TMTB: JPM TMT Conf Key Quotes (INTC, ALAB, ANET, DDOG)

Key Takeaways

• Agentic AI drives CPU-to-GPU ratio shift toward 1:1
• Intel sees 7% monthly yield gains on 18A node
• 14A risk production slated for 2028, volume 2029
• Advanced EMIB‑T packaging secured billions in pre‑pay orders
• Intel roadmap targets 10A and 7A beyond 2029

Pulse Analysis

Intel’s recent remarks signal a broader industry re‑evaluation of compute balance for artificial intelligence. While GPUs have long powered model training, the rise of agentic AI—software that orchestrates multiple specialized models—has revived demand for high‑performance, single‑threaded CPUs. Customers now report CPU‑to‑GPU ratios ranging from 1:1 to 4:1 for inference, prompting Intel to accelerate its CPU roadmap and emphasize silicon that can handle diverse, latency‑sensitive workloads.

On the manufacturing front, Intel’s 18A node is delivering a striking 7% month‑over‑month yield improvement, a metric that narrows the gap with rivals and improves cost competitiveness. The company’s 14A node is projected to enter risk production in 2028, with volume shipments slated for 2029, mirroring TSMC’s A14 schedule. By aligning its timeline, Intel aims to attract customers seeking a domestic alternative to Taiwan’s foundries, reducing supply‑chain risk while offering comparable performance.

Beyond chips, Intel’s advanced packaging strategy, exemplified by EMIB‑T, has secured billions of dollars in pre‑paid commitments. This technology enables heterogeneous integration of CPUs, GPUs, and accelerators, delivering higher bandwidth and lower latency than traditional interposers. The strong customer uptake reflects confidence in Intel’s ability to deliver end‑to‑end solutions for next‑generation AI workloads, reinforcing its position as a full‑stack provider in a market increasingly focused on speed, efficiency, and supply‑chain resilience.