Amazon Web Services is experiencing significant operational hurdles due to its massive investment in Nvidia GPUs for artificial intelligence workloads. Despite receiving approximately $5 billion in Nvidia GB300 racks in August alone, and continuing with $3 billion monthly shipments, AWS is struggling with insufficient energy capacity to power this new hardware. This situation has contributed to a drastic drop in AWS's free cash flow, reportedly moving the division into negative territory.
An internal memo cited by Business Insider revealed that AWS CFO John Felton humorously acknowledged the financial strain to an executive leading AI infrastructure, stating, "You've taken us free cash flow negative as a company, so congratulations." This highlights the scale of the challenge, as AWS, despite its experience in data center construction, is grappling with the pace and sheer volume of its AI infrastructure build-out. The company typically pays for GPU racks upon installation, but with power delivery timelines uncertain, AWS has implemented an "Early Rack Receipt" process. This means substantial capital is tied up in hardware that cannot yet be utilized, leading to the reported negative free cash flow.
Despite these immediate financial and logistical issues, AWS shows no signs of slowing its acquisition of GPU racks or its AI infrastructure expansion. The cloud division reported a 37% year-over-year revenue growth in its second quarter, its fastest growth in 18 quarters, generating $42.2 billion. This robust financial performance fuels further investment.
Amazon CEO Andy Jassy has characterized AI as a "once-in-a-lifetime opportunity" and outlined plans to invest nearly $200 billion in its infrastructure. Jassy expressed confidence to investors about achieving "strong financial returns" on this investment. However, the immediate future is complicated by the power supply issues, which could leave a significant portion of this investment idle.
This surge in demand from major cloud providers like Amazon is also contributing to the scarcity of high-end consumer GPUs. The Nvidia RTX 5090, for instance, is frequently unavailable from reputable retailers and can command prices upwards of $9,000 on reseller marketplaces. This production shift away from consumer hardware towards AI-focused enterprise solutions leaves PC gamers facing prolonged shortages and inflated prices.