AMD, the chipmaker based in Silicon Valley, is making strides in closing the performance gap with Nvidia‘s dominant artificial intelligence processors. The company recently unveiled new products aimed at capturing a market worth hundreds of billions of dollars. One of these products, the MI325X chip, is set to be released to customers in Q4 2022 and boasts “industry-leading” performance compared to Nvidia’s H200 AI chips. Additionally, AMD’s next-generation MI350 chip is on track for shipment in H2 2025 and aims to compete with Nvidia’s Blackwell system.
Under the leadership of CEO Lisa Su, AMD has made a remarkable comeback from near bankruptcy a decade ago and has emerged as a strong contender challenging Nvidia’s dominance in generative AI infrastructure. Su envisions AMD becoming the “end-to-end AI leader” within the next decade and emphasizes that this is just the beginning of the AI race.
Meanwhile, Nvidia’s customers are expected to start deploying their Blackwell system this quarter. Microsoft recently announced that it became the first cloud provider to offer its customers access to Nvidia’s latest GB200 chips. While major players like Microsoft, Google, and Amazon are developing their own in-house AI chips (“hyperscalers”), AMD has positioned itself as Nvidia’s closest competitor when it comes to off-the-shelf AI chips.
However, despite its progress, AMD still lags behind Nvidia significantly. In Q3 2021 alone, while AMD projected $4.5 billion in AI chip sales for 2024; Nvidia reported $26.3 billion in sales for its data center chips during that same period.
Nonetheless, Su remains optimistic about future demand for AI chips as she predicts that by 2027, there will be a total addressable market worth $400 billion for such technology due to tremendous demand for AI infrastructure.
In addition to unveiling new chips targeting this growing market segment on Thursday; AMD also announced networking technology advancements and upgrades to its ROCm software toolkit – all aimed at providing scalable and efficient solutions for building cutting-edge AI systems within data centers.
Su believes that offering an end-to-end infrastructure solution is crucial since large clusters of chips are required for training extensive language models – something many organizations desire.
Having joined AMD as senior vice-president back in 2012 after working at prominent companies like Texas Instruments and IBM; Su took over as CEO in 2014 when shares were valued around $4 amid concerns about bankruptcy due to intense competition with Intel.
Today though; thanks largely due diversification efforts away from traditional PC business towards server chip market domination coupled with advancements made on artificial intelligence frontiers -AMD now boasts triple Intel’s market capitalization ($275bn vs.$92bn) closing at $171 per share ahead announcement day (Wednesday).
Su sees artificial intelligence as being instrumental driving force behind future growth prospects while aiming secure same clientele base enjoyed by rival firm NVIDIA.
Microsoft along Meta have already adopted current generation MI300 GPUs manufactured by Advanced Micro Devices (AMD). Furthermore Amazon who already purchases server CPUs from them may follow suit according Lisa: “It’s point time conversation.”
To catch up with NVIDIA’s offerings such Blackwell system which sells complete server racks comprising multiple integrated circuits alongside proprietary networking equipment-AMD has pursued aggressive investment strategies including recent acquisition ZT Systems ($4.9bn) known building servers catering small group hyperscalers.
Regarding regulatory reviews pertaining aforementioned deal-Su expects US-EU checks along few other jurisdictions but not China given current thresholds aren’t met yet
