Key Points
-
Nvidia’s dominance in the market for AI processing units is well-established.
-
The rise of application-specific integrated circuits (ASICs) in the cryptocurrency sphere serves as a cautionary tale for Nvidia investors, highlighting potential market shifts.
-
Despite the parallel with cryptocurrency mining, significant differences exist that could influence the outcome for AI hardware.
Currently, Nvidia(NASDAQ: NVDA) is reaping considerable benefits from the surge in artificial intelligence (AI) technologies. However, those investing should be aware of the rapid changes markets can undergo and how quickly hardware leadership can evaporate. The shift within the cryptocurrency mining industry offers a pertinent example. Graphics processing units (GPUs) were once fundamental to crypto mining, before the advent of application-specific integrated circuits (ASICs) provided a more advanced approach.
These ASICs are designed for a single, specialized task, performing it with enhanced speed and cost-effectiveness. Consequently, the adoption of GPU-based mining for digital currencies like Bitcoin diminished rapidly. The economic advantages presented by ASICs were too significant to ignore. To remain competitive, cryptocurrency miners had to transition to ASICs or face economic disadvantage.
Now, custom-designed processors are emerging for AI-specific functions.
Image source: Getty Images.
ASICs and AI
GPUs have historically been favored for training extensive large language models (LLMs), and Nvidia has secured a dominant position in the GPU market, largely due to its robust software platform, CUDA. An entire ecosystem has been created around the company’s processors, propelling its data center revenue. However, the high costs and energy demands of AI workloads are driving major hyperscale computing firms to explore more cost-effective and energy-efficient solutions.
This impetus led Alphabet to develop Tensor Processing Units (TPUs), while Amazon engineered its Trainium and Inferentia chips. Other major players are also pursuing custom silicon. Meta Platforms and OpenAI are reportedly collaborating with Broadcom to create bespoke chips, with OpenAI potentially behind a substantial $10 billion order for the coming year. Furthermore, Microsoft, a key Nvidia customer, is also developing its own tailored AI processor.
The primary goals are reducing costs and diminishing dependency on Nvidia. Concurrently, the industry is increasingly focused on inference, leading to a changing technological landscape. Nvidia’s advantage in inference is not as pronounced as in training. Since inference is less technically demanding than training, the legacy code built on CUDA holds less weight. Moreover, because inference represents an ongoing expense, total cost of ownership and per-inference cost become more critical considerations.
In the cryptocurrency realm, the shift to ASICs caused GPUs to rapidly transition from essential to irrelevant as cost structures shifted. Nvidia’s current market value relies on the assumption that hyperscalers will continue acquiring increasing quantities of GPUs; however, past experience suggests this will only persist as long as it remains economically sensible.
However, notable distinctions exist between Bitcoin mining and AI inference which may favor Nvidia. Bitcoin mining involves repetitive, brute-force calculations, while inference requires understanding intent and utilizing trained LLM information for execution. Furthermore, the continuous development of new AI techniques, such as reasoning or multimodal AI, means that GPUs, with their greater adaptability, may be better suited than ASICs, which can become rapidly outdated.
Recognizing this potential threat, Nvidia is taking proactive measures to protect its market share. The recent $100 billion investment partnership with OpenAI exemplifies this strategy. OpenAI represents a major consumer of Nvidia GPUs, directly or indirectly, but it is also developing its own AI ASICs. This investment serves to incentivize OpenAI’s continued use of Nvidia GPUs.
Will AI ASICs replace GPUs?
Investors should closely monitor the emergence of ASICs, as it may present the most significant challenge to Nvidia’s growth trajectory. Despite the company’s strong competitive advantages, these are not insurmountable. Hyperscalers possess the financial resources and motivation to erode Nvidia’s dominance, and every dollar allocated to internal AI chip development represents a dollar not spent on Nvidia products.
This does not imply the complete obsolescence of GPUs. They are adaptable to evolving AI workloads, accommodating new models and methodologies. Nevertheless, the transition towards inference is likely to see custom AI chips gain market share.
Currently, the market appears large enough to accommodate multiple AI infrastructure providers in the future. However, the rapid shift observed in Bitcoin mining serves as a reminder of how quickly economic conditions can change, potentially impacting the AI industry similarly. Investors should bear this in mind, avoiding the assumption of Nvidia’s uninterrupted growth over the coming decade.
Should you invest in Nvidia right now?
Before investing in Nvidia, consider the following:
The Motley Fool Stock Advisor team has identified what it believes are the 10 top stocks available, and Nvidia isn’t one of them. These alternative selections could provide substantial returns in the coming years.
When Netflix was recommended on December 17, 2004, a $1,000 investment would have grown to $651,593!* Similarly, when Nvidia was recommended on April 15, 2005, a $1,000 investment would have grown to $1,089,215!*
The average Stock Advisor return stands at 1,058%, significantly exceeding the S&P 500’s 188% gain. Don’t miss the latest list of top 10 stocks; access it by joining Stock Advisor.
*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 recommends 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.