AI spending growth to slow sharply by 2028 as focus shifts to software
Morgan Stanley predicts a major shift in AI investment from chips and hardware to software and applications by 2028.
The massive wave of investment currently flooding the artificial intelligence sector is preparing for a major structural transformation. According to a new report from Morgan Stanley Research, the AI investment cycle is expected to pivot from its current hardware-heavy phase toward software and applications by the year 2028.
This shift comes as the explosive growth in capital expenditure by hyperscalers and cloud providers is projected to decelerate significantly over the next few years. The brokerage anticipates that year-on-year growth in neocloud and hyperscaler spending will drop from a staggering 93 per cent in 2026 to 61 per cent in 2027, eventually hitting a much lower 14 per cent by 2028.
From Chips to Applications
The report suggests that this deceleration does not signal a decline in interest, but rather a change in where the money is being directed. We are moving away from the era of massive spending on chips and data-centre infrastructure and entering a phase defined by AI software and services.
“This indicates we could soon see a transition from hardware to software/services in 2028, and we expect this shift heading into next year as this is priced in,” the Morgan Stanley report noted.
While the focus is moving, the transition is expected to be complex. The brokerage expects the next 6 to 12 months to serve as a bridge, where opportunities begin to broaden beyond semiconductor giants to include early software providers and businesses leveraging AI to boost their bottom line.
The Power Bottleneck: A Major Hurdle
Despite the shift toward software, the infrastructure buildout is far from over. The demand for computing power is expected to remain robust, primarily because the industry is hitting significant physical and regulatory walls. Among these, power availability stands out as the most critical constraint.
Morgan Stanley’s analysis highlights several alarming bottlenecks:
- Energy Shortage: There is an estimated 57 gigawatt power shortfall projected between 2026 and 2028, representing a deficit of roughly 30-40 per cent.
- Regulatory and Local Hurdles: Local opposition has already impacted the industry, with USD 286 billion worth of US data-centre projects being delayed or cancelled since 2025.
- Resource Scarcity: Shortages in labour and regulatory capacity continue to hamper rapid data-centre expansion.
Because of these constraints, the demand for compute is likely to exceed supply for many years to come.
Measurable Gains in Productivity
The shift toward software is being driven by real-world results. Companies that have already adopted AI are seeing tangible improvements in their operations. On average, businesses across various sectors have reported a 9.6 per cent increase in net productivity over the last year.
Looking ahead, Morgan Stanley predicts that AI adoption could contribute to a net margin expansion of approximately 100 basis points through 2027. The report expects software leadership to emerge first in the realms of infrastructure and cybersecurity before eventually trickling down into broader consumer and business applications.
Ultimately, the AI cycle is not a simple hand-off from one sector to another. Instead, it is a broadening ecosystem where infrastructure, power constraints, and software applications will continue to evolve alongside one another.