A look at Huawei and DeepSeek's mutual reliance, where Huawei provides compute hardware while DeepSeek contributes model workloads and kernel optimization
THE SO WHAT
Vertical co-design between model shops and domestic silicon — Huawei’s Ascend plus DeepSeek’s kernels and workloads — is how China works around export controls and narrows the performance gap. If you’re betting on US-centric GPU roadmaps, assume you’re competing against increasingly optimized, region-specific stacks rather than generic “inferior” alternatives.
READ THE SOURCE
MORE FROM THE WIRE
Applied AIThe biggest AI security risk is doing nothing
The real exposure isn’t just bad AI deployment—it’s letting unmanaged AI usage proliferate in the shadows. If you haven’t set basic policies, logging, and guardrails, assume your org already has AI in critical workflows and you’re blind to how it’s changing your risk surface.
Applied AIAI Spending Isn’t Enough for Investors
Markets are done rewarding AI capex without a revenue story. If your earnings calls talk about “AI investment” more than concrete uplift in ARPU, margin, or new product lines, expect pressure to cut or redirect spend in 2026–2027.
Applied AITesla-SpaceX Merger Is Elon Musk’s Biggest AI Bet Yet
Consolidating Tesla and SpaceX around an AI infrastructure thesis is a bet that owning energy, compute, and connectivity in one stack is a durable advantage. If you’re building AI-heavy products, map which of those layers you truly need to own versus rent before this integration logic spreads.
Applied AIAnthropic previews Claude Duet, an AI agent that works in the background
Background agents like Claude Duet move AI from “chat when asked” to “act while you’re away.” Before you deploy anything similar, decide which workflows you’re comfortable letting software touch without a human in the loop and how you’ll audit what it did.