The surge in artificial intelligence (AI) demand is enabling hyperscale cloud providers to exert significant control over the enterprise hardware supply chain, a development that could fundamentally alter how businesses acquire compute and storage infrastructure. This increasing market power may leave enterprises with little choice but to rent capacity from these cloud giants, according to industry observations.Nutanix chief executive officer (CEO) Rajiv Ramaswami highlighted this trend in May, noting that the fastest way for businesses to access new servers was through hyperscalers, rather than waiting for traditional hardware providers. This observation underscores the superior buying power of hyperscalers, who are prioritised by component suppliers in the current supply-constrained environment.Evidence of this preferential treatment is apparent across the hardware ecosystem. Memory manufacturers Micron and SK Hynix have publicly disclosed long-term agreements that guarantee supplies to their largest customers, securing historically high margins. Hard disk maker Seagate has adopted a similar strategy. Furthermore, AMD has established “sweetheart deals” with major AI players like OpenAI and Meta for hardware procurement.
Cloud Giants’ Economic Leverage
Meta CEO Mark Zuckerberg recently discussed the company’s ambition to establish an infrastructure-as-a-service business, leveraging its balance sheet to fund capital expenditure. Zuckerberg noted that Meta is already receiving “a lot of offers for compute at a significant premium over what we paid for it,” suggesting the company acquires hardware at favourable rates. Once Meta enters the AI infrastructure rental market, it will possess the tools to extend this model to conventional compute and storage services.Amazon CEO Andy Jassy has also detailed the economic advantages for Amazon Web Services (AWS). Jassy informed investors that AWS typically recoups its spending on servers and networking equipment in under three years, while the useful life of these servers extends to at least five to six years. He added that most AI capacity contracts are for at least five-year terms, generating “significant free cash flow” after the breakeven point. Jassy anticipates these economics will further improve as AWS shortens its breakeven times. Amazon also expects its data centres to last 30 years, with hardware margins increasing once the buildings are paid for.
Implications for Enterprise and Traditional Vendors
While Amazon currently ties infrastructure spending to predicted demand, the scenario where hyperscalers secure the majority of available hardware could force more organisations to rent capacity, thereby driving demand towards them. Traditional server manufacturers such as Dell, HPE, Lenovo, and Supermicro still offer the argument of fewer billing surprises when buying hardware, albeit at the cost of less elastic infrastructure.However, major cloud providers do not view this as a significant threat, anticipating years of sustained growth. Jassy projected that AWS, which he once believed could become a few hundred billion-dollar revenue business, could now potentially double that, or even evolve into a “trillion-dollar annual revenue business” for Amazon.Smaller cloud providers, who often compete by undercutting larger players, also appear to be affected by supply chain disruptions. OVH, for instance, has implemented steep price increases, while its larger rivals have not yet raised customer charges. This consolidation benefits hyperscalers’ suppliers, allowing them to focus on a smaller number of large customers, which reduces sales and marketing costs and helps preserve margins. The source notes that no current enterprise hardware vendor has a clear path to becoming a trillion-dollar company, with only Dell exceeding $100 billion in revenue. This leaves organisations preferring to own infrastructure in a challenging position, facing months-long waits for hardware and uncertainty regarding vendor quotes, a situation largely exacerbated by the demand for AI infrastructure.