Key Takeaways
- SambaNova raised $1 billion at an $11 billion valuation, announced at the Raise Summit in Paris
- The company focuses on AI inference chips rather than training, targeting a different market than Nvidia's core business
- Intel backs the California startup, which plans to invest across chips, systems, software, and full-stack AI infrastructure
SambaNova Systems announced a $1 billion funding round at the Raise Summit in Paris, valuing the AI chip startup at $11 billion. The California company, backed by Intel, is betting that inference-focused chips can challenge Nvidia's grip on the AI semiconductor market.
The funding comes as AI agents gain traction. These tools handle tasks from writing code to sending emails, and they run on inference, the process of generating responses from trained models. Training a model happens once. Inference happens every time a user asks a question. That distinction is central to SambaNova's pitch.
Why inference matters more than training for enterprise AI
Nvidia built its dominance on training. The company's GPUs power the data centers where OpenAI, Anthropic, and Google train their largest models. But once a model is trained, it needs to run, often millions of times per day across thousands of applications. That's inference.
SambaNova claims its chips deliver up to 8x better energy efficiency for inference workloads compared to traditional GPUs. For enterprises running AI at scale, energy costs and latency matter as much as raw performance. A company processing millions of customer queries daily will spend more on inference over time than it spent on training.
"Inference is where the action is going to be," Rodrigo Liang, SambaNova's CEO, has said in previous interviews. The company's Samba-1 model supports up to 5 trillion parameters, positioning it for enterprise deployments where customers want alternatives to hyperscaler offerings.
Where the money goes
SambaNova said it will use the funds to "continue investing across chips, systems, software, and full-stack AI infrastructure." That's a broad mandate, but it reflects the company's strategy: selling complete solutions rather than competing on chips alone.
The company has now raised over $1.1 billion total. Founded in 2017 by Stanford professors and former Oracle and Sun Microsystems executives, SambaNova spent years in relative obscurity while Nvidia captured the AI boom. The surge in AI agent usage since late 2025 has shifted attention toward inference workloads, giving challengers an opening.
Can anyone dent Nvidia's market share?
Nvidia holds roughly 80% of the AI chip market. Its data center revenue exceeded $80 billion in 2024. That dominance makes it the largest company in the world by market capitalization. Breaking into that market requires more than better specs on a slide deck.
SambaNova isn't alone in trying. Korean chipmaker Rebellions is planning an IPO in 2026, and AMD continues to push its Instinct line. But the software ecosystem around Nvidia's CUDA platform creates switching costs that pure hardware improvements can't overcome. Enterprises have built their ML pipelines on CUDA. Moving to a new chip architecture means rewriting code, retraining teams, and accepting new risks.
SambaNova's answer is the full-stack approach: provide the chips, the software, and the support so customers don't need to rebuild from scratch. Whether that works at scale remains unproven.
Another AI chip startup challenging incumbent players
The $11 billion question
An $11 billion valuation for a company that doesn't yet have meaningful market share is a bet on the future. Investors are pricing in a world where inference costs dominate enterprise AI budgets and where customers actively seek Nvidia alternatives. Both assumptions are reasonable. Neither is guaranteed.
The timing helps. AI agent adoption accelerated around the turn of 2026, creating new demand for cost-effective inference infrastructure. Every agent query burns compute cycles. Every agent task multiplied across millions of users creates the kind of sustained workload where inference efficiency pays for itself.
Logicity's Take
SambaNova's valuation reflects investor belief that inference will become the larger AI compute market, not just a complement to training. For enterprise buyers, this round validates a serious Nvidia alternative. But the real test comes in production deployments. Watch for customer announcements over the next 12 months. If major enterprises start reporting lower inference costs with SambaNova's stack, the $11 billion valuation will look prescient. If not, it's another AI bet that got ahead of execution.
Frequently Asked Questions
What does SambaNova do differently from Nvidia?
SambaNova designs chips optimized for AI inference rather than training. Its Reconfigurable Dataflow Architecture prioritizes energy efficiency and cost per query over raw processing power.
Who are SambaNova's investors?
Intel backs the company, along with other institutional investors. The latest $1 billion round was announced at the Raise Summit in Paris.
What is AI inference versus AI training?
Training builds a model by processing massive datasets. Inference runs that trained model to generate responses. Training is a one-time cost; inference happens continuously whenever users interact with AI applications.
How much has SambaNova raised in total?
The company has raised over $1.1 billion across all funding rounds, with the latest $1 billion round valuing it at $11 billion.
Need Help Implementing This?
Evaluating AI infrastructure options for your enterprise? Logicity helps technical teams assess chip architectures, benchmark inference costs, and plan migration strategies. Contact us for a consultation.
Source: Tech-Economic Times / ET
Manaal Khan
Tech & Innovation Writer
Produced with AI assistance and reviewed by the Logicity editorial team. Learn more in our Editorial Policy.
