Despite the launch of the new product claiming to surpass Nvidia's similar artificial intelligence (AI) chip in reasoning performance, AMD's stock price still ended sharply lower. Investors seem unimpressed by verbal presentations and are actually looking for signs of a return on AI investments.

On Thursday, October 10th, Eastern Time, during AMD's datacenter event introducing the MI325X AI accelerator and multiple network chips, AMD's stock price briefly turned higher in the early trading session and midday amid the morning session in the US stock market, but then continued to decline in the afternoon, hitting a new daily low of $162 before market close. The intraday decline widened to around 5.3%, ultimately closing down 4%, marking the lowest close in a week, with the largest intraday and closing decline since September 3rd.

One would expect the launch of a powerful AI chip to be bullish, especially since AMD's new product competes with Nvidia's widely-watched Blackwell architecture chips. Why then is AMD's stock price accelerating its decline?

Some media summaries indicate that the stock price decline shows investors are still waiting for returns from AMD's AI business. The commentary suggests that despite AMD's chip releases, the company's near-term financial outlook has not changed significantly. Like Nvidia, AMD also commits to introducing new AI accelerators every year to hasten innovation. However, AMD still has a long way to catch up with Nvidia, and Wall Street is eagerly anticipating signs of AMD closing in on Nvidia's progress. This progress might only be visible after AMD's next financial report, which is expected to be released around the end of this month following the announcement of the third-quarter earnings.

The second-quarter financial report released at the end of July this year showed that AMD's datacenter business's net revenue was $2.8 billion, a 115% year-on-year increase. This was mainly due to the strong demand for the AI accelerator MI300. As the flagship product competing with Nvidia's H100, the sales of MI300 in a single quarter exceeded $1 billion, far surpassing market expectations. At that time, AMD CEO Lisa Su stated that AMD's AI chip sales were 'better than expected,' with plans to launch MI325X in the fourth quarter of this year, a 'highly competitive' MI350 compared to Nvidia's Blackwell next year, and MI400 in 2026.

After AMD's financial report was released, some analysts believe that NVIDIA still has a leading advantage over AMD. Compared to NVIDIA, AMD's datacenter business scale is still very small - a quarterly sales revenue of only 2.8 billion US dollars. As for NVIDIA, the datacenter revenue for the first quarter of the 2025 fiscal year ending April reached $22.6 billion, setting a record high for the business for a single quarter, with a 23% increase over the previous quarter and a 427% increase year-on-year.

During this week's datacenter event on Thursday, AMD introduced that the company's latest AI datacenter GPU MI325X is expected to be shipped in the fourth quarter of this year and in the first quarter of next year, products carrying this GPU will be launched to the market through ADM partners such as Dell, Eviden, Gigabyte, Huizhi, Lenovo, and Super Micro Computer.

Some media outlets stated that the event on Thursday showed that AMD's long-term market potential for AI chips is becoming increasingly optimistic. Su Zifeng said that in the past year, the demand for AI has exceeded AMD's expectations. She now predicts that by 2028, the market size of AI datacenter GPUs will reach $500 billion, with an annual growth rate of over 60%. Last December, she predicted that the market would exceed $400 billion by 2027.

As for whether AMD's MI325X will surpass NVIDIA's products, the media cannot assert and implies that it is still unknown. It points out that the most comparable product to MI325X is not NVIDIA's H200, but NVIDIA's most powerful Blackwell architecture GPU, which has already been launched to customers this quarter. The performance of the Blackwell GPU may be superior to the MI325X, as the MI325X is based on the same CDNA 3 architecture as AMD's previous MI300X GPU.

Earlier, Lynx Equity Strategies commented that the significant drop in AMD's stock price on Thursday may be due to the inability of the MI325X to significantly change AMD's lagging position and effectively challenge NVIDIA's dominant status. "Unless AMD can demonstrate a clear market share growth driver, the stock may retest its annual low."

Lynx Equity Strategies also pointed out,

AMD's management may promote that the benchmark test results of its chips are better than NVIDIA's Hopper series, but in real-world workloads, we believe that the MI300X still lags behind NVIDIA's H100. Besides excellent HBM memory density, software stack, backplane network, and power consumption are equally important.

However, KeyBanc analyst John Vinh believes that even though AMD is far behind NVIDIA, the company's development trajectory is clearly optimistic, with expectations to ship 0.5 million pieces of Instinct MI300X AI accelerators this year, increasingly gaining a foothold in the AI market.

In addition, some media reports that the AI hardware sector is attracting enterprises to invest large budgets, and AMD hopes to get a share of it. Nvidia is currently the leader in the entire market, with some analysts believing that given the huge opportunities in the AI hardware sector, many companies have the potential to succeed.