Global investment in artificial intelligence is entering a new phase. After the first wave of euphoria, the focus is increasingly shifting from visionary applications to the technical basis that makes the operation of these systems possible in the first place: high-performance data centers and IT infrastructure.

Current analyses by Goldman Sachs show a clear trend towards a more selective investment strategy. Investors are increasingly focusing on companies that have physical infrastructure – especially operators of large data centers and providers of high-performance hardware. On the other hand, many providers of specialized AI software or experimental tools lose attention if they lack the necessary scalability.

The major cloud providers are pushing ahead with the expansion of their capacities at an enormous pace. Companies such as Amazon Web Services, Microsoft and Google invest tens of billions of euros annually in new data centers, specialized chips and network structures.

The background is clear: AI systems – especially large language models – require massive computing power. When training such models, thousands of high-performance processors run in parallel over long periods of time. Even during ongoing operation ("inference"), computing capacity is continuously required to provide answers, predictions or analyses in real time.

According to current forecasts, AI could take up around 30% of the total data center capacity in the next two years – a massive structural change compared to classic cloud applications.

An often underestimated aspect: The rapid expansion of the AI infrastructure is leading to an enormous increase in energy demand. Studies predict that global data center power consumption could increase by up to 175% by 2030 – largely driven by AI applications.

This makes energy a strategic bottleneck. Data centers are increasingly being built where stable and cheap power supply is available – often in regions with access to renewable energies or high-performance power grids.

At the same time, environmental aspects are also coming more into focus. Cooling systems, water consumption and space requirements are increasingly viewed critically, especially in large-scale AI training centers.

Location, infrastructure and time as limiting factors

The construction of modern data centers is complex and time-consuming. In addition to hardware, land availability, grid connections, approval procedures and long-term energy contracts play a central role.

Many projects take several years to commission. At the same time, bottlenecks in electrical infrastructure, transformers or grid capacities are leading to delays. These factors are slowing down the expansion and raising the barriers to entry for new market participants.

This is exactly where it becomes clear why investors are increasingly relying on established infrastructure operators: Those who already have functioning data center networks have a clear competitive advantage.

"Flight to quality": The market is becoming more selective

While in the first phase of the AI euphoria, numerous companies gained stock market value solely through the reference to artificial intelligence, a reassessment is now beginning.

Investors are taking a closer look at which business models are sustainable in the long term. At the heart of this are:

These companies form the foundation of the entire AI economy – regardless of which applications ultimately prevail.

Historically, a familiar pattern emerges: In phases of technological upheaval, infrastructure providers often benefit first from stable and predictable revenues, while software solutions fluctuate more strongly.

Conclusion: The next phase of AI will be physical

The development of AI is no longer just a question of algorithms and software. Success increasingly depends on very classic factors: electricity, cooling, space and hardware.

The next phase of the AI race will therefore not only be decided in software labs, but also in power plants, data centers and global supply chains.

For companies, this means that without scalable infrastructure, any AI application, no matter how innovative, remains theoretical.

For investors, the market is becoming more demanding – and rewards substance instead of hype.









Author: MF-Redaktion / Tom Weyermann
Source: AI News