Read latency reduction
Company: CoreWeave
The claim, verbatim
CoreWeave AI Object Storage with LOTA reduces p99 read latency by more than 8x for cached reads compared to bucket reads
Source (primary)
CoreWeave AI Object Storage: Cross-region writes, Archive - CoreWeave (-, news_article)
View cached copy (2026-09-19)Live source ↗
Quote: “On those hits, p99 read latency is more than 8x lower than reading from the bucket”
How we checked this
Checked on September 24, 2026. The cited source supports part of this claim, but not all of it.
The more-than-8x p99 figure comes from one named-but-unidentified customer's large deployment, not a general benchmark; the claim presents it as a general property of the product.
Confirmed in the source:
- At one leading frontier model provider's deployment of LOTA (more than 15,000 GPUs, 20 PB of cache), p99 read latency on cache hits was more than 8x lower than reading from the bucket
- The blog's subtitle claims read latency can be reduced by 8x
We could not confirm this from the cited source:
- That LOTA reduces p99 read latency by more than 8x as a general property, rather than as a result measured at one customer deployment
That does not mean it is false — only that this source does not establish it, as of the date above. If we find a source that settles it, we will re-check and update this page.
What we did: Read our cached copy of the publisher (https://www.coreweave.com/blog/new-in-coreweave-ai-object-storage-cross-region-writes-and-archive-storage) in full (16,354 characters, retrieved September 19, 2026) and checked each assertion in the claim against it.
Additional evidence
confirms CoreWeave AI Object Storage: Cross-region writes, Archive - CoreWeave
Quote: “On those hits, p99 read latency is more than 8x lower than reading from the bucket”