Australian Bitcoin miner-turned-AI infrastructure company IREN (formerly Iris Energy) has secured a $3.65 billion A-rated debt financing package to fund its Microsoft AI data center contract. The deal covers approximately 96% of the GPU spending required to fulfill the Microsoft agreement.

The financing marks one of the largest single rounds in the AI infrastructure sector and signals Wall Street’s growing confidence in Bitcoin mining companies’ pivot to AI compute. IREN’s existing power infrastructure and data center expertise positions it as a key player in the rapidly expanding AI data center market.

This move represents a broader industry shift as traditional miners leverage their cheap power contracts and large-scale operations to serve booming AI industry demand.

April 28, 2026 — Nvidia (NVIDIA) announced today a new memory solution aimed at alleviating the GPU VRAM shortage that has plagued the AI industry for over a year. The move is being hailed as a critical step toward resolving what some have called the “RAM Apocalypse.”

Background of the Memory Crisis

Over the past year, as large language models have grown in scale, the demand for GPU VRAM in AI training and inference has surged exponentially. From tens of billions to trillions of parameters, each generation of models has imposed higher memory requirements. At the same time, global production capacity for High Bandwidth Memory (HBM) has fallen severely short of demand, with major suppliers Samsung, SK Hynix, and Micron unable to keep pace.

Gizmodo reported that VRAM shortages have become one of the biggest bottlenecks constraining AI development, affecting everyone from tech giants to startups. Some AI companies have been forced to scale down their models or delay product launches, while GPU prices have continued to climb due to memory supply constraints.

Nvidia’s Response

Nvidia’s announced solution spans multiple dimensions:

1. Software-level optimization: Improved memory management algorithms and model compression techniques enable existing GPUs to utilize VRAM more efficiently. New driver and CUDA toolkit updates include several VRAM optimization features.

2. Hardware-level adjustments: Next-generation GPU architectures will employ more advanced memory packaging technology, increasing per-card VRAM capacity. Reports indicate new GPUs will support higher-density HBM4 memory chips.

3. Supply chain collaboration: Nvidia is deepening partnerships with HBM suppliers like SK Hynix and Samsung to expand production capacity and optimize supply chains, ensuring stable future memory supply.

Industry Impact

Tech media outlets such as Neowin note that Nvidia’s 596.36 driver already includes support for 12GB RTX 5070 laptop GPUs, signaling the company is adjusting its product lineup to ease consumer-market memory pressure.

Industry analysts say resolving the VRAM issue is crucial for the AI industry’s continued growth. “Without sufficient VRAM, even the most powerful GPUs can’t reach their full potential,” said one AI engineer who requested anonymity. “Nvidia’s move is heading in the right direction, but fully solving the problem will take time and coordinated effort across the entire supply chain.”

Market Reaction

Following the announcement, Nvidia’s stock rose slightly in after-hours trading. Investors broadly believe that easing the memory bottleneck will help Nvidia’s GPU shipments scale further, driving AI-related revenue growth.

However, some analysts cautioned that HBM capacity expansion is not an overnight process, and short-term memory supply constraints may persist.

Sources: Gizmodo, Neowin