AI Chips, Cheap Gas & a Stretched Deficit (e2550)

In this episode of the Telltales Podcast, Mike, Hunt, and Jason walk through the latest Cash Flow Memo—from US government finances and energy markets to a deep-dive on Nvidia and the AI data center buildout. They revisit the “Nvidia’s Five Risks” memo 18 months later and ask what could still derail the story for investors focused on cash flow.[00:00] Welcome & what’s in the Cash Flow MemoMike kicks off the episode with the usual Telltales introduction, highlights the Cash Flow Memo, and sets up a 30-minute tour through energy, technology, healthcare, and US government finances tied to Exhibits A, B, and C.[00:00:20] Disclaimer & investing ground rulesStandard disclosure that the conversation is for informational purposes only, not investment advice, and that listeners must do their own work before making investing decisions.[00:00:43] Exhibit C – Oil, China’s slowdown & a maturing global cycleHunt walks through Exhibit C, noting flat oil demand growth in China, the US, and Europe, arguing that China’s economy looks more “mature” and that global oil consumption growth should slow toward ~600,000 barrels per day, with implications for crude prices and upstream oil producers.[00:02:00] Exhibit B – Natural gas, Permian overproduction & negative Waha pricesExhibit B shows a very different picture for gas: cold weather, near-month prices spiking toward $5, the 2026 strip around $4+, and yet associated gas from the Permian keeps pushing production higher, leading to bizarre negative prices at the Waha hub and a tricky setup for upstream gas investors.[00:04:08] Exhibit A – US deficit, healthcare costs & the path for interest ratesTurning to Exhibit A, Hunt explains why fiscal year 2026 could look slightly better than 2025, how moderating healthcare and Medicaid growth could help stabilize the deficit, and why keeping debt-to-GDP near or below ~100% is crucial for anchoring long-term interest rates.[00:06:02] Inflation targets, deglobalization & the politics of power pricesJason questions whether the Fed’s 2% inflation goal is realistic in a deglobalizing world, while Hunt focuses on gasoline and electricity bills, arguing that regulators should “wall off” AI data centers from retail power rates and that utility regulation will become a bigger political issue.[00:08:22] Setting up Nvidia: revisiting “Hyperscale CapEx and Nvidia’s Five Risks”Hunt hands the conversation to Mike to revisit their August 2024 memo on Nvidia’s five key risks, explaining how Nvidia became the most valuable company in the world and why it’s time to compare the original risk map with what actually happened.[00:09:55] Risk #1 – CapEx digestion & token demand vs. GPU buildoutMike reviews the fear that hyperscalers would overbuild data centers and then “digest” CapEx, but notes that 18 months later Nvidia’s data center revenue is running north of $200 billion and older A100 GPUs are still fully utilized, shifting the key question to whether inference token demand can keep pace with GPU capacity.[00:12:14] Risk #2 – Depreciation mismatch & AI economics vs. accountingThe team revisits the concern that extending server depreciation lives to five or six years might overstate earnings if GPUs become obsolete faster, concluding it’s mostly a non-issue for big platforms like Google and Meta but something to watch for “neo-clouds” if their average contract lengths shorten.[00:15:28] Risk #3 – The AI scaling wall, X’s Colossus & the rise of networkingMike explains why the feared “scaling wall” hasn’t arrived, how X’s massive H100 clusters (Colossus) showed that coherent mega-clusters can leapfrog competitors, and why Nvidia’s networking stack—boosted by the Mellanox acquisition—is becoming as important as the chips themselves.[00:17:55] Risk #4 – Non-chip bottlenecks: software efficiency, power & export controlsThey tackle non-chip constraints: DeepSeek’s efficiency gains and Jevons Paradox (cheaper tokens driving more demand), severe power bottlenecks that leave Microsoft with idle chips awaiting grid connections, and geopolitical export controls that periodically restrict Nvidia’s ability to sell GPUs into China.[00:20:28] Risk #5 – Competition from Google, AMD, TPUs and custom siliconThe conversation turns to competitive threats: AMD as a perpetual second source, Google’s TPU v7 and its own coherent clusters powering Gemini, and custom ASICs for workloads large enough to justify re-architecting software, even as Nvidia remains the default platform for most AI data center compute.[00:21:51] Nvidia’s valuation, free cash flow growth & token-driven demandHunt walks through Nvidia’s latest numbers and a free cash flow multiple around 40–45x, asking whether 20% annual growth is realistic, while Jason and Mike discuss the massive potential if CPU-based data centers transition to GPU/TPU-based AI infrastructure over many years.[00:25:27] Healthcare, legal work & the real drivers of token growthJason highlights early healthcare AI use cases in medical note-taking and radiology, contrasts them with heavy-token workloads like long legal documents, and explains why reasoning models that do hundreds or thousands of inference steps per answer—and multimodal, audio/video-heavy interfaces—could supercharge token consumption.[00:28:50] Programming note & teaser: Why Google needs Broadcom (and Nvidia doesn’t)Hunt closes with a scheduling note for the Christmas and New Year’s weeks, reiterates their commitment to 52 episodes a year, and teases next week’s 10-minute segment on why Google needs Broadcom while Nvidia doesn’t.Thanks for listening to Telltales—subscribe, grab the free Cash Flow Memo at Telltales.us, and share this episode with a friend who’s trying to understand AI, energy, and cash-flow-based investing. Drop a comment with the next company or sector you want the team to put through the cash flow wringer.This podcast and the information herein are intended for informational purposes only. The views expressed herein are the author’s alone and do not constitute an offer to sell, or a recommendation to purchase, or a solicitation of an offer to buy, any security, nor a recommendation for any investment product or service. While certain information contained herein has been obtained from sources believed to be reliable, neither the author nor any of his employers or their affiliates have independently verified this information, and its accuracy and completeness cannot be guaranteed. Accordingly, no representation or warranty, express or implied, is made as to, and no reliance should be placed on, the fairness, accuracy, timeliness or completeness of this information. The author and all employers and their affiliated persons assume no liability for this information and no obligation to update the information or analysis contained herein in the future. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit telltales.substack.com