New Survey Highlights Challenges in AI ROI for Companies

Artificial intelligence (AI) has demonstrated impressive capabilities, but it is crucial to acknowledge its limitations. Essentially, AI models operate by predicting the next token—whether in text or images—without genuine reasoning or comprehension, relying heavily on statistics.

These constraints lead to inaccuracies and the generation of false information, making it difficult for businesses to realize the full potential of their AI investments. International Business Machines (NYSE: IBM) recently surveyed 2,000 CEOs worldwide, revealing concerning insights for companies like Nvidia (NASDAQ: NVDA) that are heavily invested in AI.

CEO Insights on AI Investments

IBM’s survey revealed that only 25% of CEOs believe their AI initiatives have achieved the expected return on investment (ROI) in recent years. Alarmingly, just 16% have successfully scaled these initiatives across their organizations.

Despite these challenges, AI continues to provide value, albeit not always to the extent anticipated. Approximately 52% of CEOs reported that their organizations are finding value in AI investments beyond just cost savings. Interestingly, about two-thirds of CEOs attribute their AI investments to a fear of falling behind competitors, rather than a clear understanding of financial benefits.

Nevertheless, optimism remains, with 85% of CEOs predicting a positive AI ROI by 2027.

Implications for Nvidia

Nvidia’s growth relies heavily on the demand for advanced AI accelerators used to train and deploy increasingly complex AI models. However, as businesses struggle to derive positive ROI from their AI efforts, reducing AI implementation costs is likely to become a critical factor for success.

This trend could pose challenges for Nvidia. The company needs to sell millions of its data center graphics processing units (GPUs) at high prices to sustain growth. Although increasing AI usage can potentially elevate GPU demand, businesses’ difficulties in successfully implementing AI investments suggest that future AI development might focus on more efficient models that can operate on less expensive hardware.

Earlier this year, DeepSeek disrupted the AI market with affordable AI models. Additionally, Microsoft introduced a “1-bit” AI model capable of running on a central processing unit (CPU), consuming just 0.4 gigabytes of memory while delivering performance similar to larger models. Similarly, IBM recently showcased its Granite 4.0 Tiny AI model that can function on consumer-grade GPUs.

Affordable AI technologies are beneficial for the industry as they enhance accessibility and open up use cases that might otherwise remain financially unviable. Companies focused on enterprise AI, like IBM, stand to gain, while Nvidia may face obstacles if AI models continue to trend toward requiring less power.

Investors in Nvidia should remain cautious about the low success rates companies have experienced in scaling their AI initiatives and achieving financial viability.

The views and opinions expressed herein are the views and opinions of the author and do not necessarily reflect those of Nasdaq, Inc.