Concentration of Global Compute Demand
Market projections indicate a significant geographic concentration in the demand for high-performance computing. According to data from Hynix, the United States and China are expected to account for 95% of global compute demand by the first quarter of 2027. This concentration has drawn criticism from observers in Europe, where some suggest the region is being left behind in the race for hardware and infrastructure.
While some analysts suggest that emerging markets in Latin America and Africa may struggle to spin up necessary infrastructure over the coming decades, the current landscape is dominated by a few key players. This shift toward centralized, high-capacity compute is driving massive capital expenditures across the technology sector.
Infrastructure Spending and Market Realities
The scale of investment in artificial intelligence (AI) infrastructure is substantial. The “big four” hyperscalers—AMAZON.COM INC, MICROSOFT CORP, ALPHABET INC, and META PLATFORMS INC—alongside Oracle, have reportedly spent over $1 trillion on AI infrastructure and data centers since 2020. Projections suggest these companies may spend hundreds of billions more through 2026.
Despite the high level of expenditure, there is a debate regarding whether the current build-out is sufficient. While some critics argue that the industry is facing a bubble, others contend that data centers are actually being underbuilt. Unverified reports suggest that companies such as OpenAI and Anthropic are currently unable to launch certain products due to compute shortages, forcing them to use “quantized” or lower-parameter models to meet existing demand.
Concerns Over Circular Revenue Models
A more controversial aspect of the current market involves the financial relationship between hyperscalers and AI developers. Unverified claims circulating in online financial discussions suggest a potential “circular revenue” loop that could pose systemic risks.
The theory posits that hyperscalers provide capital to AI startups through massive funding rounds, which the startups then immediately use to pay the hyperscalers for compute services. This cycle may artificially inflate the revenue figures of the hyperscalers. For example, unconfirmed data suggests that OpenAI could account for approximately 70% of Microsoft’s AI-related revenue for the 2026 fiscal year. While this represents only a small fraction of Microsoft’s total revenue, the heavy reliance on a single client for a specific growth segment has led some observers to warn of a “too big to fail” scenario.
If these revenue streams are heavily dependent on promissory agreements and non-profitable AI entities, the potential for a market correction remains a subject of intense debate among market participants.