StarZero All shows Try the podcast clipper →

Jan 27, 2025 · 44m · big-technology

How DeepSeek Changes AI Research & Silicon Valley w/ M.G. Siegler

M.G. Siegler · 27m spoken Alex Kantrowitz · 14m spoken
0:00 / 0:00
▶ Watch on YouTube →

gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions

Host Alex Kantrowitz and tech investor M.G. Siegler analyze how DeepSeek R1's cost-effective reasoning breakthroughs disrupt Silicon Valley's compute-heavy scaling paradigm. They explore the architectural, financial, and venture implications of open-weight artificial intelligence models achieving frontier performance at a fraction of traditional capital expenditure.

How this conversation actually went

Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Alex holds 35% of the talking time here. How this is scored →

Alex as informed peer 6.7 Guest teaching 3.8 Guest disagreement 1.5 Alex pushing back 2.8
05100:0015:0030:003:28–10:49 · Alex as informed peer 6/10 Market Reactions and Innovation Under Hardware Constraints Alex opens with detailed benchmark metrics and pricing data for DeepSeek R1 relative to OpenAI. MG provides historical context on export controls forcing innovation under chip scarcity in China.10:50–15:39 · Alex as informed peer 7/10 DeepSeek Technical Breakthroughs: Model Distillation and Pure RL Alex demonstrates strong technical understanding by highlighting the transition from self-supervised learning to pure reinforcement learning. MG outlines DeepSeek's origins as a quant trading fund and the mechanics of distilling frontier models.15:39–22:05 · Alex as informed peer 7/10 Challenging the Scaling Hypothesis and AI Capex Spending Alex connects Demis Hassabis's comments on the AI wall and quotes MG's newsletter directly regarding the scaling hypothesis. MG explains how Project Stargate and capex offloading signal hyperscalers reconsidering massive scaling spend.22:05–31:28 · Alex as informed peer 7/10 Wall Street Reckoning and Big Tech Revenue Models Alex outlines the pre-market stock selloff and frames the AI industry's reliance on wealth transfers from search and ad revenues. MG offers a detailed analogy comparing current AI infrastructure overspend to the streaming wars.31:28–38:39 · Alex as informed peer 7/10 Jevons Paradox and the Search for Economic Utility Alex challenges the tech elite narrative around Jevons Paradox by emphasizing the persistent lack of enterprise AI adoption beyond proof-of-concepts. MG agrees that tech leadership messaging is coordinated and acknowledges consumer implementations like Apple Intelligence remain underwhelming.38:39–44:42 · Alex as informed peer 6/10 Startup Implications, M&A Dynamics, and Future Trajectory Alex counters the idea that high inference costs were holding back startups, noting VC funding was already abundant without sparking consumer breakout hits. MG explains how the chill on tech M&A trapped talent inside Big Tech and why CEOs cannot easily cut committed capex.3:28–10:49 · Guest teaching 4/10 Market Reactions and Innovation Under Hardware Constraints Alex opens with detailed benchmark metrics and pricing data for DeepSeek R1 relative to OpenAI. MG provides historical context on export controls forcing innovation under chip scarcity in China.10:50–15:39 · Guest teaching 4/10 DeepSeek Technical Breakthroughs: Model Distillation and Pure RL Alex demonstrates strong technical understanding by highlighting the transition from self-supervised learning to pure reinforcement learning. MG outlines DeepSeek's origins as a quant trading fund and the mechanics of distilling frontier models.15:39–22:05 · Guest teaching 3/10 Challenging the Scaling Hypothesis and AI Capex Spending Alex connects Demis Hassabis's comments on the AI wall and quotes MG's newsletter directly regarding the scaling hypothesis. MG explains how Project Stargate and capex offloading signal hyperscalers reconsidering massive scaling spend.22:05–31:28 · Guest teaching 5/10 Wall Street Reckoning and Big Tech Revenue Models Alex outlines the pre-market stock selloff and frames the AI industry's reliance on wealth transfers from search and ad revenues. MG offers a detailed analogy comparing current AI infrastructure overspend to the streaming wars.31:28–38:39 · Guest teaching 3/10 Jevons Paradox and the Search for Economic Utility Alex challenges the tech elite narrative around Jevons Paradox by emphasizing the persistent lack of enterprise AI adoption beyond proof-of-concepts. MG agrees that tech leadership messaging is coordinated and acknowledges consumer implementations like Apple Intelligence remain underwhelming.38:39–44:42 · Guest teaching 4/10 Startup Implications, M&A Dynamics, and Future Trajectory Alex counters the idea that high inference costs were holding back startups, noting VC funding was already abundant without sparking consumer breakout hits. MG explains how the chill on tech M&A trapped talent inside Big Tech and why CEOs cannot easily cut committed capex.3:28–10:49 · Guest disagreement 1/10 Market Reactions and Innovation Under Hardware Constraints Alex opens with detailed benchmark metrics and pricing data for DeepSeek R1 relative to OpenAI. MG provides historical context on export controls forcing innovation under chip scarcity in China.10:50–15:39 · Guest disagreement 1/10 DeepSeek Technical Breakthroughs: Model Distillation and Pure RL Alex demonstrates strong technical understanding by highlighting the transition from self-supervised learning to pure reinforcement learning. MG outlines DeepSeek's origins as a quant trading fund and the mechanics of distilling frontier models.15:39–22:05 · Guest disagreement 1/10 Challenging the Scaling Hypothesis and AI Capex Spending Alex connects Demis Hassabis's comments on the AI wall and quotes MG's newsletter directly regarding the scaling hypothesis. MG explains how Project Stargate and capex offloading signal hyperscalers reconsidering massive scaling spend.22:05–31:28 · Guest disagreement 2/10 Wall Street Reckoning and Big Tech Revenue Models Alex outlines the pre-market stock selloff and frames the AI industry's reliance on wealth transfers from search and ad revenues. MG offers a detailed analogy comparing current AI infrastructure overspend to the streaming wars.31:28–38:39 · Guest disagreement 2/10 Jevons Paradox and the Search for Economic Utility Alex challenges the tech elite narrative around Jevons Paradox by emphasizing the persistent lack of enterprise AI adoption beyond proof-of-concepts. MG agrees that tech leadership messaging is coordinated and acknowledges consumer implementations like Apple Intelligence remain underwhelming.38:39–44:42 · Guest disagreement 2/10 Startup Implications, M&A Dynamics, and Future Trajectory Alex counters the idea that high inference costs were holding back startups, noting VC funding was already abundant without sparking consumer breakout hits. MG explains how the chill on tech M&A trapped talent inside Big Tech and why CEOs cannot easily cut committed capex.3:28–10:49 · Alex pushing back 2/10 Market Reactions and Innovation Under Hardware Constraints Alex opens with detailed benchmark metrics and pricing data for DeepSeek R1 relative to OpenAI. MG provides historical context on export controls forcing innovation under chip scarcity in China.10:50–15:39 · Alex pushing back 1/10 DeepSeek Technical Breakthroughs: Model Distillation and Pure RL Alex demonstrates strong technical understanding by highlighting the transition from self-supervised learning to pure reinforcement learning. MG outlines DeepSeek's origins as a quant trading fund and the mechanics of distilling frontier models.15:39–22:05 · Alex pushing back 3/10 Challenging the Scaling Hypothesis and AI Capex Spending Alex connects Demis Hassabis's comments on the AI wall and quotes MG's newsletter directly regarding the scaling hypothesis. MG explains how Project Stargate and capex offloading signal hyperscalers reconsidering massive scaling spend.22:05–31:28 · Alex pushing back 3/10 Wall Street Reckoning and Big Tech Revenue Models Alex outlines the pre-market stock selloff and frames the AI industry's reliance on wealth transfers from search and ad revenues. MG offers a detailed analogy comparing current AI infrastructure overspend to the streaming wars.31:28–38:39 · Alex pushing back 4/10 Jevons Paradox and the Search for Economic Utility Alex challenges the tech elite narrative around Jevons Paradox by emphasizing the persistent lack of enterprise AI adoption beyond proof-of-concepts. MG agrees that tech leadership messaging is coordinated and acknowledges consumer implementations like Apple Intelligence remain underwhelming.38:39–44:42 · Alex pushing back 4/10 Startup Implications, M&A Dynamics, and Future Trajectory Alex counters the idea that high inference costs were holding back startups, noting VC funding was already abundant without sparking consumer breakout hits. MG explains how the chill on tech M&A trapped talent inside Big Tech and why CEOs cannot easily cut committed capex.

speaking balance: gold is Alex, purple is the guest (3 minute bins)

0:00 · Alex 94% · guest 6%0:00 · Alex 94% · guest 6%3:00 · Alex 15.6% · guest 84.4%3:00 · Alex 15.6% · guest 84.4%6:00 · Alex 28.6% · guest 71.4%6:00 · Alex 28.6% · guest 71.4%9:00 · Alex 31.7% · guest 68.3%9:00 · Alex 31.7% · guest 68.3%12:00 · Alex 35.2% · guest 64.8%12:00 · Alex 35.2% · guest 64.8%15:00 · Alex 36.1% · guest 63.9%15:00 · Alex 36.1% · guest 63.9%18:00 · Alex 46.6% · guest 53.4%18:00 · Alex 46.6% · guest 53.4%21:00 · Alex 55.9% · guest 44.1%21:00 · Alex 55.9% · guest 44.1%24:00 · Alex 0% · guest 100%24:00 · Alex 0% · guest 100%27:00 · Alex 29.9% · guest 70.1%27:00 · Alex 29.9% · guest 70.1%30:00 · Alex 49.9% · guest 50.1%30:00 · Alex 49.9% · guest 50.1%33:00 · Alex 32.8% · guest 67.2%33:00 · Alex 32.8% · guest 67.2%36:00 · Alex 20.7% · guest 79.3%36:00 · Alex 20.7% · guest 79.3%39:00 · Alex 18.5% · guest 81.5%39:00 · Alex 18.5% · guest 81.5%42:00 · Alex 30.3% · guest 69.7%42:00 · Alex 30.3% · guest 69.7%
Sharpest disagreement ▶ 28:08 Pushing Back on the AGI Spending Mantra

MG dismisses tech leadership talking points about spending hundreds of billions toward AGI as superficial mantras designed to placate Wall Street rather than sustainable business strategies.

 Hardest push from Alex ▶ 39:50 Rejecting the Cost-Barrier Argument for Startups

Alex firmly disputes the premise that lowered compute costs will suddenly trigger a flood of new AI startups, citing the lack of consumer app breakouts despite years of unlimited venture subsidies.

 Biggest teaching moment ▶ 28:40 Wall Street Cycles and the Streaming Analogy

MG delivers a masterclass in market dynamics by comparing tech's current AI capex boom to Hollywood's streaming bubble during the pandemic, explaining how Wall Street inevitably flips from demanding growth to enforcing efficiency.

 Alex holds their own ▶ 13:08 Technical Distinction of Pure RL and Distillation

Alex demonstrates sharp domain expertise by precisely explaining how DeepSeek bypassed traditional self-supervised learning in favor of pure reinforcement learning and efficient model distillation.
the scores for every segment, with the reasoning behind each
ChapterTopicAlex as informed peerGuest teachingGuest disagreementAlex pushing backWhy
Market Reactions and Innovation Under Hardware Constraints 6412 Alex opens with detailed benchmark metrics and pricing data for DeepSeek R1 relative to OpenAI. MG provides historical context on export controls forcing innovation under chip scarcity in China.
DeepSeek Technical Breakthroughs: Model Distillation and Pure RL 7411 Alex demonstrates strong technical understanding by highlighting the transition from self-supervised learning to pure reinforcement learning. MG outlines DeepSeek's origins as a quant trading fund and the mechanics of distilling frontier models.
Challenging the Scaling Hypothesis and AI Capex Spending 7313 Alex connects Demis Hassabis's comments on the AI wall and quotes MG's newsletter directly regarding the scaling hypothesis. MG explains how Project Stargate and capex offloading signal hyperscalers reconsidering massive scaling spend.
Wall Street Reckoning and Big Tech Revenue Models 7523 Alex outlines the pre-market stock selloff and frames the AI industry's reliance on wealth transfers from search and ad revenues. MG offers a detailed analogy comparing current AI infrastructure overspend to the streaming wars.
Jevons Paradox and the Search for Economic Utility 7324 Alex challenges the tech elite narrative around Jevons Paradox by emphasizing the persistent lack of enterprise AI adoption beyond proof-of-concepts. MG agrees that tech leadership messaging is coordinated and acknowledges consumer implementations like Apple Intelligence remain underwhelming.
Startup Implications, M&A Dynamics, and Future Trajectory 6424 Alex counters the idea that high inference costs were holding back startups, noting VC funding was already abundant without sparking consumer breakout hits. MG explains how the chill on tech M&A trapped talent inside Big Tech and why CEOs cannot easily cut committed capex.

Statements from this episode (12)

Assertion Supported
Kantrowitz: DeepSeek R1 tops OpenAI o1 on AIME math benchmark
“On the AIME mathematics test, it scored 79.8% compared to OpenAI's O-one scoring 79.2%.”
Alex Kantrowitz Jan 27, 2025 ▶ 2:05
Assertion Supported
DeepSeek R1 API costs just 3.5% of OpenAI's o1 pricing
“It costs 55 cents per million token inputs and two dollars and 19 cents per million token outputs. Just to give you a sense, OpenAI costs 15 dollars per million input tokens and 60 dollars per million output tokens. That's 3.5% of the cost that it costs to run…”
Alex Kantrowitz Jan 27, 2025 ▶ 2:29
Insight
U.S. chip restrictions forced Chinese AI innovation through compute scarcity
“The constraints that were put in place by the US because of the, you know, everything going on with chip constraints and sort of forcing AI companies not to export to China you know, led to sort of this very interesting cauldron that I think could only happen …”
M.G. Siegler Jan 27, 2025 ▶ 9:57
Assertion Partly supported
DeepSeek used model distillation to match OpenAI's o1 at lower cost
“And, you know, they've just used the process of distillation to, you know, effectively bring those bigger versions of the sort of state-of-the-art models and distill them down into you know, smaller models, which eventually led to this R-one, you know, the equ…”
M.G. Siegler Jan 27, 2025 ▶ 12:00
Insight
Siegler: Embedding all world knowledge in every AI model is overkill
“Like, do you need all of the world's knowledge, you know, in every single model for every single use case? Of course not. Like that's going to be overkill for almost everything that you're going to do.”
M.G. Siegler Jan 27, 2025 ▶ 14:46
Assertion Supported
Demis Hassabis acknowledges AI progress is slowing and taking longer
“Even Demis, you know, when you talk to him, he noted that he doesn't necessarily believe in, in, you know, a wall being hit, but he did acknowledge that things are slowing and it'll just take longer to get more, you know, juice out of the squeeze as it were.”
M.G. Siegler Jan 27, 2025 ▶ 16:28
Assertion Supported
Tech stocks plunged in pre-market trading after DeepSeek R1's release
“NVIDIA down 10%, Microsoft down four percent. Google down three percent meta down 2.6% SMP down two percent. So this is all based off of this deep seek reckoning or this deep seek realization.”
Alex Kantrowitz Jan 27, 2025 ▶ 21:50
Opinion
Meta is better positioned for open-source AI disruption than peers
“Meta's probably in a bit better position than the other ones, just because they, at the end of the day, they do want, like, you know, their whole philosophy is to open source this, not for necessarily altruistic reasons, but because they know that it's histori…”
M.G. Siegler Jan 27, 2025 ▶ 25:50
Assertion Supported
Kantrowitz: OpenAI has 300M ChatGPT users but loses billions annually
“We have OpenAI who has ChatGPT with three hundred million users, which is okay. But still losing billions a year to run that thing.”
Alex Kantrowitz Jan 27, 2025 ▶ 33:14
Opinion
Apple has failed to successfully implement its Apple Intelligence features
“Apple intelligence is almost the perfect example of the problem that I'm pointing toward, which is that we have this technology that's so promising, and yet even Apple cannot implement it successfully. And that might, I mean, obviously it says something about …”
Alex Kantrowitz Jan 27, 2025 ▶ 37:17
Opinion
Antitrust chill stopped Big Tech talent from leaving to found startups
“I do feel like there was a bit of a chilling effect the past year because M&A had, you know, basically been shut off. That sort of kept people staying at Google and staying at Meta and staying at OpenAI, not forming new startups as they might have in, in years…”
M.G. Siegler Jan 27, 2025 ▶ 41:19
Prediction Held up
Big Tech cannot cut AI CapEx due to catastrophic FOMO risk
“I think that it won't be so easy also for, as I noted, for all these companies to pull back spend because they've already committed to buying X number of H. 200 chips, and then soon enough, we'll get the next iteration, you know, announced down the road, and s…”
M.G. Siegler Jan 27, 2025 ▶ 43:16
Made with StarZero

Turn any episode into a week of clips.

This entire site, over 300 episodes transcribed, diarized, checked and made playable, runs on the StarZero media pipeline. Drop in your own episode and the podcast clipper finds the moments worth sharing, cuts them, captions them, and reframes them for every feed.

Try the podcast clipper Learn more