TMT Breakout

TMT Breakout

TMTB Podcast Intel: OpenAI’s Sam Altman, Micron’s (MU) Scott DeBoer, Blackstone’s (BX) Jon Gray, Cognition CEO Scott Wu, Travis Kalanick on NVIDIA (NVDA), MongoDB’s (MDB) Dev Ittycheria

Ethmos's avatar
TMT Breakout's avatar
Ethmos and TMT Breakout
Nov 18, 2026
∙ Paid
TMTB Podcast Intel — Powered by Ethmos

TMTB Podcast Intel is produced by Ethmos, research agents that listen to every episode in TMTB’s coverage universe and write the briefs below.

Follow OpenAI, Sam Altman, MU, BX, NVDA or MDB on Ethmos to get a brief whenever they or their executives go on a podcast. TMTB readers get 14 days free and 25% off their first month.


Today’s Essential Episodes:

  1. Sam Altman on Jensen Huang, an Eventual OpenAI IPO, Chip-Demand Double Counting and the AI Backlash
    Future Proof from Semafor Technology, with Semafor’s Reed Albergotti and OpenAI CEO Sam Altman

  2. Micron President and CTO Scott DeBoer on the Memory Roadmap, Global Fab Expansion and AI in Chip Design
    TBPN, live from TomorrowX Day 2, with John Coogan, Jordi Hays and Micron President and CTO Scott DeBoer (the DeBoer interview)

  3. Blackstone President Jon Gray on AI’s Demand for Compute, Energy and Infrastructure
    TBPN, live from TomorrowX Day 2, with John Coogan, Jordi Hays and Blackstone President and COO Jon Gray (the Gray interview)

  4. Cognition CEO Scott Wu on Devin Bot, AI Cybersecurity and Uneven AI Progress
    TBPN, live from TomorrowX Day 2, with John Coogan, Jordi Hays and Cognition co-founder and CEO Scott Wu (the Wu interview)

  5. Travis Kalanick on Why He Backs Jensen Huang’s Return to NVIDIA and His Playbook for Fighting Back
    TBPN, with John Coogan, Jordi Hays and Uber co-founder Travis Kalanick (full interview, recorded at TomorrowX)

  6. MongoDB’s Dev Ittycheria on Returning to Run the Company Ten Months After Retiring, With HubSpot’s Brian Halligan
    Long Strange Trip: CEO to CEO, with HubSpot co-founder Brian Halligan and MongoDB’s Dev Ittycheria


Produced by Ethmos: AI-assisted summaries of public podcast episodes. For information only. May miss context or contain errors. Verify before relying on them. Summaries are not substitutes for the episodes, which belong to their creators. Not investment advice, offer, or solicitation. Not affiliated with or endorsed by the shows, hosts, guests or companies discussed. Corrections and publisher opt-outs: support@ethmos.ai.
Sam Altman on Jensen Huang, an Eventual OpenAI IPO, Chip-Demand Double Counting and the AI Backlash, on Semafor’s Future Proof

Listen · Follow OpenAI on Ethmos · Follow Sam Altman on Ethmos

THE SIGNAL

  • Altman’s double-counting warning, reportedly made privately to Samsung executives: many new chip vendors each tell the silicon supply chain the labs will buy most of their output, so the same demand gets counted repeatedly (our read).

  • Altman puts the DRAM price rise at roughly 5-7x and says memory makers held back capex because the business historically meant investing $20B to earn $19B; he says production is now ramping.

  • Altman calls iPhone photos and iMessages tech’s biggest lock-in and says Dots’ learned behavioral preferences are harder to export, and says OpenAI is still working out what export would look like, implying strong consumer switching costs (our read).

  • After Albergotti says voters turned against data centers in all but one referendum, Altman says people now put data centers far out in the desert, sympathizes with not wanting one nearby, and says the country has room.

01 Huang’s ouster and NVIDIA

Altman calls Jensen Huang a longtime hero of his, adding that heroes have flaws, and expresses personal empathy given his own 2023 board ouster; Albergotti notes a letter circulating about Huang appeared copied from one Altman’s supporters drafted in 2023. He declines to say whether Huang should return, citing insufficient information, but says he can’t picture the AI revolution without Huang and the company he built. Asked whether NVIDIA’s guarantee for OpenAI’s Ohio data center, which Albergotti puts at $105 billion, still holds the same weight post-ouster, Altman says NVIDIA remains a behemoth whose guarantees people still trust, though he concedes Huang’s ambition and scale have mattered to the industry and there could be some ripple effect — while still expecting the industry overall to be fine given strong demand.

02 Going public and chip supply tensions

Altman confirms OpenAI will eventually IPO, citing real downsides to being public — volatility, high-stakes safety decisions — but argues the world as a whole, not just private investors, deserves exposure to AI’s value creation. He dismisses the idea that OpenAI risks becoming an also-ran once public behind Anthropic, arguing investor demand has exceeded supply over the past year. On chip supply, he explains that with many more chipmakers now in the market versus the historical single-provider era, several vendors are each telling the silicon supply chain that OpenAI, Anthropic and others will buy most of their chips — purchases that can’t all materialize, even as he still sees the underlying data-center buildout as solid. He notes supply constraints have shifted over time between memory, rack assembly, data-center capacity, and components like PCB laminate.

03 Data-center backlash and AI politics

Altman argues the industry’s sales pitch has been self-defeating — offering a cure for cancer and cheap goods in exchange for people losing agency — and says data centers serve as a visible proxy for deeper anxiety about power concentration and loss of human control. He says the industry needs to take a firm position against any gradual or sudden handover of decision-making to AI. On politics, he reaffirms his earlier view that he is politically homeless and hopes both parties move toward a more innovation-friendly posture, saying he last spoke with Trump at a China state dinner and would try to reach officials like Stephen Miller if needed.

04 Security, safety, and RSI

Albergotti raises what he calls OpenAI’s own spy drama: a hidden microphone in an office and warnings to employees about possible foreign agents. Altman answers that people look for an edge on AI in all sorts of ways, citing a camera put in a conference room and attempts to get engineers to give up pretraining secrets over drinks; he says OpenAI has had more practice with this than peers now suddenly confronting it, while admitting its security hasn’t been perfect. He frames OpenAI’s safety approach as centrist and pragmatic, rejecting both total secrecy and treating AI like any other technology. Asked about recursive self-improvement, he says OpenAI isn’t letting models self-improve without human oversight, though development is already faster using AI tools, and says OpenAI deliberately holds back on RSI based on the principle that people must stay in control of the future. Asked about a METR report that Anthropic doubled its capability-development speed on specific tasks, he says he doesn’t know of a comparable measurement and that blended numbers are hard to generalize, and notes OpenAI now works with embedded evaluators including METR.

05 Products, memory, and the buildout

On Dots’ memory feature, Altman agrees people should be able to see, modify and export their data, though says Dots’ deep behavioral understanding makes this harder than standard ChatGPT memory export. He defends his controversial RAM comments, clarifying he wants people to buy lots of RAM, but argues shared cloud memory is a more optimal use of scarce supply than idle local RAM given the shortage he flagged years earlier. On debt-funded buildout, he says he hasn’t seen revenue and margin growth like this in any past tech cycle he’s witnessed, expects some excess that needs to come off but not current recklessness, and points to AI-driven science — citing math, physics, chemistry and materials breakthroughs, plus newly expanded free academic compute — as evidence the buildout’s value remains underpriced.


Micron President and CTO Scott DeBoer on the Memory Roadmap, Global Fab Expansion and AI in Chip Design, Live From TomorrowX Day 2 on TBPN

Listen · Follow MU on Ethmos

THE SIGNAL

  • DeBoer says memory has been tight for several years and calls that the reason for the global build-out; the 250B+ US plan plus expansions in Asia suggests a large wave of supply arriving later (our read).

  • Micron’s technology priorities have changed: for about 20 years, bits per wafer and cost drove it, but recently performance, customization and customer co-design matter more than raw bit growth, which supports a less commodity-like margin argument (our read).

  • Even with AI targeted at sharply raising designs per engineer, Micron is still adding staff because the work is labor-heavy, so productivity gains likely become more custom designs, not lower opex (our read).

01 US fab expansion rationale

DeBoer details more than $250 billion in planned investment, including a major new facility underway in Boise, Idaho (Micron’s headquarters city), a larger facility in Clay, New York, and an expansion in Virginia — among the largest industrial projects in US history, in his telling. Pressed on why Micron is building in the US rather than cheaper labor markets, he says the company is balancing its output across the globe, with additional expansion projects ongoing in Asia, but that the US push reflects the concentration of Micron’s largest customers domestically and the strategic value of matching US supply to US demand.

02 Community strategy for new sites

DeBoer says the company’s approach to building large fabs centers on upfront partnership with local communities, community colleges, and schools, citing recent Micron donations in Boise toward roads and housing. He frames the jobs created as durable, given ongoing manufacturing needs, and notes the broader multiplier effect — partner companies supporting fab infrastructure plus retail and restaurant growth in surrounding areas — particularly in New York, a region without prior heavy industry presence.

03 Capacity constraints and customer co-design

Asked when Micron has to refuse customer requests, DeBoer says increased R&D spending has let Micron say yes more often by better understanding customer roadmaps three to four years out and co-designing features jointly. Still, he concedes Micron cannot support every request and must align expectations when capacity or technical roadmap timing doesn’t match customer needs — describing deep, forward-looking roadmap discussions with large customers that didn’t exist when memory was treated as a commodity.

04 Memory’s shifting role in AI

Keep reading with a 7-day free trial

Subscribe to TMT Breakout to keep reading this post and get 7 days of free access to the full post archives.

Already a paid subscriber? Sign in
© 2026 TMT Breakout · Publisher Terms
Substack · Privacy ∙ Terms ∙ Collection notice
Start your SubstackGet the app
Substack is the home for writing, podcasting, and media founders.