Rebellions NPUs

REBEL

Shaping the Future of Gen AI

The emergence of LLMs presents a unique opportunity for our team to make a lasting impact. In the evolving landscape of Generative AI, our next-generation chip, REBEL, stands as a game-changer.
Designed with chiplet architecture, exceptional latency, and remarkable energy efficiency, REBEL is a superlative choice for the Gen AI market.
Enhanced further by HBM3E technology support, this chip sets new standards in AI processing power.
Without a doubt, REBEL stands as the epitome of best-in-class AI chips, catering to the distinctive needs of Generative AI.

ATOM™

Redefining Datacenter Efficiency with Accelerative AI

Following the triumphant reception of our inaugural product, we've expanded our horizons and developed a cutting-edge datacenter inference AI chip, essential for businesses and cloud service providers. Backed by impressive global benchmark(MLperf™ v3.0) results, this product addresses the critical needs of the datacenter industry. Starting from 2023, our innovative solution accelerates transformer models with unparalleled latency, paving the way for datacenter commercialization on a global scale.
Explore ATOM™
markets-ion

ION™

Elevating HFT Market Performance

At ION™, we firmly believe that the High-Frequency Trading (HFT) market can truly make a difference with just 1 microsecond of improved performance. Fueled by this conviction, our dedicated team embarked on a journey that led to the creation and launch of our groundbreaking product, ION™, specifically tailored for the finance industry. Within just one year of our team's inception, we introduced this powerful solution to the market, marking a significant milestone in our pursuit of technological excellence.
Explore ION™

Roadmap

Explosive expansion on compute and external memory

64x DNCs

REBEL-Quad
(4x chiplet)

  • 175B+ LLM acceleration
  • Samsung 4nm EUV
  • HBM3E 144GB (4.8TB/sec)
16x DNCs

CXL-PIM

  • CXL-based PNM/PIM

REBEL-Single
(1x chiplet)

  • sLLM acceleration
  • Samsung 4nm EUV
  • HBM3E 36GB (1.2TB/sec)
8x DNCs

ATOM™ Plus
(high perf)

ATOM™
(low power)

  • BERT, Vision Transformers, 7B+ generative AI
  • Samsung 5nm EUV (Mass in 4Q 2023)
  • PCle gen5 64GB/s, GDDR6 256GB/s

ION™

2021~2023 2024 2025 2026

* DNC (Dual Neural Core) : Our fundamental design granule for multiple product lines. This proprietary IP is continually maintained, updated and patented by Rebellions' DNC team.