Should the acquisition take place, Intel would gain a developer of chips, systems, and clouds aimed squarely at AI inference and give more direction to Intel’s AI strategy which has seen several ups and downs.
Intel is exploring beefing up its AI capabilities by acquiring SambaNova Systems, a developer of custom-designed chips and hardware, software, and pre-trained models for AI, Bloomberg reported Thursday.
Bloomberg, citing people familiar with the move, said that SambaNova would likely be valued well below its peak valuation of over $5 billion.
Intel declined to respond to a CRN request for more information. SambaNova did not respond to a CRN request for more information by press time.
[Related: As Intel Creates New AI Group, Data Center Division To ‘Refocus’ On CPUs: Memos]
Palo Alto, Calif.-based SambaNova develops a number of product lines related to AI infrastructure, including:
- SN40L RDU (reconfigurable dataflow unit) chip – This is a fourth-generation RDU aimed at providing the performance needed for running AI models.
- SambaRack – A system built of multiple modules featuring 16 SN40L RDUs.
- SambaCloud – A cloud built on the SN40L RDU chips to support a range of inference models including DeepSeek, Llama, and Qwen.
- SambaManaged – A fully managed inference cloud that can be built in a customer’s current data center infrastructure.
News of the possible acquisition comes just a few days after The Information reported that SambaNova is exploring a sale after struggling to complete a fundraising round. SambaNova has raised a total of $1.14 billion in funding and had a valuation of $5.1 billion as of 2021, according to PitchBook.
Intel CEO Lip-Bu Tan is well connected with SambaNova. Tan has served as executive chairman of SambaNova since May 2024. In addition, Tan is the chairman of Walden International, one of SambaNova’s investors.
Intel Capital is also an investor in SambaNova, as well as SoftBank’s Vision Fund, which is also an investor in Intel. Nvidia is also a SambaNova investor.
Intel has struggled with the AI market. The company in January canceled plans to sell its next-generation Falcon Shores AI accelerator chip, which it had hoped to release sometime this year. The company at the time said it wanted to focus on developing a “system-level solution at rack scale” with a successor chip it’s calling Jaguar Shores.
Intel this month unveiled a 160-GB energy-efficient GPU code-named “Crescent Island” for use in inference workloads running on air-cooled enterprise servers. Crescent Island is slated to start shipping in the second half of 2026. At the same time, the company did not provide an update for its Jaguar Shores next-generation GPU.
