Imbue, formerly Generally Intelligent, which claims to have access to 10K Nvidia H100 GPUs to train its AI “agents”, raised a $200M Series B at a $1B valuation
Forbes
Context & Ripple Effects
Imbue’s financing tied an AI-agent developer’s prospects to access to scarce high-end training hardware. That connection became more concrete when the company later reached a $150M hardware agreement with Dell to support development of AI systems and advanced models.
Related coverage also shows the same infrastructure layer becoming a business in its own right: Together AI raised capital to give developers access to Nvidia server chips, including a $106M round at a $1.25B valuation.
First-order effects
Imbue gains $200M of new financing and a $1B valuation benchmark, strengthening its capacity to pursue compute-intensive AI-agent development.
The reported access to 10,000 Nvidia H100 GPUs makes compute availability a central operational input rather than a background technical detail for Imbue.
Second-order effects
Hardware vendors and infrastructure partners gain a well-funded customer pursuing large-scale AI development; Imbue’s later Dell agreement shows how model funding can translate into equipment demand.
Startups building AI agents face a sharper capital-and-compute comparison: funding alone is less differentiating when access to advanced GPUs is also required to train systems.
Third-order effects
If this financing-and-hardware pattern persists, AI startup competition may increasingly sort around the ability to finance and secure compute, concentrating advantage among companies with both capital and infrastructure relationships.
The emergence of providers such as Together AI suggests that compute access could also become a commercial service layer, lowering some entry barriers while leaving top-tier training capacity scarce.
The trend: AI development is becoming increasingly capital-intensive, with startup valuations and product ambitions tied closely to access to specialized compute infrastructure.
We believe seriously using our agents every day is needed to push both reasoning models, and agent tooling & interface. Many of our agents write code, since that's our main work. Ultimately, we aim to release systems that allow anyone to build robust, safe, custom AI agents.
@kanjun
@kanjun on x
Today's AI systems are powerful, but they have limited ability to complete even simple goals for us. Reasoning helps AI systems know when to ask questions, change approach, revise conclusions, break down problems, manage risk, & generally deal with the real world's complexity.
@arian_ghashghai
Arian Ghashghai on x
People going nuts about the “bubble” Like Mistral, Imbue is an AI research company, not an applied AI company. Research is resource intensive and requires a lot capital to have any chance of succeeding— big rounds are the only way to play the game. Not sure if the game is smart
@basedbeffjezos
@basedbeffjezos on x
Me: let me raise several rounds at incremental valuations to make sure all my employees get to eat. The market: *$200M@1B pre-product*
@drjimfan
@drjimfan on x
Congrats Kanjun! I had fond memories of us discussing Avalon & MineDojo at NeurIPS last year. Glad to see the company behind Avalon grow to unicorn!
@nealkhosla
Neal Khosla on x
Big fan of @kanjun and @joshalbrecht. They've were talking about AGI 6 years ago when I met them in YC. Now that compounding work is paying off. Excited to see where this goes next.
@kenrickcai
Kenrick Cai on x
Exclusive from @alexrkonrad and I on a new $1B valuation for an AI research lab with an unusual backstory and an even more unusual lead backer. Imbue (fka Generally Intelligent) says its AI agents, which might one day solve complex tasks on their own, are years away from launch.
@cailen
@cailen on x
The Dropbox mafia likes to speak softly, but we carry big sticks! Congrats @kanjun!
@kanjun
@kanjun on x
I'm excited to share our $200M Series B at a $1B+ valuation to develop AI systems that reason! We believe reasoning is the main blocker to effective AI agents. We train large models tailor-made for reasoning on our ~10K GPU cluster. On top, we prototype agents we use every day. […