AI startup Rebellions raises $400 million at a $2.3 billion valuation ahead of IPO

The AI chip market has another contender for billion-dollar contracts: South Korean startup Rebellions closed a pre-IPO round at $400 million with a company valuation of $2.3 billion. The company plans to go public in 2026 and is vying for a share of the market that Nvidia currently controls almost monopolistically.

Rebellions was founded in 2020 by a team of engineers with experience from Samsung and major American tech companies. Unlike many competitors, the startup focused not on neural network training but on inference—that is, running already-trained models.

This is a pragmatic choice: inference actually constitutes the majority of operational expenses for companies deploying AI in production. Every query to a language model, every AI recommendation in an app—it's all inference, and its volume is growing exponentially.

The company's flagship chip is designed specifically for transformer architecture tasks—the foundation of GPT, Claude, Gemini, and other large language models.

According to the company, the solution demonstrates high energy efficiency when handling LLM queries, which is critical for data centers with limited power supply.

Among the clients are major South Korean banks and telecommunications companies interested in a local alternative to American hardware.

The pre-IPO round of $400 million is one of the largest in Korean tech sector history.

The full list of investors is not disclosed, but it is known that both Korean institutional funds and international venture players showed interest in the company.

The $2.3 billion valuation places Rebellions alongside other AI accelerator contenders like Groq, Cerebras, and Tenstorrent.

Context matters: the AI chip market is experiencing a boom in 2025–2026, accompanied by shortages and geopolitical constraints.

The US has introduced export barriers on high-performance chip shipments to China and several other markets—creating a window of opportunity for alternative manufacturers.

South Korea, with its strong semiconductor ecosystem built around Samsung and SK Hynix, is in an advantageous strategic position.

Nevertheless, the path to real competition with Nvidia remains steep.

Nvidia sells not just chips, but the CUDA ecosystem that AI developers have relied on for years.

Switching to alternative hardware requires rewriting code and accepting compatibility risks.

It's this software superiority that protects the company's position more reliably than any hardware specifications.

Rebellions is betting on open standards and integration with popular frameworks—PyTorch, JAX, vLLM—to lower the barrier to adoption.

The 2026 IPO will be a crucial test: a successful listing will confirm that investors are willing to pay for diversification of AI hardware supply chains.

And for the industry as a whole, it's a signal—the battle for the inference market is just beginning.