Will Broadcom Chips End AMD Stock's AI Dreams?

AMD’s (NASDAQ:AMD) shares experienced a decline of more than 6% during Friday’s trading session following the announcement of impressive quarterly results by semiconductor and infrastructure software provider Broadcom, which revealed a $10 billion order for custom AI chips from a new customer, widely believed to be OpenAI. This development caused Broadcom’s stock to rise nearly 11%, raising worries about AMD’s position in the increasingly competitive AI hardware sector. Nvidia (NASDAQ:NVDA) stock also dropped by approximately 3% in response to the news, although its well-established dominance in GPUs, along with its extensive software ecosystem, is providing investors with less cause for concern. The key takeaway is that the AI hardware market may be entering a new phase – one in which custom silicon begins to redefine infrastructure spending. GPUs have fueled the surge in generative AI training thus far, but ASICs are starting to emerge as more efficient options for inference workloads.

AMD vs Nvidia vs Broadcom

While AMD is still significantly trailing Nvidia in the AI accelerator market, it has achieved some advancement in recent months. Data center revenue, often seen as a proxy for AI chip sales, increased by 57% in Q1 to $3.7 billion, partially due to strong demand for Instinct GPUs. However, the subsequent quarter presented a different picture, with revenue of $3.2 billion falling short of consensus expectations, and guidance indicating an 11% sequential drop and roughly 14% year-over-year growth.

In contrast, Broadcom reported a 63% year-over-year increase in Q3 AI revenue, reaching $5.2 billion, while Nvidia’s data center sales skyrocketed by 56% last quarter to $41.1 billion, based on a much larger foundation. This has reinforced the perception that AMD is still very much in the early stages of its AI journey, with Nvidia controlling a dominant share of training workloads, and Broadcom now threatening to gain market share with its custom silicon. Nvidia Stock To Fall 50% As AI Cycle Turns?

For investors seeking growth opportunities while reducing exposure to the high volatility of individual stocks, we encourage you to consider the High Quality Portfolio. This portfolio has demonstrated an impressive performance history, consistently surpassing its benchmark—a mix of the S&P 500, Russell, and S&P MidCap indices—and achieving cumulative returns exceeding 91%, as shown in HQ performance metrics.

GPU To Custom Chips

The AI landscape may be approaching a pivotal moment. Over the past three years, there has been significant capital invested in GPUs for training extensive models, but this phase appears to be becoming more saturated. The incremental performance enhancements from larger models are tapering off, and the availability of high-quality training data is becoming a limiting factor – much of the easily obtainable data on the Internet has already been utilized. This indicates that the focus of AI expenditure may shift from training, which occurs upfront, to inference, where models are implemented at scale and consistently used in practical applications.

This is specifically where ASICs have an advantage. The parallels with cryptocurrency are noteworthy: Bitcoin mining initially started with GPUs, but efficiency and cost requirements quickly led to a complete shift to ASICs. A similar rationale applies to AI inference, where custom chips can outperform GPUs in terms of power consumption and cost-efficiency. Broadcom seems well-positioned for this transition, with CEO Hock Tan indicating that hyperscale customers are increasingly adopting XPUs – the company’s custom chip architecture – for a growing portion of their computing needs.

If the client behind Broadcom’s $10 billion order is indeed OpenAI, it would represent the clearest indication yet of a shift away from Nvidia GPUs for ChatGPT workloads. Such a change would likely have industry-wide implications, prompting other hyperscalers to reduce their dependency on GPUs and be more willing to experiment with custom silicon. OpenAI, which has traditionally operated its workloads on Microsoft Azure utilizing Nvidia or AMD GPUs, may now be transitioning towards in-house equipment utilizing custom Broadcom silicon.

AMD Challenges

For AMD, this presents challenging questions. Its dilemma is now dual-faceted: defending against Nvidia, whose GPU ecosystem and CUDA software are deeply rooted, while also addressing the rise of ASICs that could diminish demand for general-purpose accelerators. The company’s MI300 series shows potential, but still lacks substantial customer validation at scale. Meanwhile, Broadcom’s strengths in other areas of the AI stack, including networking through its Tomahawk switches and interconnects, bolster its position, providing hyperscalers with a bundled offering. Currently, AMD faces the risk of being pushed into a middle position, sandwiched between Nvidia at the high end and ASIC suppliers at the efficiency frontier.

Even after the recent sell-off, AMD stock is trading at approximately 40 times the estimated 2025 earnings. Revenues are expected to increase by roughly 28% this year, with recovery in its CPU business being a contributing factor. Comparatively, Broadcom commands an even higher valuation at roughly 49 times forward earnings, despite forecasts suggesting a slower revenue growth of about 23% for this year. Nevertheless, the premium multiple appears justified given Broadcom’s accelerating momentum in AI. Its $10 billion custom silicon deal, combined with a 63% year-over-year increase in AI revenues last quarter and strength in networking interconnects, could present a more sustainable growth story for investors.

The Trefis High Quality (HQ) Portfolio, consisting of 30 stocks, has a proven history of consistently outperforming its benchmark, including the S&P 500, Russell, and S&P midcap. What drives this? Collectively, HQ Portfolio stocks have generated superior returns with decreased risk when compared to the benchmark index; providing a smoother experience, as observed in HQ Portfolio performance metrics.