AI-Briefing: DeepSeek’s emergence from nowhere shows open-source is eating the world

A key question facing the $225 billion-plus U.S. digital media sector is, how will its key players respond?

Developments last week hint at such players adopting an open-source approach in a rapidly evolving industry landscape. Meanwhile, Digiday’s ad tech sources noted that, while DeepSeek poses a credible alternative to Big Tech, clients are in a cautious mood, particularly around privacy.

The market disruption wrought by the technical feats of China-based DeepSeek’s new R1 large language model developed at a fraction of the cost of U.S. rival and AI talisman OpenAI and ChatGPT was demonstrated in stock prices.

Notably, Nvidia, a leading supplier of GPUs essential for AI development, experienced a record one-day market value loss of approximately $593 billion, marking the largest single-day loss for any company to date.

Meanwhile, the resulting impact on the market capitalization drops of Alphabet, Amazon, Apple, Meta and Microsoft — collectively, these companies dominate the ad-funded internet — were declines ranging from 5% to 10% in a similar period.

Opening up

The key to DeepSeek’s emergence was its open-source approach to developing its R1 model. This contrasts with OpenAI’s more proprietary strategy, where models are typically closed-source and access is provided through APIs.

Despite the political tenor of the development, as exemplified by the privacy concerns flagged by Digiday sources, some of the internet economy’s foundational names are making moves that hint at an open future.

For example, DeepSeek’s R1 model is now available on AWS, a sign (for some) of how the Amazon machine views others’ margin as its opportunity and will do whatever it takes to push its infrastructure strategy.

Related Insights

Beyond Ads

A deep(er) dive into DeepSeek’s privacy policies  Read More

One ad tech founder said potential clients are asking them to make sure they don’t use tech owned by companies like Amazon and Google, which makes having an alternative even more appealing. However, ad tech companies are still wary of using DeepSeek’s API and instead are looking for other options like running R1 on-premise.

Elsewhere, some have noted the timing of Google’s Meridian — the online advertising giant’s open-source marketing mix model (MMM) — last week as significant. As Digiday sources recently observed, measurement is at the core of Google’s indispensableness in any media plan.

However, as discussed extensively at last week’s annual leadership meeting hosted by the IAB, the loss of traditional audience-targeting signals, primarily third-party cookies, means Google’s ability to calculate how ads drive purchases is waning.

Some interpret this embrace of the open approach from traditional walled garden players as a significant shift in their approach. Given the disruptive impact of AI on Google’s search hegemony, observers speculate if similar significant shifts in outlook can be expected from the titans of Big Tech.

And the potential for downstream impacts are manifold, with ad tech sources telling Dgiday in separate conversations last week that they believe open-source models give them a chance to “control our own destiny.”

Buy-side and sell-side sources told Digiday that R1’s transparency creates an alternative to black boxes like OpenAI and also offers them the opportunity to rely less on storing data with competitors like Google and Amazon. That doesn’t mean the Big Four won’t have a moat, but it does mean the moat’s eroded — at least for now.

Market reactions

Meanwhile, OpenAI-patron Microsoft reported its latest earnings report last week, with CEO Satya Nadella declaring its AI business was approaching a run rate of $13 billion per year, up 175% year over year. This lofty figure failed to impress markets, though, as a comparatively modest revenue outlook for the coming quarter, along with residual concerns over the impact of DeepSeek, proved a drag on Microsoft’s stock price after earnings.

Elsewhere, Apple also issued its earnings last week, and while its subsequent stock price fluctuations may have fared more stably than others in its Big Tech cohort, some noted how it will have to accelerate its AI deployment, i.e. Apple Intelligence, if it is to gain momentum.

Given Apple’s increasing interest in the “services” sector, which increased 14% year on year for the reporting period to surpass $26 billion, and advertising in particular, it will be interesting to see just how AI will play a role in its Madison Avenue aspirations.

Chris Vanderhook, CEO and co-founder of publicly-listed ad tech company Viant, noted how the recent developments have been significant. “Because AI adoption is so early, they [Google, Meta, Amazon, etc.] didn’t get their hooks and claws into everyone yet,” he told Digiday.

Prompts and Products: AI news and announcements

• The U.S. Copyright Office said it will allow applicants to copyright AI-generated works as long as a human played enough of a role in the creation or editing process.
• Tech and consumer advocacy groups urged the White House to keep rules for responsible AI development in place.