What's Important About GPUs?
As Nvidia's financial numbers continue to impress, let's look at why GPUs are important in the AI industry.
On Wednesday 28 August, the markets eagerly anticipated Nvidia’s Q2 Earnings call to gain insights into how the business is performing. With a market capitalisation of $3.09 Trillion, it’s fair to say it’s doing well. The global chip maker reported revenues of $30 billion for the 2nd quarter, which is up 15% on the last quarter and up 122% on the same period a year ago. Those numbers are truly staggering. Nvidia has been a successful business for a long time, but generative AI and the demand for increased computer processing power have catapulted it to the next level and made Nvidia the largest company in the world, recently overtaking Microsoft and Apple.
What’s Important About GPUs?
Amongst other products, Ndvida produces a computer chip known as a Graphics Processing Unit or GPU. GPUs were historically used in computers to power graphic applications such as games and image/video editing tools. However, their ability to process mathematical calculations extremely quickly has made them ideal for generative AI and the increasingly complex models that are being created.
AI to U is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.
For example, I have a machine learning platform installed on my Mac. With this program, I can easily create a range of ML models, such as clustering or time-series forecasting. What I cannot do, however, is train a deep learning model, as my laptop does not have the computational power to do the required calculations. To do so, I’d need to hand this off to another environment (such as Amazon EC2) with available GPUs attached to do the computation.
And here’s where AI can be expensive. On the one hand, we have a huge demand to buy GPUs themselves where the simple economics of supply (low) and demand (high) dictates high prices. And on the other hand, we have the costs of running these solutions. Infrastructure fees can be extremely costly. Compute costs are almost like a tax for running an AI project. Anyone who’s played around with an LLM via API might have already learnt how quickly the costs can wrack up.
Is It Only Nvidia That Make GPUs?
The good news is that it’s not only Nvidia that make GPUs and this can only be a good thing. Not only does competition drive innovation, it can also help reduce prices. AMD is another leading manufacturer of computer chips, and in a sign that it’s not given up, it recently acquired ZT Systems in a $4.9bn deal to challenge Nvidia’s dominance.
If you have a friend or colleague who might be interested in these articles, please don’t hesitate to share!
Until next time.
AI to U is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.
