Source item

Here’s a focused analysis of the widely discussed The Information report about Oracle. I believe this analysis helps clarify the report’s logic. I’m not commenting on the report’s motives, nor on Oracle’s share price. The following two passages from the report are pivotal: they indicate the period covered (June–August 2025) and the scope (rentals of small quantities of both new and older Nvidia GPUs). -- “In the three months that ended in August, Oracle generated around $900 million from rentals of servers powered by Nvidia chips and recorded a gross profit of $125 million.” “In some cases, Oracle is losing considerable sums on rentals of small quantities of both newer and older versions of Nvidia’s chips, the data show.” -- From an industry-research perspective, here is what Oracle was doing—and encountering—during June–August 2025. Three key points: 1. Based on my industry surveys, Oracle accounted for roughly 12% of global GB200 NVL72 shipments in 2025. That makes it a major buyer, but still behind Microsoft (~30%), Google (~16%), and Dell (~14%). As a result, Oracle’s deliveries arrived later—around late 2Q25—vs. initial small-batch GB200 NVL72 shipments starting in 1Q25. 2. In June–August 2025, Oracle was only beginning to receive and deploy GB200 NVL72 systems. Consequently, it could offer just limited Blackwell compute capacity, and rentals remained weighted toward the prior-generation Hopper. 3. Early GB200 NVL72 deployments are unlikely to be profitable because: 1) AI server costs rise significantly with the new platform; 2) infrastructure retrofit expenses are front-loaded; and 3) limited initial compute/service scale constrains revenue and margins. These are standard early-phase dynamics in infrastructure transitions (e.g., conventional server platform transition of data center or 4G to 5G upgrades). The preceding analysis should help to clarify the logic of The Information's report, which can be broken down into two key points: 1. Why “small quantities”? Because Blackwell/GB200 NVL72 had only just arrived and was being readied for service, leaving limited compute capacity available to rent. 2. Why the losses? Because the calculation conditions focus on “small quantities of both newer and older Nvidia chips.” Losses largely reflect the early-phase costs of the Hopper-to-Blackwell transition. If the lens were instead “small quantities of new chips plus large quantities of older chips,” or “large quantities of older chips alone,” the result would likely not show a loss. Report source: https://t.co/IHBjeqRi03