Key Takeaways
-
Nvidia currently dominates the market for AI processing units.
-
The rise of ASICs in cryptocurrency mining serves as a cautionary tale for investors regarding market shifts that can impact Nvidia’s position.
-
Significant differences exist between the cryptocurrency mining scenario and the current AI landscape.
Nvidia(NASDAQ: NVDA) has reaped considerable benefits from the surge in artificial intelligence (AI) adoption. However, stakeholders should be mindful of the potential for rapid alterations in hardware leadership as the AI market matures. The trajectory of cryptocurrency mining provides a relevant parallel. Initially, graphics processing units (GPUs) were essential for crypto mining, but application-specific integrated circuits (ASICs) soon emerged to optimize the process.
These ASICs performed specialized functions but delivered enhanced efficiency and cost-effectiveness. Consequently, GPU-based mining, particularly for currencies like Bitcoin, diminished significantly as ASICs became the superior choice. Miners who desired to remain competitive were compelled to transition to ASICs.
Looking for promising investment opportunities? Our expert analysts have identified their top 10 stock picks for immediate investment. Discover Now »
Currently, ASICs are being engineered to tackle specific AI tasks.
Image source: Getty Images.
AI ASICs: A Potential Shift
GPUs have been consistently used for training large language models (LLMs). Nvidia’s robust software environment, CUDA, has solidified the company’s lead in the GPU sector. The entire ecosystem built around Nvidia’s processing units is the reason for the dramatic increase in data center revenue. However, AI applications are extremely expensive and power-intensive, making it advantageous for large hyperscale entities (companies that operate extensive data centers) focused on AI to find more cost-effective and efficient alternatives.
This motivation led Alphabet to create Tensor Processing Units (TPUs) and prompted Amazon to develop its Trainium and Inferentia chips. Other organizations are now adopting similar strategies. Reportedly, Meta Platforms and OpenAI are collaborating with Broadcom on custom chip development. It is suggested that OpenAI placed a significant $10 billion order for the next year. Furthermore, Microsoft, a major Nvidia client, has also invested in creating its own tailored AI chip.
The underlying objective is to reduce expenses and reliance on Nvidia’s products. With the market increasingly focusing on inference, the competitive dynamics are evolving. Nvidia’s competitive advantage in inference is less substantial than in training. Inference requires less technical sophistication than training, diminishing the impact of the established CUDA-based code. Since inference represents an ongoing expense, overall ownership costs and cost per inference become more crucial.
When cost considerations drove the transition to ASICs in Bitcoin mining, GPUs quickly became obsolete. Nvidia’s present market valuation assumes continuing GPU demand from hyperscalers. However, historically, they will continue purchasing only if the economics remain favorable.
Several key distinctions exist between Bitcoin mining and inference that could benefit Nvidia. Bitcoin mining involves a repetitive process, while inference requires understanding the meaning behind inputs (such as questions) and leveraging LLM training data to execute instructions. The continuous creation of novel AI approaches such as reasoning and multimodal AI favors GPUs because they are more easily adaptable than ASICs, which are more likely to become outdated quickly.
Nvidia recognizes this potential risk and is implementing safeguards. The recent $100 billion investment partnership with OpenAI exemplifies this. OpenAI relies heavily on Nvidia’s GPUs; however, they have been developing their own AI ASICs. By investing, Nvidia is effectively incentivizing OpenAI to continue using its GPUs.
The Future: ASICs vs. GPUs in AI
The threat posed by ASICs warrants close monitoring as it represents a considerable risk to Nvidia’s growth. While the company has a significant advantage, it isn’t impenetrable. Hyperscalers have both the financial resources and the motivation to erode Nvidia’s dominance. Funds allocated to internal AI chip development reduce Nvidia’s potential revenue.
GPUs are unlikely to disappear completely, given the evolving nature of AI workloads. GPUs can adapt to handle new models and techniques. Still, custom AI chips will likely gain market share as the industry increasingly emphasizes inference.
Currently, the market appears large enough to support multiple AI infrastructure leaders. However, the Bitcoin experience demonstrates how quickly market conditions can shift. AI may follow a similar trajectory. Investors should consider this possibility before assuming Nvidia’s sustained growth over the coming years.
Nvidia: A Smart Investment Today?
Before investing in Nvidia, consider the following:
The analysts at Motley Fool Stock Advisor have identified their top 10 stocks to buy now – and Nvidia didn’t make the list. These 10 stocks have the potential to produce significant gains in the years ahead.
Consider Netflix‘s inclusion on this list on December 17, 2004. A $1,000 investment at that time would be worth $651,593!* Similarly, a $1,000 investment in Nvidia when it was listed on April 15, 2005, would now be worth $1,089,215!*
It’s also important to note that Stock Advisor’s average return is 1,058% – a substantial outperformance compared to the S&P 500’s 188%. Don’t miss out on the latest top 10 list; join Stock Advisor today.
*Stock Advisor returns as of September 22, 2025
Geoffrey Seiler holds positions in Alphabet. The Motley Fool holds positions in and recommends Alphabet, Amazon, Bitcoin, Meta Platforms, Microsoft, and Nvidia. The Motley Fool recommends Broadcom and suggests the following options: long January 2026 $395 calls on Microsoft and short January 2026 $405 calls on Microsoft. The Motley Fool has a disclosure policy.