Babyfolio(@babyfolio):Meta scaling its custom MTIA chips into data centers is another sign hyperscalers are increasingly going custom to lower inference costs. Google has TPUs. Amazon has Trainium. Meta is now taking MTIA to multi GW scale. This is structurally bullish for $AVGO. Broadcom co-develops Meta's custom XPUs and supplies the Ethernet infrastructure connecting them. So as Meta shifts more compute toward its own silicon, Broadcom still captures a meaningful piece of that.

2026.09.16 13:12

Meta scaling its custom MTIA chips into data centers is another sign hyperscalers are increasingly going custom to lower inference costs. Google has TPUs. Amazon has Trainium. Meta is now taking MTIA to multi GW scale. This is structurally bullish for $AVGO. Broadcom co-develops Meta's custom XPUs and supplies the Ethernet infrastructure connecting them. So as Meta shifts more compute toward its own silicon, Broadcom still captures a meaningful piece of that.

Paradis@ParadisLabs

2026.09.15 14:18

The end result is that the chips can perform more efficiently than "whatever Nvidia is currently shipping" when running AI models. What did Meta cook up?!