As the doomsday clock ticks 85 seconds to midnight, we’re viewing the AI sell-off as a kind of countdown of its own… and investors are reading it backwards
Listen to the audio version of this article (generated by artificial intelligence).
On a warm day in mid-July, about 80 Nobel laureate scientists and nuclear security experts gathered in a 10th-floor conference room at the University of Chicago. They were then asked to imagine their death…
An announcer directed the group’s attention out the window, past the campus’s gothic towers, and traced the neighborhoods that could disappear from nuclear explosions of varying sizes. Exercise, which was recently dated Popular Mechanicsis part of the process behind one of the most famous codes on Earth: Doomsday Clock.
When artist Martel Langsdorff depicted it on the first cover of the Bulletin of the Atomic Scientists in 1947, it set the hand at seven minutes to midnight for no scientific reason at all… the position was simply “Fits my eyes“The hand has always been a call to judgment.
Last January, the publication’s board moved it to 85 seconds to midnight, the closest in its history, citing nuclear arsenals, the climate, and the unlimited rise of unregulated artificial intelligence.
It turns out that Wall Street keeps its own AI clock. And every sell-out like the one we just experienced is the masses grasping the hand and holding it back… a collective judgment that the promise of technology to change the world lies further away than feared, that disruption is overblown, and that midnight is drawing down.
Of course, two-hour servings vary by size. But the mechanism is the same. In the last episode of Being exponential With Luke LangoWe read the machine, and we find it working faster than ever before.
Watch the full episode here. Also check that Subscribe to Being exponential On X (Formerly Twitter) For more exclusive content:
What went wrong in the sales process?
Some blamed it Broadcom Company (Afgo) To launch it. However, we don’t buy this story. Broadcom had nearly 50% revenue growth, nearly 80% semiconductor revenue growth, more than 140% AI semiconductor revenue growth, and a $30 billion backlog…records across the board. As we say in the episode, there is no fundamental weakness in that report.
And the spending headlines continue to pile up. China has reportedly allocated nearly $300 billion over five years to a national network of AI data centers. Nepheus Nevada Group (NBIS) Investing €1.7 billion to build capacity in the UK Advanced Micro Devices Company (AMD) It has just announced plans to invest up to £2bn there over the same period. SK Telecom plans to build a gigawatt-sized AI cloud in South Korea. OpenAI It just raised $122 billion and filed for an IPO. SpaceX It is set to raise approximately $75 billion in its private offerings.
So what really scared the market? We look at three real risks: escalation with Iran sending oil past $110 to $120 and reigniting inflation, policy shifts that shouldn’t matter until 2028, and a creeping sense that the market has become too cheerful. in The full episodewe learn why each one, upon closer examination, seems more manageable than the tape suggests.
The game every company wants
Then there’s the viral story that… Uber (Uber) blew through her Anthropic The symbolic budget… has been hung in some corners as evidence that AI is not paying off. Our read turns that on its head: Buy a child a new toy, and of course they play with it every waking hour until you set some boundaries. Limits do not mean the game was a mistake.
Companies are moving from token cap to token budget – and the budget line is exactly what the institutionalization process looks like. Read conference calls and reports company after company is improving AI for operations across software, hardware and consumer businesses alike.
$5 Trillion Flood Gate
The centerpiece of the episode is the wave of “kilicorn” IPOs — SpaceX approaching $1.75 trillion, OpenAI and Anthropic heading toward roughly $1.5 trillion. Consider $5 trillion of new market capitalization hitting the public markets in one year. Many investors read that as a top signal.
History reads it differently.
The giant IPOs of the dot-com era came in 1998 and 1999, and smaller companies poured through the gates after that. Big bulls open the gates first. There’s also one development in particular that could bolster one corner of this trade: reports that the White House is considering direct stakes in frontier AI labs, which we interpret as an OpenAI story… and a bullish story for pre-IPO vehicles like Soro Capital Corp (Ssss).
Where does all the new IPO capital go? Right back to computing…which means more networking, more memory, more chips, more cooling, and more power.
Mirage jobs report
One last dissenting call. Last week’s strong jobs report – which added more than 170,000 jobs – has revived claims that the AI employment apocalypse has been exaggerated. We leaf through Challenger, Gray, and Christmas data that tells a different story: Nearly 100,000 jobs were cut in May, the largest May number since 2020, with AI cited in about 40% of them. AI-driven reductions have already exceeded 88,000 this year – about 60% more than in all of 2025, just five months later.
In the full episode, we have identified the specific accumulation area we are watching on Van Eek semiconductor box (Name it)And why we think the summer of AI will resume once SpaceX’s IPO is over, and the only scenario that will actually change our minds.
Watch the latest episode of Being Exponential with Luc Lango here. And make sure of that Subscribe to Being exponential On X (formerly Twitter) for more exclusive content.
