
Let’s see how things are going to look when this tech lands in the hands of actual developers and companies and how people are going to react.
This year looks extremely promising, we will see the raise of a ton of new AI tech, a bunch of new AI hardware, and we’ll still see how Nvidia’s stock goes over the moon.
Just to remind you, this year starts, with relatively accessible hardware, that allows you to train your own AI model. Isn’t this a promising bright future?
