Key Takeaways:
A San Francisco-based startup has raised $300 million from investors including Nvidia itself to create a "great equalizer" for the AI chip market.
Decart, an AI software startup, has secured $300 million in a new funding round to accelerate the adoption of its hardware-agnostic software, a direct challenge to the vendor lock-in that has helped make Nvidia the dominant force in AI computing. The round, led by Radical Ventures, values the one-year-old company at nearly $4 billion, a significant increase from its $3.1 billion valuation in August.
“Moving models between different hardwares is incredibly hard, and we’ve seen companies signing contracts worth hundreds of millions of dollars to help them shift models from one hardware to another,” Dean Leitersdorf, co-founder and chief executive of Decart, said in an interview. “We’re kind of the great equalizer.”
The company's core product, the Decart Optimization Stack (DOS), is a software engine that allows AI developers to move their models between processors from Nvidia, Amazon.com, Google, and others without costly custom programming. In a notable sign of the technology's potential, Nvidia joined the funding round, a counterintuitive move given Decart's platform directly reduces the friction of using competing chips. Other backers include Atreides Management and OpenAI co-founder Andrej Karpathy.
The investment highlights a critical pain point in the AI industry: a scarcity of computing power and a deep dependency on Nvidia's hardware, which controls an estimated 90 percent of the AI chip market. As companies pour billions into AI infrastructure, the ability to use any available chip from any vendor could dramatically lower costs and accelerate product timelines, a market opportunity Decart is now valued at nearly $4 billion to capture.
For years, Nvidia's primary competitive advantage has been not just its powerful GPUs, but its proprietary CUDA software environment. This software ecosystem made it difficult and expensive for developers to switch their AI models to run on hardware from other manufacturers. However, this software "moat" is being challenged as the industry increasingly adopts universal frameworks like PyTorch and JAX, which are designed to be hardware-agnostic.
Decart's DOS platform is a key enabler of this shift. It acts as an abstraction layer, hiding the underlying hardware's complexity from the developer. To a programmer, a cluster of mixed-and-matched chips can function like a single, unified system. This promises to commoditize the underlying hardware, forcing chipmakers to compete more directly on price and performance rather than relying on software lock-in.
Nvidia's participation in the funding round is a telling move. While DOS could erode its CUDA-based market dominance, the investment suggests Nvidia sees the shift to hardware abstraction as inevitable and wants a stake in a potential kingmaker. By investing, Nvidia gains intelligence on a crucial part of the ecosystem and hedges against the risk of being displaced by a purely software-driven solution.
"Every chip will be better for this," said Victor Lazarte, general partner at Benchmark, one of Decart's earlier investors. The sentiment suggests that a more fluid and competitive hardware market benefits the entire AI industry by speeding up innovation and lowering costs, a tide that could still lift all boats, including Nvidia's.
The $300 million infusion will allow Decart to aggressively hire engineers and pull forward its product roadmap. The company, founded in 2023 by Israeli engineers Dean Leitersdorf, Orian Leitersdorf, and Moshe Shalev, also develops "world models" for 3-D simulation, with Amazon's Twitch and retail divisions among its customers.
For investors, Decart's rise puts a spotlight on the software layer of the AI stack as a critical battleground. While Nvidia (NVDA) shares have soared on hardware sales, the success of companies like Decart could pressure its long-term margins by reducing switching costs for its customers. The move mirrors the dynamic seen in other tech sectors, where open standards and interoperability eventually triumph over closed, proprietary systems.
This article is for informational purposes only and does not constitute investment advice.
