Goldman Sachs, Accenture & KPMG: The AI Scaling Crisis

The AI revolution has hit a wall – and it’s made of power cables and server racks. 

While tech giants pour billions into data centres and enterprises launch endless pilot projects, a peculiar paradox is emerging. 

We’re building the most powerful computational infrastructure in human history, yet somehow most companies cannot figure out how to actually use it.

Goldman Sachs research predicts that global data centres will consume 165% more electricity by 2030. 

Meanwhile, Accenture finds that only 8% of enterprises have managed to scale AI projects beyond the pilot stage.

The future challenges of AI infrastructure

Current data centres consume around 55GW of electricity globally.

Today, business applications like email consume about a third of that power, while AI workloads account for just 14%. 

But Goldman Sachs expects this picture to flip dramatically by 2027, with total consumption jumping to 84GW and AI claiming more than a quarter of all power usage.

A single ChatGPT query burns through 2.9 watt-hours of electricity – nearly ten times what a Google search requires. 

Where traditional cloud server racks might be less hungry for power, their AI equivalents are hungrier, consuming ten times as much electricity due to the intensive requirements of training and running machine learning (ML) models.

Europe faces a particularly daunting reality. 

“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.

Goldman Sachs estimates the continent needs a data centre pipeline of roughly 170GW – equivalent to a third of Europe’s entire current electricity consumption. 

US utilities alone need to spend US$50bn on new generation capacity just for data centres, while global grid upgrades could cost US$720bn through 2030.

Frank Long, a Vice President at the Goldman Sachs Global Institute, suggests: “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.”

The enterprise scaling crisis unveiled 

While this infrastructure race accelerates, most companies are struggling to even deploy AI.

Accenture’s survey of 2,000 C-suite and data science executives from billion-dollar companies finds a growing gap between ambition and execution. 

They sorted organisations into three categories: 42% are still struggling with limited pilots, 43% are making some progress with implementation – and just 15% achieve “AI reinvention-ready” status.

Within that final group, only 8% qualify as genuine front-runners who’ve successfully scaled multiple strategic AI initiatives across their operations. 

The barriers are both technical and deeply human. 

Poor data readiness remains the biggest stumbling block, especially with unstructured data. 

While legacy IT systems create friction – and building multi-disciplinary teams proves difficult. 

Gartner predicts at least 30% of AI pilots in 2025 will be abandoned due to poor data quality, weak governance, rising costs or unclear business value.

“We are writing the playbook for how to be the most AI-enabled, client-focused professional services company in the world,” says Julie Sweet, Chair and CEO, Accenture. 

The scaling priority: Walking before running 

So how do the 8% crack the code? 

KPMG’s research reveals that deploying AI isn’t enough – businesses need a structured approach and strong foundation to overcome scaling challenges. 

Its analysis identifies five critical strategies.

Adopt a phased approach: Focus on high‑impact, low‑risk AI use cases first to assess readiness, prove early value and build alignment for scaling.

Build a strong data foundation: Invest in governance, cleansing and integration to ensure quality, accessible data, since poor data guarantees poor AI outcomes.

Implement MLOps: Standardise deployment, monitoring and maintenance of AI models to improve reliability, efficiency and performance in production environments.

Upskill the workforce: Develop internal AI skills through training and expert partnerships, involving users early in pilots to reduce resistance and drive adoption.

Foster cross‑functional collaboration: Align AI initiatives with real business goals by engaging leadership, IT and operations, defining clear KPIs and consistently demonstrating value.

It seems that success isn’t about having the grandest infrastructure – it’s about knowing how to use it.