Tae Kim Sounds Off π£οΈπ£οΈπ£οΈ, Big Companies Are Hiring Again, HBO Goes Short Form
Speakers
Tae Kim
β Senior writer, Barron's; author of The Nvidia Way
John Coogan
β Co-Host, TBPN
Ben Zweig
β Labor economist, Founder & CEO, Revelio Labs
Aakash Thumaty
β Founder & CEO, Takeoff
Summary
The show covers big-company hiring trends, the Anthropic open-model policy fight, Nvidia's reported $50B Texas data center tenant deal, and a long interview with Tae Kim arguing AI semiconductor demand remains strong. Guests Ben Zweig and Aakash Thumaty discuss AI's labor-market effects and revenue-generating AI agents, respectively. The main market takeaway is that AI infrastructure and semiconductor demand are still expanding despite negative sentiment and sensational headlines.
- WSJ reports large companies are hiring again alongside AI; jobless claims remain low.
- Nvidia is revealed as a tenant for a $50B Texas data center; Tae Kim says the real lease is much smaller and misread.
- Anthropic responds to the open-weight letter with proposals on chip controls, distillation, and safety testing.
- Tae Kim argues AI and semiconductor sentiment is overly negative, with agentic AI, RSI, and enterprise adoption driving demand.
- Tae Kim flags HBM memory shortages and Nvidia supply-chain/CUDA advantages.
- Revelio Labs data shows AI adoption correlates with growth while freelancing is being disrupted.
- Takeoff and Sierra launch Horizon for revenue-generating AI agents.
- Zuckerberg publishes an AI-for-everyone op-ed; Apple announces its Upgrade leasing program.
Ideas
AI compute demand remains exceptionally strong.
The Nvidia $50bn Texas data center tenant headline is sensationalized; the real commitment is about $20bn over 15 years and $50bn over 30 years only if renewal options are exercised, a rounding error against Nvidia's run-rate. Nvidia's actual edge is CUDA reliability, co-designed systems, and its balance-sheet ability to lock up supply of TSMC wafers, HBM memory, and optical components, which should let it add revenue faster than consensus and make its infrastructure stakes look prescient.
AI compute demand remains exceptionally strong.
Sentiment on AI and semiconductor stocks is overly negative because geopolitical headlines, out-of-context AI quotes, and vendor-financing fears are obscuring strong underlying demand for compute, memory, and data-center infrastructure. Agentic AI adoption is accelerating, recursive self-improvement may arrive within 6-12 months, AMD's CEO just raised the 2030 AI CPU TAM from $120bn to $220bn, and HBM memory is in shortage. This supports the broad AI semiconductor/compute trade.
Meta AI spending fears are overblown.
Meta AI capex fears are overdone. The Reuters Zuckerberg quote about agentic AI progressing slower was taken out of context; reports after it said Meta is actually raising capex dramatically, Meta is already getting ROI from AI on ads, and its recent open model was better than expected. The market's fear of a Meta pullback in AI spending is false.
Cloud growth justifies AI capex.
AI capex fears ignore that hyperscale cloud revenue is growing very fast β Azure about 40%, Google Cloud about 80%, and Amazon high double digits β and inference GPU cloud margins are estimated at 60-80%. If growth continues, today's data center capex generates large free cash flow in 12-18 months, making Microsoft, Alphabet, and Amazon attractive as cloud beneficiaries.
