Corbenic's Galahad puts KV cache on server SSDs; its 14x claim draws scrutiny
TechRadar reports that Belgian startup Corbenic AI's Galahad, in beta since October 1 2026, writes a model's KV cache to encrypted local NVMe storage and reloads it after GPU memory is cleared or the server restarts, so the model need not reread the same text. Founder Sietse Schelpe says it needs only Linux, an Nvidia GPU and a standard NVMe drive, with no special hardware. The report notes, however, that the headline 14x figure owes most to a retrieval component called Blaise, which hands the model roughly 668 tokens instead of the whole corpus; with the KV component Taliesin alone, vLLM took 15.56 seconds per question versus an 8.11-second baseline, and SGLang 22.02 versus 10.29. It also notes that LMCache, Nvidia Dynamo, WEKA and Phison already persist KV cache to storage, and that Corbenic's numbers come from the founder's own runs on rented cloud GPUs in self-published preprints, with raw logs not yet released.
The report says paying a GPU to reread the same contracts, manuals and code on every request is a genuine waste, and that rivals such as LMCache and Nvidia Dynamo already persist KV cache to storage, so Galahad's edge is exact, fail-closed reuse in one small library rather than its speed numbers.
- Corbenic AI's Galahad beta saves a model's KV cache to encrypted local NVMe so servers can skip rereading text after GPU memory is cleared or after a restart
- Founder Sietse Schelpe tells TechRadar Pro no special hardware is needed, but the paper credits most of the 14x speedup to Blaise, a retrieval component, not SSD loading
- LMCache, Nvidia Dynamo and CMX, WEKA and Phison already persist KV cache to storage; Galahad's edge is exact, fail-closed reuse in one small library
Belgian startup Corbenic AI says it can stop GPUs from rereading documents they have already processed, and its founder told TechRadar Pro the trick needs nothing more exotic than the NVMe drive already sitting in the server.
Galahad, which entered beta on October 1 2026, saves a model's key-value (KV) cache, the attention state built while a model reads a prompt, to encrypted local storage and reloads it the next time the same text appears.
In an email to TechRadar Pro, Corbenic introduced Galahad with the line "the GPU never reads the same text twice," and founder and CEO Sietse Schelpe summed it up as "the SSD a server already has becomes part of the AI's working memory."
#Corbenic AI#Galahad#Taliesin#Blaise#LMCache#Nvidia#Phison#vLLM#SGLang#Gemma 4 31B#Tensormesh