Nebius (NBIS) Stock Jumps as Company Agrees to Acquire Eigen AI for $643M

TLDR

• Nebius (NBIS) announced an agreement to acquire inference and model optimization company Eigen AI for approximately $643 million in cash and Class A stock.
• Eigen AI’s technology will be integrated into Nebius Token Factory, its managed inference platform for enterprise AI.
• The founding team from MIT’s HAN Lab will establish Nebius’s first Bay Area engineering and research hub.
• Jointly optimized model implementations from the two companies have already ranked among the fastest on Artificial Analysis benchmarks.
• NBIS stock rose 8.51% on the news, trading at $150.00, despite a 6.07% decline over the prior week.

Nebius (NBIS) announced on May 1, 2026 that it has agreed to acquire Eigen AI for approximately $643 million. The deal will be paid in a mix of cash and Nebius Class A stock, based on the company’s 30-day weighted average stock price at signing. NBIS jumped 8.51% on the news, hitting $150.00.

The transaction is expected to close within weeks, pending antitrust clearance and standard closing conditions.

Eigen AI is an inference and model optimization company. Its technology helps AI teams run open-source models faster and cheaper in production, without having to build optimization tools themselves.

Nebius Group N.V., NBIS

Nebius plans to fold Eigen AI’s technology directly into its Token Factory platform. Token Factory provides autoscaling endpoints and fine-tuning pipelines across major open-source models including Llama, DeepSeek, Qwen, Gemma, and others.

The two companies already have a working history. Before this deal, they co-developed optimized model implementations that ranked among the fastest on Artificial Analysis, a widely used AI benchmarking platform.

What Eigen AI Brings to the Table

Eigen AI was founded by researchers out of MIT’s HAN Lab. Co-founders Ryan Hanrui Wang and Wei-Chen Wang are behind two of the most widely cited techniques in production AI deployment.

Ryan’s Sparse Attention work (SpAtten) is the most-cited HPCA paper since 2020. Wei-Chen’s Activation-aware Weight Quantization (AWQ) received the MLSys 2024 Best Paper Award and is now the standard approach for 4-bit model serving.

Co-founder Di Jin holds a PhD from MIT CSAIL and contributed directly to Meta’s Llama 3 and Llama 4 post-training. He also co-authored the CGPO reinforcement learning from human feedback framework.

Following the close, the team will relocate to the San Francisco Bay Area and establish Nebius’s first US engineering and research presence.

The Inference Market Context

Inference is now the fastest-growing part of the AI compute market. It is forecast to make up roughly two-thirds of total AI compute demand in 2026.

Running inference efficiently is technically complex. It involves model representation, GPU kernel execution, and real-time workload scheduling — most teams don’t have the in-house expertise to handle all of it.

Open-source models, which typically ship unoptimized, add further complications. Newer architectures like Mixture-of-Experts and Compressed Sparse Attention introduce challenges around memory and compute efficiency that require specialist knowledge.

Eigen AI’s full-stack optimization approach covers post-training, fine-tuning, and production inference across all major open-source models. Its kernel and model-level techniques are designed to extract better performance from existing hardware without extra engineering work.

The post Nebius (NBIS) Stock Jumps as Company Agrees to Acquire Eigen AI for $643M appeared first on CoinCentral.