A&M: AI Infrastructure Market in Structural Growth... Emergence of 'Neoscalers'

Securing Power, Land, and Long-Term Customer Contracts Is Key
Emergence of Neoscalers
Responding to Dedicated GPU and Sovereign AI Demand

There is growing analysis that the global artificial intelligence (AI) infrastructure market has entered a structural growth phase, moving beyond a period of temporary investment expansion. With demand for 'inference' rising rapidly, so-called 'neoscales'—companies that complement traditional hyperscalers (large data center operators)—are emerging as a new axis of competition.

On August 12, global management consulting firm Alvarez & Marsal (A&M) announced the publication of its latest report, "The Rise of the Neoscaler: Global Competition for Building AI Infrastructure," which analyzes changes in the global AI infrastructure market and investment opportunities.

A&M projected that total capital expenditure plans this year by the United States' five major hyperscalers—including Google and Amazon Web Services (AWS)—will exceed 750 billion dollars (approximately 1,061 trillion won), diagnosing that demand for AI infrastructure has entered a period of structural growth.

The Five Major U.S. Hyperscalers to Surpass 750 Billion Dollars in Capital Expenditure Plans This Year

First, as the AI market's focus expands from model development and training to real-world service applications, the importance of inference infrastructure is increasing. Currently, inference accounts for more than 40% of AI computing, and this share is expected to rise to over 75% of new demand in the future.

However, expanding supply will take time. Not only do electricity grid connections, site acquisition, regulatory approvals, and data center construction require significant time, but procurement of key components is also lengthy, resulting in supply failing to keep pace with the surge in demand.

As it becomes increasingly difficult for traditional hyperscalers alone to meet the various demands of AI, neoscales such as OpenAI and CoreWeave are emerging. Neoscales are businesses that specialize in AI model or graphics processing unit (GPU) infrastructure, enabling large-scale training, inference, and responding to sovereign AI demands, among others.

A&M Global Report Cover

"Korea Has Opportunities in Industry-Specialized & Sovereign AI... Real Demand and Contract Stability Must Be Verified"

A&M expects that neoscales will play complementary roles rather than fully replacing existing hyperscalers, who have strengths in broad customer bases and general-purpose cloud services. Going forward, the market is likely to segment into differentiated players, each addressing diverse needs—such as dedicated GPU clusters and sovereign AI infrastructure projects tailored to specific customer demands.

Myeonggu Kim, Partner at A&M, stated, "In today’s AI infrastructure business, competitiveness depends not only on being able to secure large volumes of GPUs, but on how quickly and reliably one can build and operate custom infrastructure for clients, and whether these can be converted into long-term contracts."

He added, "In Korea as well, investment opportunities can be found in industry-specific AI tailored to sectors such as manufacturing and finance, as well as sovereign AI and enterprise private AI. It is essential to evaluate not only infrastructure conditions such as power and land, but also customers’ creditworthiness and contract stability as a whole."

This content was produced with the assistance of AI translation services.