Reka AI, which aims to make more efficient LLMs, raised $110M from Nvidia, Snowflake, and others at a $1B+ valuation, up from ~$300M in 2023, per PitchBook
Snowflake had previously been in talks to buy the three-year-old AI startup. — Artificial intelligence company Reka AI raised …
Context & Ripple Effects
Reka emerged with funding to build custom enterprise AI models, including backing from Snowflake and others in its earlier $58M launch financing. That positioned the company as a specialist model developer rather than a general-purpose consumer AI provider.
Snowflake later explored acquiring Reka for more than $1B, but those discussions broke down. The new round gives Reka a separately financed path at a valuation above that earlier reported threshold.
First-order effects
- Reka gains $110M to continue developing efficient LLMs while remaining independent after Snowflake's unsuccessful acquisition discussions.
- Nvidia and Snowflake become investors in Reka's new round, tying two important AI ecosystem participants more directly to the startup's outcome.
Second-order effects
- The outcome illustrates an alternative to an outright purchase for Snowflake: retaining exposure to a model developer through investment after an M&A process fails.
- Other enterprise-focused LLM developers may face a higher bar to show both model efficiency and strategic relevance as investors back Reka at a sharply higher valuation than its 2023 round.
Third-order effects
- If similar financings persist, strategic AI buyers may increasingly use minority investments alongside, or instead of, acquisitions to preserve access to specialized model developers.
- That would reinforce a more financially interconnected AI stack, where infrastructure and data-platform companies participate in startup financing as well as serving or partnering with those companies.
The trend: AI infrastructure and enterprise-platform companies are increasingly using strategic capital to secure exposure to specialized model developers without necessarily buying them outright.