Skip to content

Search the site


AI startup Imbue spends $150m on Dell servers as it eyes foundation model progress

The startup is running workloads in the data centres of non-profit AI compute provider Voltage Park.

Imbue CEO Kanjun Qiu, right, with investor Reid Hoffman.

AI startup Imbue, which is building its own foundation models, has inked a $150 million deal with Dell for servers to train its emerging models.

The two have co-designed a compute setup running inside AI cloud provider Voltage Park’s facilities, built on Dell PowerEdge XE9680 servers.

Imbue, formerly known as Generally Intelligent, raised $200 million in a Series B in September 2023. Investors in that round included Astera Institute; NVIDIA; Cruise CEO Kyle Vogt; Notion co-founder Simon Last.

The compute cluster, underpinned by NVIDIA H100 GPUs, includes smaller clusters “to support rapid experimentation on novel model architectures as well as rapid networking into a large cluster” the two said today.

Imbue, led by CEO Kanjun Qiu and CTO Josh Albrecht, said it is already using the cluster to develop “early prototype agents that can correct bugs in code and analyze lengthy documents” among other challenges. 

Imbue is among a burgeoning cohort of AI startups including AI21 Labs, Anthropic (backed with $2 billion by Google), Adept, Cohere, and others, as well as competing with the cloud hyperscalers and the likes of OpenAI. 

The news is good for Dell, which is keen to capture more of the AI workload opportunity as well as for specialist AI “cloud” provider Voltage Park, which runs a data centre packed with 24,000 NVIDIA H100 GPUs.

The organisation, set up as a non-profit, is backed by billionaire Jed McCaleb and lets organisations lease GPUs using an auction process

Its clusters consist of 80GB H100 SXM5 GPUs interconnected with 3.2T InfiniBand and the non-profit offers bare-metal access for large-scale users that need peak performance; it also plans to add short-term leases and hourly billing along with support for tools like Slurm, Kubernetes, and Mosaic for “easy integration into existing training frameworks” it said.

See also: With $1.3 billion MosaicML acquisition, Databricks sets itself up for open source generative AI fight