Texas Instruments launches chips to boost AI data centres - IT Brief UK

TI's move is not about competing with NVIDIA for compute—it's about owning the "vascular system" of AI data centres. When per-GPU power crosses 700W and rack density pushes toward 100kW, the binding constraint shifts from FLOPS to power delivery and signal integrity. TI's analog and embedded DNA sits precisely at that fault line. The supply-chain ripple is immediate: upstream SiC/GaN power devices face repricing, while downstream server OEMs see analog content per BOM climb materially. Yet compliance risk is non-trivial. TI's fabs in Hefei and Wuxi, though on mature nodes, could get swept into export-control expansions if products are reclassified as "AI infrastructure components"—a structural overhang on revenue mix. Competitively, Analog Devices and Infineon are the most direct threats. TI's moat is vertical integration across analog and embedded, but if NVIDIA integrates power-management IP into next-gen architectures, the mid-layer value gets squeezed. Over the next 18 months, the AI data-centre battleground pivots from "who computes faster" to "who consumes less." TI is betting on that paradigm shift—and the bet is structurally sound, provided the geopolitical perimeter holds.