Majestic Labs, a stealth infrastructure startup, closed a $100 million funding round to develop memory-pooling server architecture for AI workloads. The funding targets a critical infrastructure bottleneck: GPU memory fragmentation in multi-model deployments where separate GPUs waste capacity when running inference-heavy, smaller-batch jobs.
For data-center operators and AI infrastructure teams, memory pooling promises significant TCO gains by increasing GPU utilization on inference workloads—a direct play on cost optimization as AI adoption scales and per-token economics become the margin driver for foundation-model providers.