AI adoption will “explode” as costs come down

IBM CEO Arvind Krishna, in an interview with Bloomberg at the World Government Summit in Dubai, claimed that, “The usage [of artificial intelligence] will explode as costs come down.”

In other words, lower-cost models such as Deepseek’s R1, released in January, will be the impetus for a new crest of artificial intelligence (AI) experimentation in 2025.

While the investor response to DeepSeek’s release displayed a momentary lapse of confidence in US AI companies, the long-term trend will likely be positive for AI globally.

“I think it is a validation,” Krishna said. “We have been on the point that you do not have to spend so much money to get these models.” Krishna sees the emergence of DeepSeek as a validation of smaller, fit-for-purpose machine learning models that are built for narrow use cases.

Other AI companies aren’t slowing down. OpenAI announced later in February that they’d achieved 400 million weekly active users, up 33% in less than 3 months.¹

Much of this growth came from the enterprise sector. McKinsey reports that over the next 3 years, 92% of companies plan to increase their AI investments.²

“Only 1% of enterprise data has found its way into any form of AI model so far,” Krishna said, implying that the majority of value from AI solutions has yet to be unlocked.

It’s helpful to think of Deepseek’s place in a long chain of innovation that will see further discoveries as generative AI (gen AI) continues to expand within the enterprise context.

Another promising trend that we see in 2025 is the acceleration of AI infrastructure solutions—new chips, new higher-performance networking solutions and new energy-efficient platforms.

Some AI companies are experimenting with making their own custom chips, with the goal of decreasing their dependency on NVIDIA, which controls most of the market.

There is also significant investment from tech giants in new AI-oriented data centers. Microsoft is planning a USD 80 billion investment for data centers, Amazon is expected to invest more than USD 75 billion and Meta is rumored to spend upwards of USD 200 billion.³

To meet the dynamic demands of AI workloads, there's a shift toward disaggregated and composable infrastructure. This approach allows for flexible allocation of computing resources, enabling data centers to optimize performance and efficiency by tailoring hardware configurations to specific AI tasks.

As data centers consume substantial energy, there's an increasing emphasis on sustainability. Companies are exploring energy-efficient hardware solutions and renewable energy sources to power AI operations. Some are aiming to use AI capabilities to reduce AI’s impact on their expanding infrastructure.

It’s not just early-adopting tech companies that are prioritizing AI initiatives. The adoption rate of AI-driven solutions has expanded beyond early adopters and beyond the tech sector. AI technology is addressing specific challenges in every conceivable industry.

In healthcare, AI-driven predictive models are improving diagnostics and treatment planning. Financial services firms are using AI applications for fraud detection, risk management and algorithmic trading.

Supply chain management has seen significant improvements through AI solutions that optimize logistics, demand forecasting and inventory management.

Telecommunications providers are integrating AI chatbots to enhance customer experience, automate service operations and streamline digital transformation efforts.

Within organizations and across various business functions, AI is playing a critical role. In human resources, the use of AI tools assists in talent acquisition, performance evaluation and workforce planning.

AI-driven automation and optimization are reshaping business processes, reducing operational costs and increasing efficiency. AI adoption has also strengthened cybersecurity by using algorithms to detect threats and prevent cyberattacks.

AI adoption does not come without challenges. The IBM Institute for Business Value has identified several common challenges for AI adoption. Organizations must navigate data privacy and responsible AI imperatives.

Promoting ethical AI systems and transparency is critical to maintaining trust. Businesses must also invest in AI skills development to build expertise and support AI-driven initiatives. Allocating budgets for AI strategy and addressing workflow integration concerns remain key factors.

The Institute for Business Value, in partnership with Oxford Economics, surveyed 400 global leaders across 17 industries and 6 geographies in October and November 2024. They found that respondents are “still struggling” to transform business operations with AI, but believe they’re “on the cusp” of a breakthrough.

If 2024 was a year of experimentation, 2025 would be a year where those experiments move toward business as usual. However, the path from experimentation and low-risk, isolated use cases to a grand, enterprise-wide business strategy vision for AI adoption isn’t a straight line.

As barriers to new technologies lower with cheaper models, we can expect a renewed focus on responsible deployments. Leaders know that AI can be done faster and cheaper than ever, but doing it safely and responsibly is something enterprises are pursuing across 2025.