Feb 6, 2025 · 44m · a16z
What DeepSeek Means For The Future Of AI | Tech Veterans Weigh In
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
Tech veterans Steven Sinofsky and Martin Casado analyze the global implications of DeepSeek's open-source reasoning model release on the a16z podcast. They argue that DeepSeek represents a structural shift toward algorithmic efficiency, local edge compute, and application-layer value capture rather than a geopolitical threat or market crash.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. How this is scored →
The host as informed peer 2.8 Guest teaching 5.5 Guest disagreement 1.8 The host pushing back 1.6
speaking balance: gold is the host, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 35:47 Premise rejection on boardroom focus
Martin directly interrupts and rejects the host's framing ('I don't want to talk to the boardroom. I want to talk to the US government, right?'), steering the conversation forcefully toward policy errors.
Hardest push from the host ▶ 18:41 Host challenges core historical analogyHost explicitly pushes back on Steven's foundational premise ('I just want to probe you, is there something different here?'), forcing the guests to justify why AI will follow the exact pattern of the internet.
Biggest teaching moment ▶ 22:36 Explanation of scale-out computing historySteven educates the host and audience on the transition from mainframes and workstations to microcomputers, demonstrating why distributed endpoint MIPS inevitably beat centralized data centers.
The host holds their own ▶ 35:06 Host synthesizes enterprise insights and historical parallelsHost demonstrates expertise by connecting Scott Belsky's insights on enterprise software needs with JFK's post-Sputnik moonshot speech to challenge the guests on actionable AI policy.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The host as informed peer | Guest teaching | Guest disagreement | The host pushing back | Why |
|---|---|---|---|---|---|---|
| Media Frenzy & DeepSeek's Sudden Arrival | 3 | 4 | 1 | 1 | Host sets up the episode by asking for a TLDR on DeepSeek's sudden viral arrival and market overreaction. Steven and Martin explain that while the $6M figure triggered market panic, the release was the culmination of long-term research. | |
| Separating Signal from Noise: Chinese AI Talent | 4 | 5 | 1 | 1 | Host synthesizes details about DeepSeek's V3 base model and recent R1 release to ask for the signal from the noise. Martin clarifies that strong Chinese AI talent explains the achievement rather than an unexplained miracle. | |
| The Hyperscaler Trap vs. End-Point Compute Trajectories | 0 | 6 | 2 | 0 | Host does not speak in this segment. Steven critiques the hyperscaler trap of assuming infinite compute and data, arguing compute will naturally shift toward billions of endpoint devices. | |
| Engineering under Constraints & Data Advantage | 5 | 5 | 1 | 2 | Martin discusses engineering under constraints and data advantages in China. Host intervenes to highlight overlooked technical factors like released reasoning traces and permissive MIT licensing. | |
| Business Models & The Internet Infrastructure Analogy | 0 | 6 | 1 | 0 | Host remains silent throughout this segment. Steven compares selling base LLMs to monetizing early HTTP web servers, predicting value will migrate to higher application layers. | |
| Vertically Integrated Applications vs. Model Commoditization | 0 | 6 | 2 | 0 | Host is absent during this guest exchange. Martin and Steven debate whether apps will require vertically integrated models or if models will commoditize like legacy desktop software. | |
| Capital Expenditure, Tech Balance Sheets, and Market Stability | 5 | 5 | 2 | 6 | Host actively probes Steven's claim that AI must follow the internet trajectory, asking if something fundamental is different this time. Martin clarifies why strong cloud balance sheets prevent a 2000s telecom crash. | |
| Scale-Up vs. Scale-Out & On-Device Local AI | 4 | 6 | 1 | 2 | Host explicitly references Steven's article on scale-up versus scale-out compute dynamics. Steven explains how cheap endpoint computing historically defeated expensive mainframes, pointing to local AI models. | |
| Best-Effort Delivery: The 'Switching Wars' Parallel | 0 | 6 | 1 | 0 | Host does not participate in this segment. Martin and Steven compare LLM imperfections and hallucinations to best-effort delivery protocols during the 1990s telecom switching wars. | |
| Expanding the TAM and New Product Benchmarks | 5 | 5 | 1 | 3 | Host asks whether parameter count and coding test benchmarks are becoming obsolete in favor of device fit and operational cost metrics. Steven agrees that benchmarks quickly become irrelevant. | |
| US Policy, Export Controls, and Regulatory Realities | 6 | 5 | 7 | 4 | Host brings domain knowledge from Scott Belsky on enterprise AI needs and frames DeepSeek as a Sputnik moment requiring boardroom action. Martin forcefully rejects the premise, pivoting to critique US export policy. | |
| Technology Diffusion & Historical Export Control Failures | 0 | 6 | 2 | 0 | Host does not speak in this segment. Steven and Martin argue that government export controls on software and chips are historically ineffective, referencing past failed encryption and gaming console controls. | |
| Strategic Imperatives for AI Labs & Unexpected Innovation Sources | 4 | 6 | 1 | 2 | Host asks if DeepSeek originating from a quant hedge fund indicates that a wider group of entrants can compete. Steven confirms that major tech shifts historically originate outside dominant legacy labs. |
Statements from this episode (36)
Sinofsky: $1T market selloff over DeepSeek was a complete overreaction
“Like on a Friday, the whole weekend was like everybody whipping themselves into a frenzy so they could wake up Monday morning and trade away a trillion dollars of market cap, which seems to be a, Complete overreaction and craziness, but that's not what we're h…”
Casado: DeepSeek aggregated existing public research rather than inventing core techniques
“All of the contributions that they've done have been in the public literature somewhere. Just nobody really aggregated.”
Casado: DeepSeek's $6M reasoning model cost aligns with OpenAI and Anthropic
“The fact that they spent six million dollars just in the chain of thought is actually not out of whack with what Anthropik has now said they've spent in OpenAI has said that they spent.”
Casado: GPT-style base models reached a performance plateau around GPT-4
“Yeah, like the previous base models, you know, like the GPT lineage seemed to have asymptoted around GPT-IV.”
Sinofsky: Endpoint devices will eventually surpass centralized data center compute capacity
“Well, at some point you're going to end up breaking the problem up to the seven billion endpoints of the world, which will have vastly more compute than you can ever squeeze into one giant nuclear power data center.”
Sinofsky: Overcapitalized Western AI companies lost focus on efficiency
“And I feel like that's where the AI community in, in the U S particularly, or the West, if you will, got just a little carried away. And it was, you know, just like every, you know, startup that has too much money, you know, the snacks get a little too good.”
Sinofsky: The Chinese internet is a far superior AI training dataset
“Well, even on the data, you know, their starting point is, you know, the Chinese internet per se, those, that has much more structure to it. It's a much better training set.”
Casado: China offers unbeatable global arbitrage for high-quality AI data annotation
“If you want to look at a place to arbitrage really smart, educated people and relatively low cost, it's hard to beat China globally, right? And so they definitely have access to a bunch of, you know, potentially highly educated annotated data, which is very re…”
Casado: DeepSeek's success is legitimate engineering, not a PSYOP
“So I, you know, I happen to be able to believe it sounds like you are too, that this did not come out of nowhere. It's not a PSYOP. It's not, it's like, this is a great team taking advantage of what it has closely”
Casado: Every a16z App-Layer AI Startup Uses Multiple Models
“I mean, I think at A-Six and Z, we have one of the large portfolio of AI companies, both at the model layer and at the app layer. And I will say any company at the app layer is using many models. Like, it's not just like one model. Like, I have yet to see the …”
Casado: DeepSeek Reasoning Traces Enable Model Distillation for Edge Devices
“It turns out that that chain of thought, if you have access to that, it allows you to train smaller models very quickly and very cheaply, and that's called distilling. So like the general, Term of distilling in LLM world means you have a teacher model, train a…”
Sinofsky: Monetizing base AI models directly is like charging for HTTP servers
“And the rest of the companies are still trying to figure out their revenue models, which I would argue was probably premature. And it starts to look a little to me like, hey, let's charge for a web server. And it's like the business of selling, like serving HT…”
Sinofsky: AI stacks will standardize and licensing choices will become critical
“The licensing model really matters because what's going to happen is that there's going to end up being some level of standardization. Now, I don't know where in the stack or in what level, but there is going to be some level of standardization and the model, …”
Casado: DeepSeek's Threat to OpenAI Is Limited if AI Apps Require Proprietary Models
“It could be that the apps actually require you to own the model. And in that case, DeepSeek is less relevant because they're not building apps. And then, you know, this means that the impact to the opening AIs or Anthropics are not as great, right?”
Sinofsky: AI trajectory will mirror the decentralized internet
“There's no way this doesn't play out like the internet.”
Sinofsky: LLMs will struggle to replace traditional search engines
“These LLMs are going to replace search. But it turns out that's actually going to be really, really hard because there's a lot of things that search does that, that the models are bad at it really bad.”
Sinofsky: Category-defining AI apps will be entirely novel software concepts
“The apps that ended up mattering on the internet literally didn't exist. Before the internet. And I think that's what people are losing sight of.”
Casado: DeepSeek R1 standalone will not have a deep impact
“I don't think that R-One itself as a standalone is going to have that deep of an impact.”
Casado: Strong balance sheets prevent AI infrastructure market crash
“You have a much better foundation. Much, much better foundation for the AI wave, right? Like the primary investors are the big three cloud companies. They've got hundreds of billions of dollars on the balance sheet. Even if all of this goes away, they'll be fi…”
Sinofsky: Microsoft spent $30B to $40B developing Bing
“People forget that Microsoft poured, I don't know, 30, forty billion dollars into Bing.”
Casado: Meta currently spends more on VR than AI
“I would bet, I don't know this as a fact, I'll bet Meta's spending more money on VR than it is on AI right now.”
Sinofsky: AI will evolve toward local, cost-free on-device execution
“And then you look at Apple and their strategy, which the execution hasn't been great, but the idea that all of these things will just surface as features popping up all over my phone and they're not going to cost anything. My data is not going to go anywhere. …”
Casado: Flawed AI models unlock a rapidly growing white space
“They enable an entirely new set of stuff like Creativity and coding. And it's an entirely white space and it's going to grow very quickly.”
Sinofsky: New tech paradigms succeed because liabilities become features
“They can't understand that these paradigms where, where like the liabilities either don't matter or just become features.”
Casado: DeepSeek is a step toward AGI in your pocket
“This is another step to basically AGI in your pocket.”
Casado: Shorting Nvidia over DeepSeek's launch is the wrong answer
“My reaction was not, oh, shit, I need to, like, short NVIDIA or whatever. I think that's actually the wrong answer.”
Sinofsky: DeepSeek's efficiency leap is AI's JavaScript moment
“And it's, to me, this is all the equivalent of the browser getting JavaScript. Because once the browser got JavaScript, then all of a sudden you could do anything you needed without going to some standards body or building your own browser.”
Casado: Base AI models lack defensibility without retentive application workflows
“The thing is, is all the other models catch up very quickly, because they distill so well, so like, that's not defensible, in a way. And so the companies that are defensible, that I've seen, is they'll put out a model that's very compelling, And then once the …”
Casado: US AI export controls failed; regulation is sole obstacle to winning
“China is a lot of very smart people. They're incredibly capable. They're great researchers. They can build stuff as great as well as we can, and they can open source it. We did not enable them. They do this even with export controls on chips. Right? So there's…”
Casado: Global AI development is a space race the US must win
“It really is the AI race, just like we went through the space race, and we need to win.”
Casado: Historical US hardware export controls were a total failure
“We've actually put export controls on GPUs before. I mean, like a perfect analog. We were like, oh, listen, you can do weapon simulation on these things. Like a PlayStation was the first to actually use the SGI. Remember that we're gonna export control that. L…”
Sinofsky: DeepSeek DAUs reached roughly 35% of OpenAI's total
“DeepSeq, I think is, was the number I saw this morning is like 35% of the DAUs of open AI.”
Casado: DeepSeek release is not a crisis for OpenAI, Anthropic, or Nvidia
“I personally don't think this is a crisis moment for open AI or Anthropic. I think like apps are hard to build. I think that like right now the apps that they put out are very complex. They actually know their users. They have very specific use cases. And so, …”
Sinofsky: Building first-party apps is key to creating great software platforms
“Because the best feedback loop to build a great platform for other people to use is to be building apps.”
Sinofsky: AI TAM will grow 100x, with revenue captured primarily by applications
“And so I think that that the TAM is going to be a hundred X. It's going to be every endpoint. The revenue is going to come from the apps side of it, and then there'll be a developer side of it. It'll just be a different pricing model for different sets of scen…”