A.I. Chip Challengers Draw Fresh Bets

The rush to find winners in artificial intelligence hardware is no longer confined to Nvidia and its closest suppliers.

Two younger chip companies, SambaNova and Rebellions, are drawing fresh money and making plans that suggest investors still see room for new contenders in a market dominated by the world’s most valuable semiconductor company. SambaNova said it had secured $1 billion in the first close of a new financing round that values the company at $11 billion, while Rebellions, a South Korean A.I. chip startup, said it was aiming for an initial public offering in South Korea in the first half of 2027.

Taken together, the moves offer a more concrete sign that the A.I. boom is broadening into a wider financing wave for companies promising alternatives to Nvidia’s chips, especially for so-called inference — the stage when trained A.I. models are put to work answering questions, generating content or carrying out tasks.

Capital Chases Alternatives

SambaNova’s latest financing was led by General Atlantic, with additional investors expected to join later in the round. The company had already announced a $350 million fund-raising in February that included Vista Equity Partners, Cambium Capital and Intel Capital.

The new valuation underscores how heavily investors are still betting on specialized A.I. infrastructure, even after years in which Nvidia tightened its grip on the market through a mix of powerful chips, software tools and close ties to the largest cloud computing companies.

SambaNova, based in Palo Alto, Calif., has spent much of the past several years trying to distinguish itself not simply as a chip designer, but as a provider of integrated A.I. systems for enterprises. More recently, it has sharpened that pitch around inference, an area many startups believe offers an opening because customers are increasingly focused on lowering the cost of running A.I. models at scale, not just training them.

That positioning has been reinforced by its partnership with Intel, announced earlier this year, part of a broader effort to tie its technology to established manufacturing and enterprise channels. Intel’s chief executive, Lip-Bu Tan, has served as SambaNova’s executive chairman, and Reuters has previously reported that Intel had at one point discussed acquiring the company.

A South Korean Contender Looks to Go Public

Rebellions is making a different bet: that public investors will be ready to back an A.I. chip challenger with ambitions beyond South Korea.

Park Sunghyun, the company’s chief executive, said Rebellions was leaning toward a listing on the benchmark KOSPI exchange rather than the tech-heavy KOSDAQ market, with a potential U.S. listing possible later. The company previously raised about $400 million in a pre-I.P.O. financing at a valuation of roughly $2.3 billion.

Rebellions has emerged as one of South Korea’s most closely watched semiconductor startups, aided by backing from prominent domestic investors and ties to Samsung. Like other upstarts in the field, it has been trying to move beyond selling individual chips and instead offer more complete systems for data centers, including rack-scale products meant for real-world deployment.

That matters because customers buying A.I. infrastructure increasingly want a fuller package: hardware, software, networking and support that can be installed and operated without having to assemble a complex system themselves.

Why the Market Is Opening

The willingness to fund and potentially list these companies comes at a moment when the economics of A.I. are shifting. During the first phase of the generative A.I. race, spending centered on the immense computing power needed to train frontier models. Nvidia became the clear beneficiary, with its graphics processing units becoming the default choice across the industry.

But as companies move from experimentation to deployment, inference is becoming a larger piece of the business. Running chatbots, search tools, coding assistants and enterprise A.I. services continuously and cheaply is emerging as a commercial problem in its own right. That has created an opening for chip startups that argue they can offer lower costs, better efficiency or systems more tailored to specific workloads.

Investors appear willing to finance that argument, at least for now, if startups can show they have customers, manufacturing partners and a plausible path to scaling production.

The Obstacles Ahead

Still, raising money is not the same as breaking Nvidia’s hold on the market.

Any challenger faces a formidable set of barriers: Nvidia’s deep software ecosystem, the familiarity developers have with its tools, its entrenched position with cloud providers and its command of supply chains that remain constrained across advanced semiconductors. For customers, switching away from Nvidia can mean technical risk, retraining engineers and uncertainty about long-term support.

That leaves open the central question for companies like SambaNova and Rebellions: whether financing momentum can be converted into durable commercial traction.

For SambaNova, investors will be watching whether the latest round attracts more backers and whether its enterprise-focused inference strategy produces sizable deployments rather than pilot projects. For Rebellions, much depends on whether its timetable for a Korean listing holds and whether public market investors prove as enthusiastic as private ones.

What is becoming clearer, though, is that the surge in spending on artificial intelligence is creating enough demand — or at least enough belief in future demand — to support a broader cast of hardware hopefuls. Nvidia remains the dominant force, but the capital now flowing to challengers suggests investors are increasingly willing to wager that the next phase of the A.I. build-out will require more than one kind of chip.