TickerTain — Make Analysis Fun Again

▶ 28:00 · AI Infrastructure · Huang: CUDA ecosystem moat = install base + programmability + cloud ubiquity + TCO leadership

episode shorts · 8

Jensen Huang on Why Nvidia Passed on Anthropic the First Time

Jensen Huang on Nvidia's Competition

How Nvidia Actually Allocates GPUs - Jensen Huang

Why Nvidia Invests Billions in Companies That May Fail - Jensen…

The Idea That China Can't Have AI Chips Is Nonsense - Jensen Hu…

AI Doomers Were Wrong About Radiology - Jensen Huang

Jensen Huang Makes the Case for Selling Chips to China

Jensen Huang Fires Back on China Chip Ban

now playing · AI Infrastructure

AI Applicationstailwindscore 7/10jensen huang

Huang: AI agents will cause exponential growth in EDA and software tool usage (Synopsys, Cadence)

Agents remove the human-engineer bottleneck on tool usage; design space exploration will explode, driving skyrocketing license revenue for Synopsys Design Compiler, Cadence tools, and Nvidi…

Semiconductorstailwindscore 8/10jensen huang

Huang: Nvidia orchestrates supply chain via commitments, ecosystem alignment, and technology invention (COUPE, photonics)

Nvidia doesn't just buy capacity; it shapes upstream supply by sharing roadmaps with CEOs (Micron, TSMC, Lumentum, Coherent), co-inventing technologies (COUPE, double-sided probing), and li…

AI Infrastructuretailwindscore 9/10jensen huang

Huang: All AI compute bottlenecks (EUV, CoWoS, HBM, energy) solvable in 2-3 years with demand signal

No bottleneck is fundamental; each resolves in 2-3 years once a credible demand signal exists. Nvidia's purchase commitments and roadmap visibility provide that signal, enabling TSMC, ASML,…

Energy & Power Generationriskscore 8/10jensen huang

Huang: Energy is the longest-pole bottleneck for AI factories; chip/packaging bottlenecks are 2-3 years

While logic, CoWoS, HBM, and EUV scale in 2-3 years, energy infrastructure (power plants, grid, permitting) takes far longer — making energy policy the critical constraint for US AI leaders…

Semiconductorstailwindscore 9/10jensen huang

Huang: Architecture and computer science (not just 50x Blackwell leap) matter more than lithography scaling

Moore's Law contributes only ~25%/year; 10x-100x leaps come from algorithm-system co-design (MoE, disaggregation, new numerics, NVLink, Spectrum-X) — Nvidia's extreme co-design capability i…

AI Infrastructuretailwindscore 10/10jensen huang

Huang: CUDA ecosystem moat = install base + programmability + cloud ubiquity + TCO leadership

Three reinforcing pillars: (1) hundreds of millions of GPUs everywhere, (2) programmable architecture enabling rapid algorithm invention, (3) best performance-per-TCO proven by InferenceMAX…

Semiconductorstailwindscore 9/10jensen huang

Huang: Annual architecture cadence (Blackwell → Vera Rubin → Vera Rubin Ultra → Feynman) creates investable predictability

Predictable yearly product generations with 30-50x performance-per-watt leaps let customers bet entire businesses on Nvidia roadmap; no other semiconductor company offers comparable cadence…

Geopolitics & Trademixedscore 8/10jensen huang

Huang: Conceding China market to Huawei accelerates their ecosystem; US should compete and win all five AI stack layers

Export controls forced China to build domestic stack (Huawei, SMIC, open models); with 50% of AI researchers, abundant energy, and manufacturing scale, China will advance regardless. US lea…

AI Infrastructuretailwindscore 7/10jensen huang

Huang: Inference market segmenting into high-throughput vs ultra-low-latency premium tokens; Groq integrated for latter

As token value rises, inference economics bifurcate: high-throughput for batch, ultra-low-latency (high ASP) for interactive coding/agents. Nvidia expands Pareto frontier by folding Groq in…

←

▶ 28:00 · AI Infrastructure · Huang: CUDA ecosystem moat = install base + programmability + cloud ubiquity + TCO leadership

episode shorts · 8

Jensen Huang on Why Nvidia Passed on Anthropic the First Time

Jensen Huang on Nvidia's Competition

How Nvidia Actually Allocates GPUs - Jensen Huang

Why Nvidia Invests Billions in Companies That May Fail - Jensen…

The Idea That China Can't Have AI Chips Is Nonsense - Jensen Hu…

AI Doomers Were Wrong About Radiology - Jensen Huang

Jensen Huang Makes the Case for Selling Chips to China

Jensen Huang Fires Back on China Chip Ban

now playing · AI Infrastructure

AI Applicationstailwindscore 7/10jensen huang

Huang: AI agents will cause exponential growth in EDA and software tool usage (Synopsys, Cadence)

Agents remove the human-engineer bottleneck on tool usage; design space exploration will explode, driving skyrocketing license revenue for Synopsys Design Compiler, Cadence tools, and Nvidi…

Semiconductorstailwindscore 8/10jensen huang

Huang: Nvidia orchestrates supply chain via commitments, ecosystem alignment, and technology invention (COUPE, photonics)

Nvidia doesn't just buy capacity; it shapes upstream supply by sharing roadmaps with CEOs (Micron, TSMC, Lumentum, Coherent), co-inventing technologies (COUPE, double-sided probing), and li…

AI Infrastructuretailwindscore 9/10jensen huang

Huang: All AI compute bottlenecks (EUV, CoWoS, HBM, energy) solvable in 2-3 years with demand signal

No bottleneck is fundamental; each resolves in 2-3 years once a credible demand signal exists. Nvidia's purchase commitments and roadmap visibility provide that signal, enabling TSMC, ASML,…

Energy & Power Generationriskscore 8/10jensen huang

Huang: Energy is the longest-pole bottleneck for AI factories; chip/packaging bottlenecks are 2-3 years

While logic, CoWoS, HBM, and EUV scale in 2-3 years, energy infrastructure (power plants, grid, permitting) takes far longer — making energy policy the critical constraint for US AI leaders…

Semiconductorstailwindscore 9/10jensen huang

Huang: Architecture and computer science (not just 50x Blackwell leap) matter more than lithography scaling

Moore's Law contributes only ~25%/year; 10x-100x leaps come from algorithm-system co-design (MoE, disaggregation, new numerics, NVLink, Spectrum-X) — Nvidia's extreme co-design capability i…

AI Infrastructuretailwindscore 10/10jensen huang

Huang: CUDA ecosystem moat = install base + programmability + cloud ubiquity + TCO leadership

Three reinforcing pillars: (1) hundreds of millions of GPUs everywhere, (2) programmable architecture enabling rapid algorithm invention, (3) best performance-per-TCO proven by InferenceMAX…

Semiconductorstailwindscore 9/10jensen huang

Huang: Annual architecture cadence (Blackwell → Vera Rubin → Vera Rubin Ultra → Feynman) creates investable predictability

Predictable yearly product generations with 30-50x performance-per-watt leaps let customers bet entire businesses on Nvidia roadmap; no other semiconductor company offers comparable cadence…

Geopolitics & Trademixedscore 8/10jensen huang

Huang: Conceding China market to Huawei accelerates their ecosystem; US should compete and win all five AI stack layers

Export controls forced China to build domestic stack (Huawei, SMIC, open models); with 50% of AI researchers, abundant energy, and manufacturing scale, China will advance regardless. US lea…

AI Infrastructuretailwindscore 7/10jensen huang

Huang: Inference market segmenting into high-throughput vs ultra-low-latency premium tokens; Groq integrated for latter

As token value rises, inference economics bifurcate: high-throughput for batch, ultra-low-latency (high ASP) for interactive coding/agents. Nvidia expands Pareto frontier by folding Groq in…