Is AMD The Dark Horse Of The AI Silicon Race?

Following the launch of OpenAI’s surprisingly competent generative AI tool, ChatGPT, major tech players appear to be doubling down on the artificial intelligence space. This has turned the spotlight on companies that make graphics processing units, the de-facto chips for running AI-related workloads. Nvidia – the dominant player in the GPU space – has been the biggest beneficiary of the trend thus far. Nvidia stock is up over 3x this year and the company just published a blowout set of Q2 results, with revenue growing 2x and profits rising over 5x compared to last year as sales to data center customers surged.

We note that AMD stock has had a Sharpe Ratio of 0.8 since early 2017, compared to close to 0.6 for the S&P 500 Index over the same period. This compares with the Sharpe of 1.3 for the Trefis Reinforced Value portfolio. Sharpe is a measure of return per unit of risk, and high-performance portfolios can provide the best of both worlds.

Advanced Micro Devices stock (NASDAQ
NDAQ
: AMD), too, is poised to emerge as a key player in this AI silicon space. Although AMD’s GPUs have typically been used more for gaming and professional applications compared to Nvdia’s chips, which are the go-to GPUs for accelerated computing applications, AMD is also now focusing on winning over the AI market. The company recently unveiled the MI300X chip which is targeted specifically at large language model training and inference for generative AI workloads. The company expects to begin offering the technology for customers to try sometime in Q3 2023, with production slated to ramp up from 2024. The chip’s specs indicate that it could compete favorably with Nvidia. The chip will house 192 gigabytes of high-bandwidth memory, well ahead of the 120 GB found on Nvidia’s H100 data center GPU. This should imply that AMD’s new processors can train more sizable large-language models for generative AI applications.

AMD appears to be prepping to scale up production for the chips meaningfully. DigiTimes has reported that foundry major TSMC – which produces most of the GPUs on sale today and offers an advanced chip packaging technology called Chip on Wafer on Substrate (CoWoS) – indicated that AMD’s new chips using the CoWoS technology could be as much as half of Nvidia’s CoWoS-equipped GPU output. This is notable, considering that Nvidia currently holds over 90% of the compute-focused GPU market.

So, is AMD stock still a buy at the current market price of about $103 per share? AMD stock has gained almost 60% this year and its valuation isn’t exactly cheap, considering that the stock trades at about 28x consensus 2023 earnings, despite revenue expected to contract for the year. Moreover, Nvidia has been looking to cement its dominant position in the AI chip space, developing an ecosystem of sorts around its AI chips, with its own programming languages, and software, which the company is hoping will allow it to better lock in customers. However, the AI processor space is still in its infancy and it is unlikely to be a “winner takes all” type of market and AMD still has a good shot at scaling up if its new chips impress. We value AMD stock at about $107 per share, which is marginally ahead of the current market price. See our analysis on AMD Valuation: Is AMD Stock Expensive Or Cheap? for more details on what’s driving our price estimate for AMD. Also, check out our analysis of AMD Revenue for more details on the company’s key revenue streams.

Invest with Trefis Market Beating Portfolios

See all Trefis Price Estimates