TWiT 1062: The Architects of AI - Can Small Models Outrun the Data Center Boom?

Podcast: This Week in Tech (TWiT)
Date: December 15, 2025
Host: Leo Laporte
Guests: Ian Thompson (SiliconLimey.com), Jason Heiner (Editor-in-Chief, The Deep View), Owen Thomas (Managing Editor, San Francisco Business Times)

Overview

This week’s episode dives deep into the current state and future of AI: who are its chief architects, is the AI/data center boom sustainable, and what are the broader societal and regulatory consequences? In classic TWiT fashion, Leo Laporte hosts a roundtable of seasoned tech journalists—each bringing decades of experience to topics ranging from the TIME Person of the Year (the "AI architects"), the environmental and economic impacts of AI’s rapid expansion, skepticism about big model dependence, data privacy, regulatory overreach, and whether small models and direct-to-consumer tech journalism signal the future.

Panel Intros & Setting the Scene

• [00:00] Leo welcomes panel of seasoned insiders: Ian Thompson (SiliconLimey.com), Owen Thomas (SF Business Times), Jason Heiner (formerly ZDNet, now The Deep View).

• Ian describes launching "Silicon Limey" and reprising a "Letter from America"-style column for PC Pro, explaining tech trends to the UK.
  "Just trying to explain to the Brits what exactly is going on over here, which is doing my head in." (Ian, [01:32])

• Jason Heiner shares moving on from ZDNet after 24 years to become Editor-in-Chief of The Deep View, an independent AI-focused publication. Motivations: Direct audience relationships, mission-driven journalism in times of AI confusion and disruption.

Key Discussion Points

1. TIME's "Architects of AI" – Cop-Out or Accurate?

• [09:39] TIME’s Person of the Year: “The Architects of AI” (Altman, Musk, Li, etc.)
  • Owen (who worked at TIME) calls it a "total cop out" ([10:30]):
    "They couldn't settle on you. And so they went with, you know, this hand-wavy architects of AI."
  • Leo: “It’s... the most important and exciting thing that’s happened in technology since the PC.” ([11:51])
  • Comparison to 1982’s “Machine of the Year” (PC) instead of Jobs.

Panel’s Alternative Picks (Jensen Huang, Dario Amodei)

• Jason:
  "Jensen really driving a lot of this, like, AI factories movement... Dario and what Anthropic has done... be the safe model, set the standards... and they're making money, doubling growth every year." ([15:34]-[19:13])
• Discussion of how OpenAI, Google, Anthropic, and others have recently shifted places in terms of model quality and innovation.

2. The AI Wildfire: Bubble, Crash, or Correction?

• [21:01] Referencing Dion Lim’s essay, is the AI/data center boom about to “burn” us?
  • Energy/water use, RAM price spikes, construction labor shortages—all blamed on AI.

Market Correction Analogies and Skepticism

• Ian: "Whenever you see stuff like this going on, my instinctive reaction is, yeah, there’s going to be a market correction. The winners will win, the losers will lose." ([30:16])
• Owen (41:25):
  "A forecast about tech based on current technology is always going to be wrong... this is like when AT&T forecast that everyone in America was going to be a switchboard operator..."

Small Models vs. Big Models

• Jason ([33:41]):
  "What's happening now is... if companies use smaller [AI] models, they're faster and a lot cheaper to run... most workloads that will be growing are going to run on these smaller models."
• Implication: This could break the economic rationale for the big model, ultra-data-center strategies.

3. AI in Healthcare—Real World Value

• [36:10] Discussion of AI “ChatGPT for Doctors” (Open Evidence) and peer-reviewed journal integration.
  • Jason shares anecdotes of patients uploading test results to AI for interpretation, with doctors impressed ([37:26]):
    "The doctor was, you know, pretty forthright about the fact that this is good and it was helpful."
• Emphasis on AI’s real-world, immediate value—especially for synthesis and diagnostics—vs. hype.

4. Environmental Impact and Tech Hysteria

• AI data centers’ impact on power/water (exaggerated?), while streaming/video uses more capacity ([43:19]).
• Nuclear power as a comeback solution.
• Bubble narratives may be overblown; signs to look for in a real bubble:
  "When people are running around saying it's a bubble, there's usually not a bubble. Bubbles usually sneak up on you..." (Jason, [43:41])

5. AI Regulation, Lobbying, & Geopolitical Rhetoric

• Trump’s executive order to head off state-level AI regulation—industry wants national uniform rules.
• Regulatory capture: Key US AI players (Anthropic/Altman vs. Sacks/PayPal) accused of seeking "regulatory capture" ([50:03])

US vs. China "AI Race" Realities

• [63:57] Tim Wu essay: Is "AI race" a straw man to funnel resources to entrenched Western firms?
  • Jason: "The US and China are making each other better in AI... Competition is always good..." ([65:08])
  • Real action is in smaller, more efficient models—China's DeepSeek and others pushing the US to adapt.
  • Owen: “Here in San Francisco, there's obviously a lot of bullishness...” but “extraction” concern: Is AI sucking all venture funding from other sectors? ([70:57])

6. AI Glasses, Wearables, and Privacy

• Google & Meta's XR/Smart Glasses Arms Race
  • Google partners with Xreal, Samsung, Warby Parker (“first glasses with Gemini will arrive 2026”).
  • Meta Ray-Ban stories: Border patrol using glasses for raid recordings ([87:12]); cruise lines banning smart glasses.
  • Societal privacy discomfort and predictably “chunky” early hardware ([89:08]).
  • Jason: Many will wait for Apple/Google hardware because “they trust [them] a little bit more than meta” ([84:10]).

7. Right to Repair—Military Edition

• Congressional block on US military’s right to repair their own equipment ([134:47]): bureaucratic and lobbying pressures keep built-in obsolescence/proprietary repair contracts in effect.
  • Ian: “For the military, [not allowing repair] is completely ridiculous.” ([134:47])

8. Miscellaneous Tech News Highlights

• Social media and age verification ([104:13]):
  Australia bans social media for under 16s. Platforms racing to comply; questions about enforceability, privacy, and unintended consequences (VPNs, lost anonymity).
• Trademark battle for ‘Twitter’ ([167:03]):
  A former Twitter lawyer claims X Corp’s abandonment of the name opens up the trademark.
• Apple App Store antitrust ([140:55]):
  9th Circuit upholds previous ruling: Apple’s 27% fee for alternative payments—still “prohibitive” and non-compliant with the court’s direction.
• AI at CES 2026 ([90:27]):
  All expect CES to be dominated by AI announcements, especially in wearables and XR hardware.
• Excel’s 40th birthday ([145:44]):
  “The spreadsheet remains stubbornly unkillable.” Even die-hard Google Workspace users acknowledge finance teams’ loyalty to Excel.
• Game Awards ([164:29]):
  French indie game "Expedition 33" wins game of the year, signaling indie resurgence.

Notable Quotes & Memorable Moments

On TIME’s Person of the Year:

• Owen Thomas ([10:31]):

  “This is a total cop out by the editors of Time. They couldn't settle on you. And so they went with, you know, this hand-wavy architects of AI. Multiple. Multiple people.”

On the AI Bubble:

• Jason Heiner ([43:41]):

  “When people are running around saying it’s a bubble, there’s usually not a bubble. Bubbles usually sneak up on you...”

On Small AI Models’ Impact:

• Jason Heiner ([33:41]):

  “What's happening now is... if companies use smaller [AI] models, they're faster and a lot cheaper to run... most workloads that will be growing are going to run on these smaller models."

AI in Healthcare Diagnoses:

• Jason Heiner ([37:26]):

  "These two separate people got their lab report, uploaded it, you know, to an AI, got the results and... in one case...the doctor was like, 'this is pretty good... Let’s go on this together.'"

On Streaming vs. AI Power Usage:

• Owen Thomas ([43:19]):

  "By the way, streaming still uses more data center capacity... but you know, no one complains about that because we all take it for granted."

On Right to Repair in the Military:

• Ian Thompson ([134:47]):

  “For the military... not allowing repair is completely ridiculous.”

Timestamps for Key Segments

• [01:32] Ian on “Letter from America” column and launching Silicon Limey
• [05:43] Jason on Deep View and state of AI media
• [09:39] Panel debates TIME’s Person of the Year
• [15:34]-[19:13] “Who really matters in AI?”: Jensen Huang and Dario Amodei praised for impact
• [21:01] “AI Wildfire is coming”—environmental & economic pressures
• [33:41] Jason: Small models & the challenge to Big AI/Data Centers
• [36:10]-[39:42] Real-world AI in medicine: ChatGPT for doctors
• [41:25] Owen: Forecasts based on current tech are always wrong—switchboard operator analogy
• [49:16] Regulatory capture—Sacks’ & PayPal’s history with regulation
• [63:57] Tim Wu essay: “Is the US-China AI race a myth?”
• [84:10] Who will consumers trust with AI wearables?
• [87:12] Border patrol agent caught using Meta Ray-Bans on raid
• [90:27] CES preview: XR, AI everywhere—panel increasing interest
• [104:13] Age verification and kids’ social media bans: issues and workarounds
• [134:47] Military right to repair blocked by Congress
• [140:55] Apple’s latest App Store antitrust defeat
• [145:44] Excel turns 40—and isn’t going anywhere

Tone & Atmosphere

As always, the show is spirited, skeptical, sometimes snarky, but fundamentally engaged and optimistic about independent journalism and innovation. There’s camaraderie, and the banter is both nerdy (Excel esports, F1 jokes) and pointed (regulatory capture, media consolidation, privacy).

Final Thoughts

The panel concludes that while AI’s disruption is real, the stories we tell—"runaway AI," "the China threat," "catastrophic data center overload"—are often motivated by lobbying, self-preservation, or misplaced tech pessimism/hype cycles. The reality is more nuanced: Small models are rising. Regulation is being shaped actively by entrenched powers. Media, like AI architectures, is also decentralizing as independent, niche, and trust-centric outlets rise.

Leo sums up:
"I'm proud of both of you guys for going independent. That's why I'm proud to be doing what we do. The more voices, the better. The less beholden to government or Wall Street, the better." ([124:29])

TL;DR

• TIME’s “Architects of AI” Person of the Year is critiqued as a cop out, but AI’s impact is undeniable.
• Big model/data center mania is reaching its limits—small, efficient, and specialized models are rising fast.
• Real-world value is emerging in places like medical diagnostics ("ChatGPT for doctors").
• Environmental impacts/rate of expansion may spark correction, but it's not 2000 or 2008 redux.
• US/China “AI race” is more rhetoric than zero-sum fact; both drive each other’s progress.
• Ongoing privacy, data, and right to repair debates reflect wider anxieties about consolidation and control.
• Tech journalism is rapidly fragmenting and going direct/independent—just like the future of AI.
• The panel is bullish on independent voices, skeptical about hype, and united in tech’s unpredictable evolution.

For those who didn’t listen: This episode is your one-stop catch-up on the true tectonic shifts in AI and its ecosystem—beyond the hype, beyond the easy scare stories. It’s a conversation only long-timers can have: skeptical, insightful, but still excited for the future.

[Timestamps are MM:SS from the beginning of the published episode. All quotes are as near verbatim as possible.]