NVIDIA Raises Prices Amid Memory Crunch

NVIDIA is flexing its pricing power, hiking the RTX Pro 6000 Blackwell GPU to $13,250—a 55% jump in just a year—while a global “memory famine” squeezes supply and forces older GPUs back into rotation, especially in Asia. For buyers, this means ballooning costs as memory now claims about a quarter of AI hardware budgets, leaving large studios and enterprises to swallow the price hikes while smaller businesses scramble for previous-generation deals or cloud rentals. If memory shortages don’t ease by the end of the year, expect these hefty premiums and supply distortions to stick around through 2026.

But there’s more at play than supply and demand. NVIDIA is pulling more of the AI stack in-house, launching NVLink Fusion, rolling out Dynamo 1.0 for smarter inferencing, and splitting its Data Center business to highlight architectural dominance—not just chips, but the whole blueprint. This strategy is paying off: networking revenue soared 199% year-over-year to $14.8 billion last quarter, and hyperscalers like Microsoft and Meta are backing neocloud players with massive power commitments. The catch? If neoclouds can’t ramp up power and utilization fast enough, NVIDIA risks getting caught in a circular financing trap—shipping hardware to debt-laden partners betting on future demand.

Featuring reporting from Tom’s Hardware, AlphaStreet, IO Fund, and Electronics For You.

Powered by Apisod.com