The funding frenzy in the AI inference chip sector is intensifying, with SambaNova positioned at the forefront of this wave.

According to a recent report by The Information, citing informed sources, chip startup SambaNova is seeking to raise $8 to $10 billion in a new funding round. Upon completion, this would value the company at approximately $100 billion, a staggering fivefold increase from its valuation just four months ago. Private equity giant General Atlantic is reportedly in talks to lead the investment, with a deal potentially finalizing as early as next week. This news was initially disclosed by SambaNova's Executive Chairman, Lip-Bu Tan, who also serves as the CEO of Intel (INTC), during an event.

This potential funding round, which could exceed $10 billion, would mark the largest in SambaNova's history. The dramatic valuation leap is driven by a combination of factors: steadily increasing customer orders, significant partnerships with Intel and private equity firm Vista Equity Partners, and the explosive market demand for AI inference chips. Vista Equity Partners announced earlier this month plans to create a cloud service provider combining Intel CPUs, SambaNova RDUs, and NVIDIA (NVDA) GPUs. SambaNova CEO Rodrigo Liang revealed this partnership comes with a $3.5 billion revenue commitment for the company.

Partnerships and Orders Fuel Growth

SambaNova's valuation trajectory highlights the intense heat in the AI inference chip arena.

Reports indicate that the company completed a $350 million funding round in February at a valuation of around $20 billion, with investors including Intel. Prior to that, Intel had reportedly engaged in talks to acquire SambaNova for approximately $16 billion, but negotiations ultimately fell through.

The current round would propel its valuation to roughly $100 billion. It is noteworthy that this figure remains below the $50 billion valuation SambaNova achieved during the peak of the tech bubble in 2021. However, that previous valuation was built on historically low interest rates and pandemic-driven tech euphoria, a stark contrast to the current growth narrative driven by business fundamentals.

Lead investor General Atlantic is not traditionally known for semiconductor investments but has previously placed bets in the AI space, including investments in Anthropic. Its move into SambaNova signals a rapidly expanding interest from traditional growth-focused private equity firms in the AI infrastructure sector.

A key catalyst for the valuation surge is the materialization of SambaNova's collaborations with Intel and Vista Equity Partners.

Intel holds approximately a 9% stake in SambaNova and is also a customer. In February, Intel announced an AI inference product that combines its CPUs with SambaNova's Reconfigurable Dataflow Unit (RDU).

This month, Vista Equity Partners furthered this integration by announcing plans for a cloud service provider that would combine Intel CPUs, SambaNova RDUs, and NVIDIA GPUs. While Liang mentioned a $3.5 billion revenue commitment from this deal, the company's current actual revenue scale remains undisclosed.

Additionally, SambaNova is collaborating with Saudi Aramco, providing hardware and software for the development of large language models. The company plans to release its fifth-generation chip this year.

Technological Edge: Challenging NVIDIA on Efficiency

SambaNova's core selling point lies in the energy efficiency advantage of its Reconfigurable Dataflow Unit (RDU). Rodrigo Liang stated that SambaNova's chips run AI models faster while consuming only about one-tenth the power of NVIDIA GPUs.

In terms of application, SambaNova has gradually shifted its positioning from being an "NVIDIA alternative" to an "NVIDIA collaborator." Under this architecture, NVIDIA GPUs handle the computationally intensive "prefill" stage of AI inference, while SambaNova's RDU takes on the subsequent "decode" part.

This strategy somewhat mitigates direct competition with NVIDIA and may ease its entry into the existing NVIDIA ecosystem.

Liang stated in an interview, "What we're really focused on is high-end inference. Everyone's services are built around NVIDIA; they've done an excellent job serving a broad market. But there is still a need for differentiation."

Despite the heated funding environment for AI inference chips, NVIDIA's market dominance remains unchallenged. Estimates suggest NVIDIA's share in this market has actually risen to 74% in recent years. NVIDIA CEO Jensen Huang maintains that the company's GPUs are more efficient than competitors' across all inference scenarios.

SambaNova faces a field of formidable competitors: Cerebras recently completed its public listing; Groq has been conditionally acquired by NVIDIA; cloud giants like Google, Amazon, and Microsoft are developing their own AI inference chips; Meta Platforms is also working on its own inference chip; and OpenAI announced this week a collaboration with Broadcom to co-design AI inference chips.

The report also notes that SambaNova faces potential supply chain bottlenecks. Its chips rely on High Bandwidth Memory (HBM), which is currently in tight supply, potentially constraining the company's scaling and path to profitability. Some competitors, like Cerebras, do not rely on HBM but could still be affected by broader memory shortages.