Asian tech stocks are cratering this morning as investors finally blink at the eye-watering cost of building AI infrastructure. SoftBank Group led the plunge, dragging down chipmakers and cloud providers in a selloff that’s raising urgent questions about whether the AI boom’s economics actually work. The rout comes as companies burn through billions on GPUs and data centers with little clarity on when they’ll see returns.

SoftBank Group is bleeding red this morning, and it’s taking the entire Asian tech sector down with it. The Japanese conglomerate’s shares tumbled in Friday trading, sparking a cascade that’s hitting everyone from chipmakers to cloud infrastructure providers. The culprit? A sudden crisis of confidence about whether pouring hundreds of billions into AI infrastructure makes any financial sense.

The panic isn’t coming out of nowhere. Companies have been on an unprecedented spending spree, snapping up Nvidia GPUs like they’re going out of style and building data centers at a pace that would make previous tech booms look quaint. But investors are starting to ask the uncomfortable question: where’s the payoff?

SoftBank’s portfolio is particularly exposed to this reckoning. The company has been one of the most aggressive investors in AI startups and infrastructure plays, betting big that the technology would justify whatever price tag came with it. That conviction is getting tested right now. When a bellwether like SoftBank stumbles, it sends a clear signal that the easy money phase of the AI boom might be over.

The infrastructure cost problem is real and getting worse. Building the computing power to train and run large language models requires massive capital expenditure upfront. We’re talking about specialized chips that cost tens of thousands of dollars each, deployed in facilities that need industrial-scale power and cooling. Microsoft, Google, Amazon, and Meta have all signaled they’re spending more than initially planned, but revenue from AI products hasn’t scaled at the same pace.

This creates a dangerous dynamic. Companies can’t afford to stop spending because falling behind in AI capability could be existential. But continuing to spend at this rate without clear monetization paths is straining even the biggest balance sheets. It’s a classic arms race scenario where everyone keeps buying weapons they’re not sure how to use.

The selloff is hitting chip stocks particularly hard. If demand for AI infrastructure is about to cool because companies are tapping the brakes on spending, that’s terrible news for semiconductor makers who’ve been riding the AI wave. Supply chains that ramped up production to meet seemingly infinite demand could suddenly find themselves oversupplied.

What makes this especially worrying is the timing. We’re not in the early experimental phase of AI anymore. Companies have had time to figure out business models and revenue streams. If they’re still struggling to justify the infrastructure costs now, that suggests the economics might be fundamentally harder than the hype implied. Building AI capabilities is turning out to be more expensive and monetizing them more difficult than many investors bet on.

The contagion risk is significant. Asian markets often serve as a early warning system for global tech sentiment. If this selloff reflects a genuine reassessment of AI infrastructure economics rather than just regional jitters, US tech stocks could face similar pressure when markets open. The mega-cap tech companies that have led market gains on AI enthusiasm would be particularly vulnerable.

SoftBank’s decline also raises questions about the broader AI startup ecosystem. The company has been a major source of capital for emerging AI companies, many of which are burning cash to build infrastructure and acquire users. If SoftBank and similar investors start pulling back because the infrastructure cost equation doesn’t work, that’s going to create funding pressure throughout the sector.

The market is essentially calling for proof of concept at scale. It’s no longer enough to show that AI technology works or that it has potential applications. Investors want to see that the revenue generated can actually cover the staggering infrastructure costs and deliver returns that justify the investment. That’s a much higher bar, and it’s not clear how many companies can clear it right now.

This selloff represents more than just a bad trading day. It’s a fundamental challenge to the narrative that’s driven tech valuations for the past year. The AI infrastructure build-out has been predicated on the assumption that massive upfront costs would be justified by transformative revenue opportunities. But if investors are losing faith in that equation, we could be entering a period where AI companies need to prove profitability rather than just promise it. The next few weeks will reveal whether this is a temporary panic or the start of a serious recalibration in how the market values AI investments. Either way, the era of spending whatever it takes on AI infrastructure without hard questions about returns appears to be ending.