Goldman Sachs: Will AI Data Centres Drive 165% Power Surge?

The rapid rise of AI is reshaping global energy demand, with data centres at the heart of this transformation. 

According to Goldman Sachs, electricity consumption from AI-driven facilities will climb 165% by 2030 compared with 2023 levels.

This shift is already evident in soaring construction spend and occupancy rates, with the biggest impact yet to come.

AI reshaping data centre demand

Goldman Sachs reports that US data centre construction spend has tripled in three years as hyperscalers accelerate build-outs to meet AI requirements. 

Despite this, third-party facilities remain close to full capacity across most major markets.

“Over the next five to six years, we forecast substantial demand growth in the global data centre market,” says James Schneider, Senior Equity Research Analyst at Goldman Sachs covering digital infrastructure.

Current global consumption stands at around 55GW, more than half of which powers cloud computing. 

Traditional workloads such as email and storage account for about a third, while AI represents just 14%. 

By 2027, total demand is expected to reach 84GW, with AI’s share rising sharply.

Rising power density

The growth is not only about volume but also intensity.

Goldman Sachs forecasts that average power density will increase from 162kW per square foot today to 176kW by 2027.

“Data centre supply – specifically the rate at which incremental supply is built – has been constrained over the past 18 months,” James adds, pointing to rising density and power requirements as key challenges.

One example illustrates the scale: a single ChatGPT query uses around 2.9 watt-hours of electricity, nearly 10 times that of a Google search, according to International Energy Agency (IEA) data. 

Multiplied across millions of interactions, the load quickly escalates.

Goldman Sachs models three scenarios for growth. Its base case projects demand rising 50% to 92GW by 2027, with slower adoption leading to 14% growth or faster uptake pushing annual growth to 20%.

European and global expansion

Europe faces a pipeline of about 170GW of data centre projects, equivalent to a third of current electricity use across the continent.

“Inflecting power demand is monumentally important, because it’s been declining for 15 years in Europe,” says Alberto Gandolfi, Managing Director, Equity Research at Goldman Sachs. 

By 2030, European data centres will require as much electricity as Portugal, Greece and the Netherlands combined.

In the Asia Pacific region, Beijing and Shanghai remain key hubs, while North America leads in future development. 

Across all markets, hyperscale operators such as AWS, Microsoft Azure and Google Cloud are consolidating their dominance. Goldman Sachs expects these providers to control 70% of capacity by 2030, up from 60% today.

The infrastructure challenge

Meeting this demand will require unprecedented investment.

US utilities alone will need an additional US$50bn in generation capacity for data centres, while global grid upgrades could reach US$720bn by 2030.

“Retrofitting existing facilities to support these massive jumps in power density is becoming complex and compromised. We will need new, purpose-built AI infrastructure to power the next generation,” says Frank Long, Vice President at the Goldman Sachs Global Institute.

Occupancy rates highlight the pressure, with utilisation projected to peak above 95% in 2026 before easing as new builds come online.

Energy mix, sustainability and the future

While efficiency gains once kept overall consumption stable, those improvements have slowed just as AI demand accelerates. 

Goldman Sachs expects 40% of new capacity to come from renewable sources, with onshore wind and solar offering competitive costs compared with gas. 

Yet the intermittent nature of renewables means hybrid solutions, combining green power with batteries and natural gas, remain essential.

Nuclear power is regaining momentum as technology companies look for reliable, low-carbon baseload options.

Contracts for more than 10GW of new nuclear capacity have been signed in the US alone.

Brian Singer from Goldman Sachs Research highlights how the sector’s sustainability commitments align with nuclear investment, while shifting political attitudes in markets such as the US and Switzerland are opening the door for more reactors.

New AI models, including China’s DeepSeek, may alter projections by delivering greater computational efficiency. 

Yet Goldman Sachs maintains that even with improved hardware and algorithms, the scale of AI adoption will continue to drive massive data centre demand.

Cooling systems, which account for up to 40% of hyperscaler energy consumption, will remain a central focus for efficiency improvements. 

As Goldman Sachs analysts note, “Longer term, we see potential for a significant reduction of data centre emissions intensity and potentially in absolute emissions” as new power sources and infrastructure strategies take hold.