Nvidia CEO Jensen Huang has named a new constraint on artificial intelligence: memory. In comments highlighted by The Motley Fool — in a piece syndicated to Yahoo Finance and AOL — Huang said memory is now AI's biggest bottleneck, a notable shift for a company whose fortunes have been tied to graphics processors.
The reason, according to The Motley Fool's analysis, is that Nvidia no longer just sells GPUs and AI hardware. It now builds systems with multiple working parts, including memory chips packed into that hardware. That makes Nvidia dependent on outside memory suppliers in a way it wasn't when a GPU was the whole product.
Think of it this way: a processor can only compute as fast as data reaches it. If the memory feeding an AI chip can't keep up — or can't be bought in sufficient quantity — the expensive silicon sits idle. Raw compute stops being the limiting factor.
The squeeze is already rippling outward. Chosunbiz reports that smartphone system-on-chip shipments are falling as memory prices soar, while AI chip shipments climbed 24%. In short, AI demand appears to be bidding memory away from consumer electronics.
Companies that locked in supply are treating it as an achievement worth announcing. Reuters' Sam Nussey reports that Sony posted a 40% year-over-year rise in first-quarter operating profit to roughly $2.92 billion, beating analyst estimates, and raised its full-year profit forecast by 8% to about $10.56 billion — while stating it has secured enough memory chip supply for the fiscal year.
Why it matters: if memory rather than processing power is what limits AI, then the chip industry's pecking order, and the price of everything from phones to data centers, gets rewritten around who controls memory supply.