Rebellions Raises $400M at $2.3B Valuation in Pre-IPO Round, Launches Two New Inference Platforms

South Korean fabless AI chip startup Rebellions has raised $400 million in a pre-IPO funding round led by Mirae Asset Financial Group and the Korea National Growth Fund, the company announced Monday. The round values Rebellions at approximately $2.34 billion — and pushes its total capital raised to $850 million, with $650 million of that arriving in the last six months alone.

A Funding Trajectory Built on Inference

Rebellions’ financial story is one of rapid acceleration. The company closed $124 million in a Series B in January 2024 in collaboration with Samsung, then followed with a $250 million Series C last November. The latest $400 million pre-IPO round is the largest single raise in the company’s history and arrives as it prepares for a public listing later this year. Chief Business Officer Marshall Choy, who is leading the company’s global push, declined to comment on IPO timing.

The company’s focus is inference — the compute workload required for deployed AI models to respond to user queries in real time. As large language models have matured and entered commercial production at scale, inference has grown from a secondary consideration into a primary cost center for AI operators. Rebellions’ bet is that the inference workload is large enough, and distinct enough in its performance requirements, to justify purpose-built silicon rather than adapting training-optimized hardware after the fact.

Two New Products: RebelPOD and RebelRack

Alongside the funding announcement, Rebellions unveiled two new inference infrastructure platforms. RebelPOD is described as a production-ready unit of inference compute — a self-contained, deployable building block for organizations running AI workloads at scale. RebelRack integrates multiple PODs into a larger cluster architecture designed for large-scale AI deployment, offering the kind of scalability that cloud providers, neoclouds, and enterprise data center operators require.

The product names signal a deliberate positioning: Rebellions is not just selling chips but packaging them into complete infrastructure units that reduce integration complexity for buyers. That stack-level approach mirrors what Amazon has done with its Trainium lab — building not just silicon but the surrounding software and hardware ecosystem to make custom silicon commercially viable.

Global Expansion: US, Middle East, Japan, Taiwan

Rebellions has recently established legal entities in the United States, Japan, Saudi Arabia, and Taiwan as part of an aggressive internationalization push. In the US, the company is targeting cloud providers, government agencies, telecom operators, and neoclouds as its primary customer segments — a go-to-market that mirrors how hyperscale custom silicon programs have typically built early traction.

The Middle East expansion is strategically timed. Gulf sovereign wealth funds and national AI programs in Saudi Arabia and the UAE have been deploying capital aggressively into AI infrastructure, and a purpose-built inference chip from a non-US vendor fits neatly into regional diversification goals for AI supply chains.

The Competitive Context

Rebellions is one of the most closely watched names in a new generation of semiconductor startups challenging Nvidia’s position. The challenge has gained credibility precisely because hyperscalers have demonstrated that custom silicon at scale is achievable: Google’s TPU program has been running in production for years, and Amazon’s Trainium has now won over Anthropic, OpenAI, and Apple as customers. Meta has its MTIA program. The question for Rebellions is whether it can do what those internal programs did — at an external commercial scale and without the captive workload advantage the hyperscalers enjoy.

CEO Sunghyun Park framed the company’s answer in terms of the economics that matter most to operators right now: “AI is now measured by its ability to operate in the real world at scale, under power constraints, and with clear economic return. That shifts the center of gravity toward inference infrastructure and software that makes that infrastructure usable.”

The pre-IPO raise suggests investors believe the answer is credible enough to fund a public market test.