Why the US & NVIDIA Just Bailed Out Intel (and What It Means for AI) | EP #195
An emergency WTF episode recorded the day the US government took a 10% stake in Intel and Nvidia followed with its own investment, which the panel frames as existential given TSMC's 66% share of AI chip fabrication and its proximity to China. They draw a close parallel to Microsoft's 1997 rescue investment in Apple, debate where to put capital (seed-stage AI startups, data-center buildout, clean energy, uranium), and argue investing against a superintelligent market is largely futile. The back half covers AI passing superhuman thresholds on hard-to-verify coding olympiad problems, Meta's new display glasses and EMG neural wristband as an early step toward brain-computer interfaces, DeepMind's progress on the Navier-Stokes millennium problem, a Nature paper on a disease-forecasting foundation model (Delphi-2M), and a wave of humanoid-robotics funding (Figure's raise, 1X Technologies) leading into Tesla's Optimus 3 teaser and a discussion of robot personhood and recursive self-improvement.
US/Nvidia bailout of Intel & TSMC dependencyGeopolitics▶ 0:00
US government bought 10% of Intel and Nvidia followed suit, driven by fear of losing the only domestic chip fab as TSMC holds 66% of AI chip manufacturing 90 miles from China.
Panel compares the Intel-Nvidia deal structure to Bill Gates' 1997 investment in a near-bankrupt Apple, which restored credibility and cash ahead of Apple's rise to the world's most valuable company.
Panel gives (non-investment-advice) picks: seed-stage AI startups, data-center buildout (CoreWeave, Crusoe), clean energy (NextEra), and uranium/nuclear, while Alex argues the market already prices in everything.
AI achieves superhuman results at coding olympiadAI▶ 28:12
DeepMind and OpenAI models hit gold-medal-level performance on competitive coding olympiads with hard-to-verify problems, which Alex frames as the next capability leap after reasoning models.
Meta smart glasses and EMG neural wristbandOther▶ 33:39
Meta unveils in-lens display glasses controlled by an EMG wristband that reads forearm muscle signals for silent typing; panel debates whether this is a step toward brain-computer interfaces.
US-China chip war and the 'digital silk road'Geopolitics▶ 46:03
David Sacks warns that export controls risk pushing the world toward Huawei and Chinese open-source models; panel debates whether this is genuine tech competition or politically driven by tariffs.
AI progress on the Navier-Stokes millennium problemCompute▶ 54:43
Google DeepMind makes progress on the Navier-Stokes fluid dynamics Clay Millennium Prize problem, raising speculation about fluid- or plasma-based computing substrates and self-replicating nanomachines.
AI disease-forecasting model (Delphi-2M)Health▶ 58:47
A Nature paper introduces Delphi-2M, a foundation model that tokenizes medical history to forecast risk across 1,000+ diseases and extrapolate a person's health decades forward, trained on ~1.9 million people.
Figure AI raises over $1 billion in a Series C at an extraordinary valuation to fund manufacturing, alongside 1X Technologies' ongoing raise, as both race to solve the robotics supply chain.
Optimus 3 teaser and robot personhoodRobotics▶ 1:13:53
Tesla teases a sleeker Optimus 3 with an OLED face display, targeting mass production in 2026 and 1 million robots/year toward a $25 trillion market; panel discusses the coming personhood debate for humanlike AGI robots.
Recursive self-improvement and the singularityAI▶ 1:20:28
Panel debates whether recursive self-improvement in AI (and soon robotics manufacturing) constitutes the singularity, predicting a sudden, unmistakable software-driven capability jump followed by uneven global distribution.
Predictions made
openAlex Wissner-Gross: The next major reasoning-style capability jump (comparable to the o1/'strawberry' moment) will become broadly deployed and generally available.
“he expected we'd be at 90% everything being coded by AI and in 2026 we'd reach 100% of all coding being done by AI”
Your call:
openDave Blundin: The current EMG wristband interface will become obsolete before it 'creeps' closer to the brain, replaced by more elegant signal-extraction methods.
“my prediction my prediction was uh boots on Mars by 2030 except they're going to be robot boots”
Your call:
openElon Musk (clip): Optimus will move from 2025 prototypes to mass production with a goal of 1 million robots per year, addressing a $25 trillion market.
“prototypes by the end of 2025. mass production in 26 with a goal of 1 million robots per year... a $25 trillion market”
Your call:
openDave Blundin: Once software-only recursive self-improvement takes hold, an unmistakable 1000x jump in neural network performance (instant disease and math solutions) will occur within roughly a 30-day window, with no change in chips.
“you're going to see a an immediate 1000x step up in performance of neural network software and the results of that are going to be unmistakable... all going to happen within say a 30-day window”
Your call:
Numbers that matter
TSMC has 66% market share in AI-driving chipsCited as the core reason the US cannot let Intel fail.
US government bought a 10% stake in IntelTriggered Nvidia's own follow-on investment.
Intel stock up 25-30% in a single dayMarket reaction to the government/Nvidia investment news.
Microsoft's 1997 investment in Apple was roughly 10% of the companyDrawn as a near-identical historical parallel to the Intel deal.
Nvidia's market cap grew from about $0.5 trillion (spring 2023) to roughly $4-4.5 trillion, about 8xAlex's MIT lecture anecdote about predicting Nvidia's AI pivot.
Leopold Aschenbrenner's hedge fund grew from about $1 billion to $2.1 billion in one quarterPer 13f.info disclosures, from inflows and gains including his Intel options bet.
Leopold's Intel call options were up 250-255% on the day of the announcementOptions were trading under $1 before the news.
Solar became cheaper to build than fossil fuel plants in 2016; cheaper to build+operate than fossil opex alone since 2019Salim's rationale for a clean-energy investment thesis.
Figure AI raised over $1 billion in a Series C at a valuation stated as $93 billion, having raised $2 billion total over two yearsFunding round for humanoid robot maker Figure, led by Brett Adcock.
Delphi-2M disease-forecasting model trained on close to 1.9 million peopleScale of the population used to validate the disease-token approach.
Optimus 3: mass production targeted for 2026, goal of 1 million robots/year, addressing a $25 trillion marketTesla's teased specs and Elon Musk's stated goals for Optimus.
Elon Musk's pay package requires growing Tesla to $8 trillion, roughly 8-12x its value at the timeTied to Optimus/robotics as the primary value driver.
Worth digging into
🕳️ DeepMind's Navier-Stokes progress and fluid computing substrates
Alex frames a constructive Navier-Stokes solution as potentially unlocking fluid- or plasma-based computing and even self-replicating nanomachines, a wild claim worth grounding in the actual paper.
🕳️ Delphi-2M disease-forecasting foundation model
A Nature paper claims a tokenized-disease-history model can forecast risk across 1,000+ diseases decades out, trained on ~1.9 million people; worth verifying methodology and real-world accessibility.
🕳️ Leopold Aschenbrenner's Intel options trade
His fund's quarterly filings show a real-time case study of a high-conviction geopolitical trade turning into 100%+ gains overnight.
🕳️ Digital silk road: Huawei's global expansion vs. OpenAI's
David Sacks's warning frames a live land-grab where Huawei/Chinese open-source models and OpenAI/Nvidia US models are competing for adoption in India, Africa, Southeast Asia, and Europe.
🕳️ Figure AI and 1X Technologies manufacturing race
Both companies are pouring huge new capital into robot manufacturing and supply chains, with hints (not yet confirmed publicly) that they may be building robots-that-build-robots for recursive self-improvement.