There’s no question that artificial intelligence (AI) has real use cases and that the latest AI models can do some incredible things. But it’s important to remember that the technology has limitations. Ultimately, all an AI model does is predict the next token — characters of text, pixels in an image, and so forth — in a stream of tokens. There’s no real reasoning or thinking, just a bunch of statistics.
Because of these limitations, AI models often get things wrong or create false information. It’s not surprising, then, that businesses are struggling to make AI investments work. International Business Machines (IBM -1.32%) recently conducted a survey of 2,000 CEOs around the world, and the main takeaway paints a rough picture for companies like Nvidia (NVDA -0.62%) that are all-in on AI.
Image source: Getty Images.
A lot of hope is driving AI investments
According to IBM’s survey, just 25% of CEOs said AI initiatives have delivered the expected return on investment (ROI) over the past few years. Worse, only 16% of CEOs report that AI initiatives have been scaled enterprise-wide.
AI is still providing value to businesses, although perhaps not as much as originally expected. About 52% of CEOs claim their organization is realizing value from AI investments outside of cost reductions. Around two-thirds of CEOs noted that the risk of falling behind, rather than a clear view of the financial benefits, is driving AI investments right now.
While AI technology is struggling to deliver for businesses, CEOs remain optimistic. A full 85% of CEOs expected AI to deliver a positive ROI by 2027.
Why this is bad news for Nvidia
Nvidia’s growth story hinges on ever-growing demand for more powerful AI accelerators to train and run AI models of increasing complexity and cost. However, with businesses struggling to produce positive returns on investment from AI initiatives, the path to positive ROI almost certainly involves bringing down the cost of AI.
That’s not a trend that will benefit Nvidia. Nvidia needs to sell many millions of its data center graphics processing units (GPUs), and it needs to sell them for tens of thousands of dollars each to keep growing. While expanding usage of AI can drive demand for GPUs, the track record of businesses making AI investments work suggests that the future of the AI industry involves more efficient AI models capable of running on cheaper hardware.
DeepSeek shook the AI industry earlier this year with its inexpensive AI models, and other developments are pointing in the same direction. Microsoft recently unveiled a new “1-bit” AI model that can run on a central processing unit (CPU) and uses just 0.4 gigabytes of memory while matching the performance of other AI models in the same size class. In the same vein, earlier this month, IBM previewed its Granite 4.0 Tiny AI model that can run on inexpensive consumer-grade GPUs.
Cheap AI is good for the AI industry in the sense that it makes the technology more accessible and unlocks use cases that wouldn’t work financially if the cost of running an AI model were prohibitive. Companies like IBM that are focused on enterprise AI solutions with clear ROIs benefit from the proliferation of affordable AI, while companies like Nvidia that need AI models to become more power hungry likely won’t.
While the AI industry is evolving quickly, Nvidia investors should be concerned about businesses’ low rate of success in scaling their AI investments and making them work financially.
Timothy Green has positions in International Business Machines. The Motley Fool has positions in and recommends International Business Machines, Microsoft, and Nvidia. The Motley Fool 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.
