Nvidia Q1 2023

Nvidia released its Q1 2023 numbers and they look better than expected. More importantly their forecast for Q2 2023 is much higher than expected.

As usual Wall street continues to forecast expectations and penalizes (or rewards) companies when they miss the expectations. To be fair a lot of companies indulge them by providing their own expectations.

Like I mentioned in one of my previous articles, companies implementing ML models don’t have any strong moat. Companies with high quality training data can probably develop more accurate models and can improve their existing products, but models themselves are not products.

With recent hype of generative AI (and previous hypes of NN, crypto mining) companies that benefit are the ones that make GPUs (or any accelerated computational processing units) that help with training and inferences. Right now Nvidia is in a best position to capitalize this new demand for providing GPUs for generative AI. AMD can benefit from this too, but AMD is much more focused on gaming segment rather than AI / ML segment. AMD’s approach with their Instinct MI300 is to develop a single chip that contains CPUs, GPUs and (huge,) memory. I am not sure if this approach will support ML workloads.

It will be very interesting next few months (years?) to see more applications of ML. Though most of the focus is on software companies (OpenAI, Google. Meta) that release products based on these models, most of innovation and profits are captured by a very few consolidated chip design and manufacturing companies.