▶ 1:56 · AI Hardware & Chip Architecture · Memory & Storage · AI Infrastructure · HBM supply crunch forces Nvidia to cut Rubin Ultra memory 75%, may drive higher GPU volumes

episode briefing

Google’s AI Shakeup, Nvidia Weighs Putting Less Memory in Rubin Chips, Meta’s New AI Coding Tools

2026-08-07 · 5 company · 11 thematic

sentiment

2 bull0 bear3 neu

speakers

rob toews

Partner at Radical Ventures, an AI-focused venture capital firm. He led the firm's investment in Discovery Loop, a startup founded by former Google AI leaders Jeff Dean, Oriol Vinyals, Sanjay Ghemawat, and Quoc Le. He evaluates nearly all 'neo lab' investment opportunities and passes on the vast majority.

phoebe liu

Covers Nvidia and semiconductor supply chain for The Information; broke the story on Nvidia testing Rubin Ultra with reduced memory.

max weinbach

Technology analyst who had early access to Meta's Muse Code and Muse Spark 1.2; provides hands-on comparison of AI coding agents including Claude Code, Codex, and Cursor.

martin peers

Writes the nightly briefing for The Information; analyzed Google's AI leadership shakeup as formalizing existing operational reality rather than a disruptive change.

episode shorts · 1

Strongest AI Founders Ever?

now playing · AI Hardware & Chip Architecture · Memory & Storage · AI Infrastructure

AI Hardware & Chip Architecturemixedscore 7/10phoebe liu

Nvidia cuts Rubin Ultra memory to 256GB from 1TB as HBM supply crunch forces redesign

Nvidia is testing Rubin Ultra GPUs with 256GB or 192GB of HBM instead of the planned 1TB due to severe supply constraints; the company is doing everything possible including a $500B partner…

Memory & Storagemixedscore 8/10phoebe liu

HBM supply crunch forces Nvidia to cut Rubin Ultra memory 75%, may drive higher GPU volumes

Advanced HBM supply cannot meet Nvidia's demand even with $500B SK Hynix partnership, pushing Nvidia to test Rubin Ultra with 256GB/192GB vs 1TB planned — potentially increasing GPU unit sa…

AI Infrastructuretailwindscore 8/10phoebe liu

Nvidia co-designs next-gen HBM with SK Hynix via $500B partnership; Google's full-stack TPU/cloud/data advantage durable

Nvidia's expanded SK Group partnership includes co-designing HBM4 with SK Hynix engineers to secure supply, while Google's vertical integration (TPUs, proprietary data, massive compute, sea…

Memory & Storagetailwindscore 7/10phoebe liu

HBM bottleneck drives Nvidia-SK Hynix $500B co-design partnership to secure future supply

Nvidia's expanded partnership with SK Group (parent of SK Hynix) involves massive capital investment to increase HBM production capacity and joint engineering of next-generation memory, hig…

AI Coding Agentstailwindscore 8/10max weinbach

Coding agents shift from assistants to autonomous multi-hour workers; Meta enters with data-sharing pricing model

The market has moved from single-file assistants to agents capable of sweeping multi-file changes and autonomous operation for hours; Meta's Muse Code lags frontier models technically but i…

AI Economics & Business Modelstailwindscore 7/10max weinbach

Model providers adopt data-for-discount pricing: Meta 95% off, OpenAI/Google free tiers for training data

A new pricing paradigm is emerging where model providers (Meta, OpenAI, Google) offer steep discounts or free tokens in exchange for user training data, creating a two-tier market: subsidiz…

AI Coding Agentsmixedscore 6/10max weinbach

Meta's Muse Code launches but lags frontier coding agents by 1-2 generations; pricing innovates with 95% data-sharing discount

Max Weinbach finds Meta's new Muse Code coding agent capable but a generation or two behind Claude Code and Codex — it requires heavy hand-holding, struggles with real data vs mock data, an…

AI Infrastructuretailwindscore 7/10martin peers

Google's full-stack AI moat (TPUs, data, compute, cloud, search cash flow) insulates it from leadership turnover

Martin Peers argues Google's structural advantages — proprietary TPU accelerator fleet, largest training data corpus, massive compute, integrated cloud, and the 'most successful cash machin…

Neo Labs / AI for Sciencetailwindscore 9/10rob toews

Discovery Loop targets recursive AI self-improvement for science; Radical Ventures says 'infinite' market justifies billion-dollar seeds for elite teams

Recursively self-improving AI applied to scientific discovery (climate, health, semiconductors) represents a vastly larger and more durable opportunity than enterprise AI agents; Radical Ve…

Venture Capitalmixedscore 8/10rob toews

Neo lab investing: most fail risk/reward test at $1B+ seed, but exceptional teams pursuing infinite markets can pay off like Anthropic

Radical Ventures' framework: neo labs typically have unclear business models, product roadmaps, and technical direction at billion-dollar entry valuations, making risk/reward unfavorable; h…

Frontier AI Modelstailwindscore 8/10rob toews

Discovery Loop pursues recursively self-improving AI for scientific discovery; neo lab model proves out with Anthropic precedent

Rob Toews argues the neo lab category — elite research teams raising billion-dollar seed rounds pre-product — has a power-law payoff profile: most will fail but the next Anthropic (now tril…

←