Hundreds of tech insiders subscribe to Big Technology’s premium tier for more reporting and to support our independent journalism. Please consider subscribing for just $8 per month.

Subscribe Today

Notes on DeepSeek: Generative AI is All About the Applications Now

Building with AI might cost 5% of what it did a week ago, so what gets built has never been more important.

I want to try to cut through some of the noise that’s circulating on the rise of DeepSeek R1, the new open source AI model from China. We’re going to see so much writing about the model, its origins, and its creators’ intent over the next few days. But no detail will be more meaningful than how cheap DeepSeek makes running AI models.

Infrastructure spending, until this point, has buoyed the entire AI industry. Tech companies spent billions of dollars on data centers and compute, and promised hundreds of billions more, grounding Wall Street’s expectations of the technology’s potential. OpenAI raised $6.6 billion last year, much of it to be spent on training, giving investors a sense of what it expected in return, and hence what they might expect on the dollars they put in.

But DeepSeek’s emergence changes the equation. The company not only learned how to build a leading AI model with far less up front investment, its architecture made cutting edge AI available at a fraction of the cost. DeepSeek today runs at 3-5% of the price of OpenAI’s comparable o1 models. And so developers can now build AI applications at a much lower cost than before.

Upgrade to paid

The focus will therefore soon turn to what you can build with AI vs. how much compute you can assemble to build it. That’s scaring everyone, both because massive infrastructure spending is no longer the benchmark, and because what developers have built with generative AI so far has been slightly underwhelming.

Yes, enterprises have used GenAI for real optimizations, and Salesforce has agents now. But take away the billions spent on infrastructure, and just show the AI products themselves, and the multi-trillion dollar hype hardly feels justified.

The good news is that building with cheaper AI will likely lead to new AI products that previously wouldn’t have existed. It will likely turn expensive enterprise proof of concepts into actual products. And it may give new hope to some working on the wasteland of consumer AI (Apple, of course, was up 3.5% today).

The bad news is we still don’t fully know what to do with generative AI. And so the promise that more efficiency will lead to greater usage isn’t a sure thing. We’re also not sure whether the DeepSeek breakthrough will lead to even greater advances in AI technology, or whether it will immediately commoditize the state of the art, creating less incentive to build it. Or perhaps something in the middle.

Even if there’s a lot to be optimistic about today, you can see why people are a bit jittery. Things are about to get real.

Share

Bonus Podcast: The DeepSeek Reckoning in Silicon Valley

M.G. Siegler is a writer and investor and the author of Spyglass. Siegler joins Big Technology for a bonus depisode to discuss DeepSeek R1, the Chinese open-source AI model and its impact on the tech industry. Tune in to hear why DeepSeek's ability to match OpenAI's performance at just 3-5% of the cost could upend the AI industry's economic model. We also cover the immediate market fallout, why Silicon Valley's scaling hypothesis might be invalidated, and what this means for companies like Microsoft, Google, and NVIDIA. Hit play for a timely analysis of one of the most significant developments in AI that could reshape the technology landscape.

You can listen on Apple Podcasts, Spotify, or your app of choice.

Thank you for reading Big Technology! Paid subscribers get our weekly column, breaking news insights from a panel of experts, monthly stories from Amazon vet Kristi Coulter, and plenty more. Please consider signing up here.

Upgrade To Paid

© 2025 Alex Kantrowitz
548 Market Street PMB 72296, San Francisco, CA 94104
Unsubscribe