Tae Kim, a technology analyst and author covering semiconductors and AI infrastructure, breaks down why Nvidia still looks strong in the AI race. The conversation hits inference demand, supply lockups, GPU and CPU shortages, open models, vertical AI agents, and why fears around depreciation, pullbacks, and a future compute glut may be overdone.
29:13
forum Ask episode
web_stories AI Snips
view_agenda Chapters
auto_awesome Transcript
info_circle Episode notes
insights INSIGHT
Why Tae Kim Thinks Nvidia Selloff Is Overblown
Tae Kim argues Nvidia’s selloff looks more like macro fear than business deterioration, echoing prior tariff and DeepSeek panics.
He says AI capex fears and the Iran oil shock are masking strong fundamentals, just as a 30% drawdown happened while business kept flying.
insights INSIGHT
Inference Demand Is Driving A Fresh Compute Shortage
Tae Kim says inference demand is exploding from AI agents and coding assistants, creating real AI compute shortages across major labs.
He cites talks with Ian Buck and engineers at Meta, Google, and Nvidia, plus users running sneaker bots for scarce B200 GPUs.
insights INSIGHT
Why Grok Fits Nvidia's Inference Strategy
Tae Kim sees Nvidia’s Grok move as a pragmatic extension of its stack, not a rejection of GPUs or a full customer-competitive pivot.
He says Grok could handle roughly 25% of low-latency inference while Vera Rubin covers 75%, matching the new coding-agent workload mix.
Get the Snipd Podcast app to discover more snips from this episode
This is our full interview with Tae Kim, recorded live on TBPN.
We discuss why he believes fears around Nvidia and AI infrastructure are overblown despite recent market pullbacks, unpack how exploding inference demand from coding agents and enterprise adoption is driving a sustained compute shortage that Nvidia is uniquely positioned to capture after locking up key supply, and debate what this next wave of AI means for everything from GPU scarcity and chip strategy to token demand, vertical agents, and whether the current boom is the early innings of a multi-year expansion or the setup for a future compute glut.
TBPN 