The search giant now says capital expenditure could reach $205 billion. Three months ago, the ceiling it gave was $190 billion. Even the floor of the fresh range — $195 billion — clears the number the company had recently described as its upper limit.
That shift, not the headline figure, is what deserves a second look from shareholders. The movement matters more than the magnitude.
It’s easy to wave off $15 billion, and plenty of observers did exactly that. Set the scale aside, though, and the message is that Google cannot forecast its own costs with any precision — and businesses in that position have a habit of delivering more surprises down the road.
The math doesn't close
More money is going out than coming in. That’s the entire problem in one line, and spotting it requires no background in finance.
Layer the pricing dynamic on top of that. Chinese AI tools are applying competitive pressure, and there’s downward pressure on what Google can charge for its models. Costs rise while prices hold steady or slide.
Spend more and take in the same amount. Or spend more and take in less. There’s no happy ending in either scenario.
Three other things went wrong at once
The market appears to have caught on to SpaceX’s problems. At the time of writing, the stock trades at roughly half its peak value.
There’s also unease around the debt behind Oracle’s datacenter expansion — a bigger deal than it first appears. Public investors treat Oracle as their OpenAI substitute. With no way to short a private company, traders punish the proxy instead.
Then there’s Nvidia, which has been working through deal negotiations adding up to three-quarters of a trillion dollars. The chipmaker, not OpenAI, is the true hub of the circular financing at play. A supplier repeatedly bankrolling the customers who buy from it isn’t necessarily a show of strength.
What a $250 billion guarantee actually signals
The arrangement in which Nvidia backstops OpenAI’s debt carries a $250 billion price tag. Call it evidence of demand if you want. Professionals who track this sector for a living see something else.
The deal is “as much a reminder of funding strain in the AI build-out as it is a demand signal,” Billy Leung, tech sector investment strategist at Global X Management, said to Bloomberg.
Strain on funding — not demand. That’s a strategist whose firm has capital deployed in the sector saying the uncomfortable part in public.
The China problem nobody has solved
Each time a Chinese start-up ships a new model, the jitters return. There’s a concrete reason for that reaction.
The central bottleneck for China is that, in theory at least, its firms can’t get GPUs on the same terms US firms can. Their AI systems keep pace regardless.
Pull on that thread and it leads somewhere Nvidia won’t enjoy. Should competitive models turn out not to need the GPU volumes the market took for granted, the chipmakers’ windfall has a visible expiry date. It could also mean the industry is putting up far more data centers than it needs.
Even the optimists expect an overbuild
Many of the sharp people I’ve spoken with are considerably more bullish than I am. For three years I’ve been asking how AI companies intend to generate actual profits, and a satisfying answer still hasn’t arrived.
Push those bulls and here’s the response. Data centers will probably get overbuilt during this bout of exuberance. When the correction arrives, plenty of AI companies won’t survive it.
They’ve put money in regardless, betting that whoever makes it through will cover the cost of everyone who didn’t. That is, genuinely, the bull case — one with a heap of casualties baked into it.
Watch the earnings calls this week
Microsoft, Amazon and Meta all deliver results in the coming days. Many expect each of them to disclose data center outlays running above forecasts as well.
None of this is unique to Google. The same forces run across the whole AI sector, which explains why a single revision to guidance shifted the discussion this far.
Even the cheerleaders are scanning for the top, same as everyone else. They understand it’s on the way. Timing it ahead of schedule is beyond anyone, and some investors are visibly losing their nerve and rotating capital elsewhere.
The remaining giants might post figures that settle nerves, and this bout of AI anxiety fades. If it’s top signals you’re after, Elon Musk is a dependable one — and SpaceX just went public.