A Business Insider investigation into Amazon Web Services’ early adoption of Nvidia’s GB300 server systems formed the basis for analyst Ed Zitron’s speculation regarding the potential impact of AI-related shipments on GeForce shortages and prices. AWS had planned to receive equipment worth approximately $5 billion in August, and subsequently around $3 billion per month. There is no evidence that Nvidia deliberately restricted supplies of consumer GPUs.
Briefly about the main points
- AWS is rolling out the GB300 systems before some data centres are ready to deploy them.
- Deliveries totalling approximately $5 billion were expected in August.
- Some of the servers are awaiting installation due to a shortage of electrical power.
- Zitron suggests that AI demand is putting pressure on the availability of consumer GPUs.
- Amazon describes temporary storage as standard practice for large-scale construction projects.
AWS is changing its procedure for accepting server systems
According to Business Insider, AWS has introduced the Early Rack Receipt procedure. It allows server hardware to be accepted Nvidia and to process the purchase at an earlier stage of the supply chain, rather than only after the finished racks have been delivered to the data centres.
At some sites, there is physical space for new servers, but there is insufficient electrical capacity to power them. Some of the systems are therefore being temporarily stored in interim warehouses; at least one AWS team did not have a clear timetable for their deployment.
Purchased accelerators that are not yet operational or are not being used to their full capacity are referred to as ‘dark GPUs’. Amazon has denied that the purchases are haphazard: a company spokesperson described the storage of ready-made equipment near future installation sites as standard practice during large-scale data centre construction.
The theory regarding early purchases and pressure on the graphics card market
Ed Zitron suggests that Nvidia may be encouraging major customers to purchase AI hardware ahead of their actual needs. According to him, buyers may agree to such deliveries in order to secure access to future batches of processors.
The practice of pushing excess stock through sales channels is known as ‘channel stuffing’. The analyst believes that this practice could intensify competition for components and affect the cost of consumer electronics. TechRadar, when considering this theory, drew attention to the rising prices of GeForce RTX cards, particularly the high-end models.
This assumption is not a conclusion reached by Business Insider. The investigation confirmed the early adoption of AWS server systems, but did not establish that Nvidia is forcing customers to make purchases or restricting the supply of gaming graphics cards.
Differing recollections, common limitations and the status of evidence
GeForce cards use graphics memory from the GDDR family, whilst high-performance AI-Graphics cards are predominantly equipped with HBM. These are different technologies, but both segments depend on the shared semiconductor industry and limited investment resources. PC Partner, which owns the Zotac, Inno3D and Manli brands, has warned of a shortage of GDDR6 and GDDR7 and a further rise in graphics card prices in the second half of 2026.
The AI sector carries significantly more weight for Nvidia: in the second quarter of the 2027 financial year, which ended on 26 July, data centres accounted for $89 billion out of $96.2 billion in total revenue. If the pace of AI infrastructure development or future revenue from it fails to meet expectations, major technology companies may reconsider their accelerator procurement volumes.
As of 25 October, both the procurement of AI equipment and the readiness of some data centres, as well as issues with the availability of certain gaming graphics cards, have been confirmed. No direct causal link between these phenomena, nor any deliberate manipulation of GeForce supplies by Nvidia, has been proven.
