llms.blog
AI

WhiteFiber Proposes 50M Convertible Debt Offering to Expand AI Data Center Capacity

AI infrastructure provider WhiteFiber announced a proposed private placement of $250 million in convertible senior notes due 2032, with an option for initial purchasers to acquire up to an additional $37.5 million in notes. The proceeds are designated to fund data center campus acquisitions, facility buildouts, utility interconnection agreements, and hardware procurement for the company's AI cloud business. The financing coincides with WhiteFiber's agreement to acquire two industrial sites in Y

axc
axc
2 min
WhiteFiber Proposes 50M Convertible Debt Offering to Expand AI Data Center Capacity

AI infrastructure provider WhiteFiber announced a proposed private placement of $250 million in convertible senior notes due 2032, with an option for initial purchasers to acquire up to an additional $37.5 million in notes. The proceeds are designated to fund data center campus acquisitions, facility buildouts, utility interconnection agreements, and hardware procurement for the company's AI cloud business.

The financing coincides with WhiteFiber's agreement to acquire two industrial sites in Yadkin County, North Carolina, for $60 million in cash. Slated to close in the fourth quarter of 2026, the properties will form the NC-2 and NC-3 campuses with an initial 60 MW of gross utility capacity and potential to scale to 200 MW. Ready-for-service status is targeted for the third quarter of 2027.

Capital Restructuring and Deployment Pipeline

The transaction includes a concurrent debt restructuring under which WhiteFiber will exchange a portion of its existing 4.500% convertible senior notes due 2031 for a combination of cash and ordinary shares.

WhiteFiber AI Data Center Architecture

WhiteFiber reported $28.8 million in revenue for the second quarter of 2026, up 54% year over year, with colocation remaining performance obligations (RPO) reaching $932.9 million. The company has secured several long-term commitments across its footprint:

  • NC-1 Campus (Madison, NC): Initiated full contracted billing across 40 MW of IT load in August 2026, with an engineering ceiling of approximately 300 gross MW.
  • European Expansion: Executed a five-year deployment contract in the Paris region exceeding $160 million in aggregate contract value, targeted for service by September 30, 2026.
  • Compute Supply: Contracted with Prime Intellect for a three-year, $108 million deployment of 576 Nvidia Vera Rubin GPUs in Canada.
  • Colocation Partnership: Signed an exclusive operating agreement with Krambu for 100 MW of liquid-cooled GPU infrastructure beginning in 2027.

The notes offering remains subject to closing conditions and the mutual execution of the 2031 note exchange agreements.

Sources

AIInfrastructureData CentersHardwareFunding
ShareX (Twitter)LinkedIn

Written by

axc

More to read

  • Grokking in Large Language Models: How Weight Decay and Circuit Efficiency Drive Delayed Generalization

    Grokking in Large Language Models: How Weight Decay and Circuit Efficiency Drive Delayed Generalization In standard machine learning paradigms, model generalization closely tracks training loss: as an optimizer minimizes loss on training data, performance on held-out validation data improves in tandem until the model begins to overfit. In 2022, researchers at OpenAI observed a phenomenon that inverted this assumption: small neural networks trained on algorithmic tasks achieved near-zero trainin

    1 min

  • Serverless GPU Inference in Production: Cold Starts, GPU Memory Snapshotting, and Weight Paging Architectures

    Serverless GPU Inference in Production: Cold Starts, GPU Memory Snapshotting, and Weight Paging Architectures Deploying large language models on dedicated cloud GPUs creates an uncomfortable financial trade-off: keeping enterprise accelerators such as NVIDIA H100s or A100s warm 24/7 costs thousands of dollars per instance each month, yet scaling instances to zero introduces severe latency penalties. When traffic arrives at a dormant node, a standard inference server cold start can take anywhere

    1 min

  • Transformer Feed-Forward Networks as Key-Value Memories: How First-Layer Keys and Second-Layer Values Store Knowledge

    Transformer Feed-Forward Networks as Key-Value Memories: How First-Layer Keys and Second-Layer Values Store Knowledge In modern autoregressive Transformers, the division of labor between attention heads and multi-layer perceptron (MLP) blocks is often summarized through a clean functional split: attention routes information across sequence positions, while feed-forward networks (FFNs) process information per position. Yet for years, the exact mechanism by which FFNs process that information rem

    1 min