Reddit founder's unexpected message about AI doomsday panic

21 Sep 2026, 13:17 · The Fresno Bee

For two weeks in September, the most powerful people in technology argued in public about whether their own products might kill everyone. Almost all of them had money riding on the answer.

Alexis Ohanian does not. He co-founded Reddit in a college dorm back in 2005, resigned from its board in 2020, and now runs the venture firm Seven Seven Six. He told CNBC on Wednesday, Sept. 16, that the industry has done a “pretty tone deaf job” of explaining AI to the public.

His objection is not that the danger is invented. It is that the version being sold is the wrong one, and that it is drowning out the version investors can actually price.

Ohanian said the risks are “more benign, or they’re more banal than people are thinking.” He pointed to agent swarms, where multiple AI agents coordinate on one complicated task, rather than the Terminator and Skynet scenarios dominating social feeds.

He compared AI to nuclear energy, which is powerful enough to do real damage or real good, depending on how it is handled.

Ohanian was sharper about the tone of the debate. “The doomer tweet was discouraging,” he said, and called it “frustrating” when “the slogans are actually not real things.”

Everyone else in the AI doomsday debate has a position

The argument he walked into started on Sept. 8, when Anthropic researcher Jacob Coxon resigned and warned that the labs are “gambling with our lives.”

CEO Dario Amodei followed on Sept. 12 with an essay urging labs to slow capability gains by one to two years. Sam Altman and Elon Musk backed him.

Then the rebuttals landed. Jensen Huang and Mark Zuckerberg both rejected coordinated pacing at Salesforce’s Dreamforce conference on Sept. 15. “Run as fast as you can,” Huang said, according to CNBC. President Donald Trump had called AI safety concerns “a hoax” a day earlier, Axios reported.

Read the roster and the pattern is hard to miss. Nvidia sells more chips when the industry moves faster. Meta runs an open-source program that prefers light regulation. Anthropic and OpenAI are the two heading towards public markets, and Michael Burry has publicly argued that a slowdown mostly protects incumbents.

Ohanian sells none of it. That’s what makes his read worth checking.

stockcam / Getty Images

Reddit is the ultimate test case

Reddit (RDDT) is the company he co-founded and the clearest public example of the mundane danger he describes. It runs a network of interest-based forums, earns most of its money from advertising, and licenses its archive of human conversation to AI developers.

Reddit posted second-quarter revenue of $805 million, up 61% year over year, with net income of $253 million, according to its Q2 earnings report.

Both figures beat Wall Street estimates. The stock fell anyway after management flagged volatile search referrals.

That is the banal risk working in real time. Google’s AI Overviews answer the question, the user never clicks through, and a business built on search traffic loses readers without any model going rogue.

Alphabet, OpenAI licensing renewals are key

The second exposure is a contract. Google parent Alphabet (GOOGL) pays Reddit about $60 million a year for training access, and OpenAI pays roughly $70 million, with both deals up for renewal, said Baird analyst Colin Sebastian, according to InsiderMonkey.

Wells Fargo estimated in June that renewed terms could push combined licensing revenue toward $550 million a year, StockTwits noted. In July, the same firm warned that walking away from Google would cost Reddit about $500 million and hurt user growth.

DA Davidson trimmed its price target to $185 from $200 on Sept. 18, according to TheFly.

Shares traded near $158 on Sept. 16, down from a record $282.95 set in September 2025, and have shed roughly a third this year. For shareholders, the entire AI story arrives as signed terms rather than a catastrophe.

More AI:

The AI pause talk has not paused anything

Six days after Amodei’s “Pace the Frontier” message that asked the industry to slow down, Reuters reported on Sept. 18 that Anthropic is weighing a new model to counter OpenAI’s GPT-6 Astra, released Sept. 3.

The company has also pushed its IPO to November at a valuation investors discuss near $2 trillion, according to The Wall Street Journal, with annualized revenue topping $65 billion by the end of July, Reuters reported.

Nothing on that calendar looks like a pause. It looks like a product cycle wearing safety vocabulary.

That is Ohanian’s complaint, and it carries a cost. Extinction headlines cannot be priced, so markets discount them and look away. The real AI repricing is happening in traffic logs, renewal negotiations, and enterprise switching costs, one quarter at a time.

Reddit is the test case because both halves of the trade run through one company. Its archive grows more valuable to AI every year.

Its traffic becomes less valuable every time a model answers instead of linking. Watch the renewal terms, because they will say more than any manifesto.

The Arena Media Brands, LLC THESTREET is a registered trademark of TheStreet, Inc.

This story was originally published September 21, 2026 at 6:17 AM.