Databricks announced on September 13, 2025, that it’s investing in a new AI hardware startup led by Naveen Rao, its former AI vice president.
The startup, yet to be named, aims to revolutionize AI computing with energy-efficient machines, tackling the massive power demands of AI training.
Here’s the key info you need, with links to dive deeper, crafted to boost discoverability with timely, engaging tech news.
What’s Happening?
CEO Ali Ghodsi revealed the investment via LinkedIn, praising Rao’s vision to “revolutionize the AI hardware space.”
Rao, who founded Nervana (acquired by Intel) and MosaicML (acquired by Databricks for $1.3B in 2023), is focusing on “brain-scale efficiency” to create a new, power-efficient computer for AI workloads.
Details on funding, company name, and products are still under wraps, but Rao will advise Databricks while leading the venture.
Why It Matters
AI training is energy-intensive, with models like GPT-4 costing millions in electricity. Rao’s startup aims to cut these costs, potentially challenging NVIDIA’s 92% GPU market dominance with custom silicon and software.
Databricks, fresh off a $10B funding round valuing it at $62B, sees this as a strategic move to boost its AI platform while fostering ecosystem innovation.
Rao’s Track Record
- Nervana Systems (2014) : Built AI chips, acquired by Intel.
- MosaicML (2021) : Pioneered efficient AI training, acquired by Databricks.
- New Startup (2025) : Focuses on low-power AI hardware, with early investor talks hinting at a high valuation.
What’s Next?
- Funding Updates : Expect details on investors and valuation soon.
- Product Timeline : Prototypes may emerge by mid-2026, targeting enterprise AI needs.
- Databricks’ Role : Rao’s advisory work could integrate his hardware into Databricks’ platform.
Why You Should Care
This startup could lower AI’s environmental and financial barriers, making advanced AI accessible to more businesses. It’s a bold bet in a market where energy efficiency is critical, aligning with trends like AWS and Google’s custom chip efforts.
Databricks announced on September 13, 2025, that it’s investing in a new AI hardware startup led by Naveen Rao, its former AI vice president.
The startup, yet to be named, aims to revolutionize AI computing with energy-efficient machines, tackling the massive power demands of AI training.
Here’s the key info you need, with links to dive deeper, crafted to boost discoverability with timely, engaging tech news.
What’s Happening?
CEO Ali Ghodsi revealed the investment via LinkedIn, praising Rao’s vision to “revolutionize the AI hardware space.”
Rao, who founded Nervana (acquired by Intel) and MosaicML (acquired by Databricks for $1.3B in 2023), is focusing on “brain-scale efficiency” to create a new, power-efficient computer for AI workloads.
Details on funding, company name, and products are still under wraps, but Rao will advise Databricks while leading the venture.
Why It Matters
AI training is energy-intensive, with models like GPT-4 costing millions in electricity. Rao’s startup aims to cut these costs, potentially challenging NVIDIA’s 92% GPU market dominance with custom silicon and software.
Databricks, fresh off a $10B funding round valuing it at $62B, sees this as a strategic move to boost its AI platform while fostering ecosystem innovation.
Rao’s Track Record
- Nervana Systems (2014) : Built AI chips, acquired by Intel.
- MosaicML (2021) : Pioneered efficient AI training, acquired by Databricks.
- New Startup (2025) : Focuses on low-power AI hardware, with early investor talks hinting at a high valuation.
What’s Next?
- Funding Updates : Expect details on investors and valuation soon.
- Product Timeline : Prototypes may emerge by mid-2026, targeting enterprise AI needs.
- Databricks’ Role : Rao’s advisory work could integrate his hardware into Databricks’ platform.
Why You Should Care
This startup could lower AI’s environmental and financial barriers, making advanced AI accessible to more businesses. It’s a bold bet in a market where energy efficiency is critical, aligning with trends like AWS and Google’s custom chip efforts.
