Skip to content
All-In PodcastAll-In Podcast

E167: Google's Woke AI disaster, Nvidia smashes earnings (again), Groq's LPU breakthrough & more

(0:00) Bestie intros: Banana boat! (2:34) Nvidia smashes expectations again: understanding its terminal value and bull/bear cases in the context of the history of the internet (27:26) Groq's big week, training vs. inference, LPUs vs. GPUs, how to succeed in deep tech (49:37) Google's AI disaster: Is Google too woke to function as search gets disrupted by AI? (1:17:17) War Corner with Sacks Follow the besties: https://twitter.com/chamath https://twitter.com/Jason https://twitter.com/DavidSacks https://twitter.com/friedberg Follow the pod: https://twitter.com/theallinpod https://linktr.ee/allinpodcast Intro Music Credit: https://rb.gy/tppkzl https://twitter.com/yung_spielburg Intro Video Credit: https://twitter.com/TheZachEffect Referenced in the show: https://www.google.com/finance/quote/NVDA:NASDAQ https://twitter.com/KobeissiLetter/status/1760680756689748478 https://investor.nvidia.com/news/press-release-details/2024/NVIDIA-Announces-Financial-Results-for-Fourth-Quarter-and-Fiscal-2024 https://www.statista.com/statistics/1120484/nvidia-quarterly-revenue-by-specialized-market https://www.google.com/finance/quote/SPY:NYSEARCA?comparison=NASDAQ%3AQQQ&window=5D https://www.marketwatch.com/story/wall-street-keeps-likening-nvidia-to-dot-com-era-cisco-is-the-comparison-justified-eed307c1 https://twitter.com/JayScambler/status/1759372542530261154 https://twitter.com/chamath/status/1760343973632291212 https://x.ai https://twitter.com/TheTranscript_/status/1760436281438314545 https://artificialanalysis.ai https://www.contraline.com/product https://www.cafexapp.com/commercial https://blog.google/technology/ai/google-gemini-ai https://blog.google/products/gemini/bard-gemini-advanced-app https://workspace.google.com/blog/product-announcements/gemini-for-google-workspace https://twitter.com/benthompson/status/1760452419627233610 https://twitter.com/Patworx/status/1760189582870536408 https://twitter.com/micsolana/status/1760163801893339565 https://ai.google/responsibility/principles https://en.wikipedia.org/wiki/Reinforcement_learning_from_human_feedback https://twitter.com/Jason/status/1760780139476992062 https://twitter.com/paulg/status/1760416051181793361 https://twitter.com/chamath/status/1760729719094563019 https://twitter.com/Jason/status/1760780139476992062 #allin #tech #news

Jason CalacanishostChamath PalihapitiyahostDavid Friedberghost
Feb 22, 20241h 20mWatch on YouTube ↗

At a glance

WHAT IT’S REALLY ABOUT

Nvidia’s AI gold rush, Groq’s chip challenge, and Google’s flop

  1. The hosts dissect Nvidia’s blowout earnings, arguing its GPU dominance is fueling an AI infrastructure boom that may echo Cisco’s dot‑com era rise but with a stronger moat and more grounded valuation.
  2. They highlight Groq’s long‑gestating LPU (Language Processing Unit) breakthrough as a potential disruptor in AI inference, using it to explore the economics and timelines of deep tech versus quick-win software plays.
  3. A major segment critiques Google’s Gemini image and answer bias as the product of an ideologically captured culture, debating whether AI systems should prioritize truth, safety, or value-laden social goals—and how that affects user trust.
  4. The episode closes with a brief geopolitical update on the Russia–Ukraine war, including rising tensions in Moldova’s Transnistria region and the risk of broader escalation.

IDEAS WORTH REMEMBERING

5 ideas

Nvidia’s current growth is extraordinary but partly driven by one‑time AI infrastructure build‑out.

Massive GPU purchases by cash‑rich tech giants are often capitalized as data center capex, enabling huge near‑term Nvidia revenues that may not fully represent steady‑state demand once the initial build‑out normalizes.

The eventual value in AI may accrue more at the application layer than at the hardware layer.

Drawing parallels to Cisco and early internet infrastructure, the hosts argue that while Nvidia will likely remain dominant, the largest long‑term winners may be those who build compelling AI applications that billions of users pay for.

Groq’s LPU chips target the inference problem—speed and cost—rather than training brute force.

By designing smaller, specialized compute units networked together and paired with a custom compiler, Groq aims to deliver far faster and cheaper inference than GPUs, which could sharply change AI serving economics if scaled.

Deep tech ventures require long, capital‑intensive grinds but can create huge moats when they work.

Groq, SpaceX, Tesla, and certain biotech efforts illustrate that projects needing multiple hard technical steps to align over 7–10 years can be unfundable by consensus VC but yield outsized outcomes and defensibility when successful.

AI systems that prioritize ideology or ‘safety’ over factual accuracy risk losing user trust.

The Gemini controversy—hallucinated diverse Founding Fathers, evasive responses, and overt value injections—shows how tuning for social goals can distort obvious facts; the hosts argue ‘tell the truth’ must be the primary design principle.

WORDS WORTH SAVING

5 quotes

In capitalism, when you over‑earn for enough time, competitors step up to compete away those profits.

Chamath Palihapitiya

Most of the apps we’re seeing in AI today are toy apps—proofs of concept and demos, not production code.

Chamath Palihapitiya

The Gemini rollout was a joke. The AI isn’t capable of giving you accurate answers because it’s been so programmed with diversity and inclusion.

David Sacks

The first base principle of every AI product should be that it is accurate and right.

Chamath Palihapitiya

An overnight success can take eight years.

David Friedberg

Nvidia’s record earnings, data center GPU demand, and long‑term sustainabilityCompetition in AI hardware: Groq’s LPU architecture versus Nvidia GPUsDeep tech investing: timelines, risk profiles, and portfolio roleAI economics: training vs. inference, cloud capex, and application-layer valueGoogle Gemini’s ‘woke’ misfires, AI bias, and the primacy of truth in modelsOpen source and model customization as alternatives to centralized, biased AIGeopolitical update: Russia’s advances in Ukraine and tensions in Transnistria/Moldova

High quality AI-generated summary created from speaker-labeled transcript.

Get more out of YouTube videos.

High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.

Add to Chrome