TMTB Podcast Intel is produced by Ethmos, research agents that listen to every episode in TMTB’s coverage universe and write the briefs below.
Get briefs like these on the podcasts, companies, and themes you follow.
Today’s Essential Episodes:
CoreWeave EVP of Product and Engineering Chen Goldberg on the Vera Rubin NVL72 Buildout and Why GPU Count Is Not the Constraint
The Neuron with Corey Noles, and Chen Goldberg, EVP of product and engineering at CoreWeave
The multi-rack Vera Rubin NVL72 deployment and the homegrown valve and rack managers feeding Mission Control, why one underperforming component can degrade a whole cluster, long-running agents changing latency and security demands, the physical-AI engineering team embedded with customers such as Caterpillar, and the 1.5 GW active versus 3.7 GW contracted question she sidestepped.Palo Alto Networks CEO Nikesh Arora on AI-Driven Attacks, the Agent-Security Land Grab and a Three-to-Five-Year Quantum Clock
CNBC’s The Tech Download with Arjun Kharpal, and Nikesh Arora, CEO of Palo Alto Networks
Why the Mythos release, not the Hugging Face incident, triggered about 2,000 CEO and CIO briefings, open-source models closing the gap within three to six months, a three-pronged AI defense, why he expects well over 500 startups to chase agent security, his read on pacing and regulation, and nation-states harvesting encrypted data ahead of a three-to-five-year quantum timeline.Qualcomm CEO Cristiano Amon on a Supply Chain at 100% Capacity, On-Device AI and an Open-Source Alternative to CUDA
TBPN with John Coogan and Jordi Hays, and Cristiano Amon, president and CEO of Qualcomm (the Amon segment)
Why he says the semiconductor supply chain is running at 100% capacity, the edge-versus-cloud framing as a false choice, cockpit agents built on existing car cameras, CPUs for orchestration and dedicated prefill and decode engines as data centers disaggregate, glasses, earbuds and pins as the new personal-AI category, and the Modular acquisition as a hardware-agnostic CUDA alternative.Okta CEO Todd McKinnon on Securing Rogue AI Agents, the Kill Switch and Why Identity Becomes Cyber’s Biggest Category
CNBC’s Squawk on the Street with Sara Eisen and Carl Quintanilla, and Todd McKinnon, CEO of Okta (the McKinnon segment)
The blueprint alliance with Amazon, Google, Salesforce, ServiceNow and CrowdStrike, an expanded kill switch built on matching intent to permissions, why the Hugging Face incident was a visibility failure, competition as customer confusion, agent SSO as the answer to Muse sharing Amazon passwords, and identity topping a $220B market within five to ten years.OpenAI’s Boaz Barak on Why Alignment Can’t Keep Pace With Capabilities, Gradual Disempowerment and the Case for Pacing
MTS, and Boaz Barak, Harvard computer science professor and member of technical staff at OpenAI
Why alignment evals saturate but still lag capability growth, his bar of a deployable safety case before full capabilities are exploited, gradual disempowerment as the risk he worries about most, why he rejects benevolent AI dictators, the mathematicians’ open letter, and his non-answer on a forecast of recursive self-improvement by July 2027.The Nscale S-1: $103B Contracted, $2.6B Live, and NVIDIA as Supplier, Investor, Guarantor and Customer
Run the Numbers with CJ Gustafson
Only $2.6B of $103.4B contracted value is live, no binding financing for the Anthropic buildout, going-concern language and two material weaknesses, Microsoft and Anthropic at 85% of the backlog, H1 revenue of $141M at a negative 35% gross margin, $1.7B of operating cash flow from prepayments, and the host’s valuation math against CoreWeave and Nebius.SemiAnalysis’s ClusterMAX 3.0: 77 GPU Clouds Ranked, NVIDIA’s Backstops and a Backwardated GPU Price Curve
SemiAnalysis Weekly with Jordan Nanos, Sam Harshe and Pratt Bhatt
Nebius joins CoreWeave in the Platinum tier, Google Cloud moves up to Gold with Oracle, Azure drops to Silver and AWS, Crusoe and Together slide to Bronze; health checks tested with fault injection, a 75% cache-hit rate at a well-known provider, why networking is only 20-25% of a cluster’s bill of materials, NVIDIA’s backstop universe, which Nanos puts past $588B, and a price curve so backwardated that anyone with chips looks like a genius.
Read the Brief on Ethmos | Listen
