10 Buzzing Stocks to Watch as AI Trade Makes a Comeback

2. NVIDIA Corp (NASDAQ:NVDA)

Number of Hedge Fund Investors: 212

Lo Toney from Plexo Capital said in a recent program on CNBC that NVIDIA Corp (NASDAQ:NVDA) is a leader in the industry, but it will start seeing rising competition in the future

“You know, obviously there’s going to be other solutions that are available. We’re starting to see that. I think there’s going to be other options available for folks. But nonetheless, I think NVIDIA Corp (NASDAQ:NVDA) at this point, clear leader and setting the tone for things to come.

However, Toney believes NVIDIA Corp (NASDAQ:NVDA) will keep benefiting from strong demand despite competition:

“It would if it were the case that we believe that the growth was stalling out for AI. In fact, it’s actually increasing. So, I think with that bigger pie, NVIDIA Corp (NASDAQ:NVDA) will still play an important role, albeit even with increased competition.”

Nvidia shares rose after the company said it will soon resume selling H20 chips in China. With this update, the company has removed yet another hurdle in its stock growth. The demand for Nvidia chips is growing worldwide. Saudi Arabia’s Humain plans to buy more than 200,000 AI GPUs from Nvidia, potentially generating $15 billion in sales. The UAE reportedly has an agreement for up to 500,000 GPUs. Even without China’s involvement, Nvidia said nearly 100 AI factories are under construction. These factories have hyperscalers deploying 1,000 GB200 NVL72 racks weekly, each with 72,000 Blackwell GPUs.

Longriver Partners Fund stated the following regarding NVIDIA Corporation (NASDAQ:NVDA) in its second quarter 2025 investor letter:

“This shift in capex priorities has reopened the debate over compute architectures. Custom silicon promises a way out of NVIDIA Corporation’s (NASDAQ:NVDA) pricing power, especially for inference. But the path is not straightforward.

For hyperscalers, the logic of custom silicon is clear. Nvidia’s pricing power is real and inference costs are spiralling. ASICs offer lower cost, better integration, power efficiency, and more control. But while this sounds ideal on paper, it is hard to deliver in the real world.

Custom chips work best when workloads are stable and scale is extreme. AI is neither. Models have evolved quickly, shifting, for example, from transformers to diffusion, and from instruction-tuned to multimodal. Fixed-function chips, by design, are not built to adapt. If the model shifts, their value evaporates. As one expert put it, “If you spend all this money building something and then you find out the workload changes underneath you, you’re basically stuck…” (Click here to read the full text)