Many tech companies are licking their wounds after a massive stock market slump. The sharp drop in market values stems both from doubts about the long-term AI buildout and from new uncertainty over political support for data centers in the US. Who lost the most, and why? And why did the biggest AI investors of all get off so lightly?
Looking at Thursday’s closing prices, the biggest losers in absolute terms are led by Nvidia, by a wide margin. TSMC, Tesla, SpaceX and Broadcom follow, each losing around $160 billion or more in market value. Remarkably, the damage at Apple, Alphabet and Amazon was relatively limited: each lost less than 3 percent. In percentage terms, the hardest hit among the big names was SanDisk (-22.2 percent). Micron (-17.5 percent), SK hynix (-9 percent) and Samsung (down more than 6 percent on both Thursday and Friday) also had to give back a good chunk of their gains. The biggest winners of the current memory shortage are, for this week at least, the biggest losers.
Losses ripple through
Those losses are anything but isolated: the tech-heavy Nasdaq shed around $2.5 trillion in market value in a single day. Of course, these aren’t physical dollars vanishing from investors’ wallets, but for riskier bets on continued growth, the impact is devastating. AMD (-17.3 percent), TSMC (-10.8 percent), Broadcom (-10.4 percent), ASML (-9.3 percent on the Nasdaq; it fell 6.5 percent in Amsterdam on Thursday) and Cisco (-7.5 percent) show that every link in the AI buildout’s supply chain is feeling the pain.
To return to Nvidia: its 9.3 percent drop amounts to $448 billion in market value. That wiped out the stock’s entire 2026 gain in one go; it is now trading below its final close of 2025. It happened on a day when the Nasdaq had its worst session since April 2025.
SpaceX (-9.1 percent) and Tesla (-11 percent) appear to be suffering from a particular ‘Musk effect’. Musk spent hundreds of millions of dollars in a failed attempt to get more Republicans elected, and investors fear his companies will face closer scrutiny under a Democratic Congress. Both companies are also based in Texas, where Democratic governor-elect Gina Hinojosa wants a moratorium on new data centers. With her as governor and Democratic majorities in both the House and the Senate, it will be harder for Musk to pursue his AI ambitions at the same pace. Incidentally, opposition to data centers isn’t a purely Democratic affair: sitting Republican governor Greg Abbott already halted permits for new data centers in September, pending an audit.
That’s not to say the AI buildout enjoyed the market’s full confidence before this. Big Tech has poured hundreds of billions into data centers that largely still need to be built or don’t yet have enough power. Then there’s the question of how you actually depreciate an Nvidia GPU or Superchip. Alphabet uses six years for its servers, Meta five and a half. Meta extended that period only last year, which added more than $2.5 billion to its 2025 net income. Amazon went the other way, shortening the period for some of its servers to five years on the grounds that AI hardware ages faster. With a new generation of Nvidia chips arriving every year, that’s not a strange thought. Investor Michael Burry calculated last year that the hyperscalers are understating their depreciation through 2028 by some $176 billion. Still, older hardware doesn’t suddenly become worthless, as IREN shows. The AI cloud provider is renting out its now elderly H100 chips under new contracts at double what it charged before, simply because there still isn’t enough compute to go around. But if power becomes the real bottleneck, you want every megawatt to do as much work as possible. Then you pick the most efficient chips, and that old H100 suddenly looks a lot less appealing.
Not panic everywhere
It should be said that the negative sentiment is limited in some crucial areas. Take the hyperscalers themselves, of all companies the biggest drivers of the physical AI buildout. Microsoft, Alphabet and Amazon rose 3 to 7 percent on Wednesday, leaving them slightly up overall since the election result, despite Thursday’s decline. Without the tech losses around them, you wouldn’t even call these notable moves. The investor logic: a slower buildout reins in these companies’ enormous spending, while they already operate large data center networks. In other words, the market is punishing the suppliers of the AI buildout and sparing the buyers. Cloudflare also escaped. After falling more than 8 percent during regular trading, the stock jumped nearly 15 percent after hours on strong quarterly results.
Lessons from the SaaSpocalypse
We saw a hefty correction earlier this year too, but on a completely different side of the IT world. Think back to the ‘SaaSpocalypse’ in the first half of 2026. Back then, the software giants seemed to be punished and marked down to significantly lower valuations, because AI could supposedly take over the functionality of entire enterprise applications. That, at least, was the sentiment the deep-red share prices conveyed.
The lesson seems to be that a correction of apocalyptic proportions only partly sticks. ServiceNow is still 33 percent below where it started 2026, and Intuit (-46 percent) and HubSpot (-49 percent) fared even worse over the same period. Atlassian (+13 percent), Workday (-7 percent) and Salesforce (-16 percent), on the other hand, have recovered all or much of their losses. It all depends on where you measure: Salesforce is 52 percent above its June 22 low, Workday 81 percent above its April 9 low, and Atlassian a dizzying 227 percent above its April 10 low. That said, Atlassian took another hit of around 16 percent in after-hours trading on Thursday evening following its own quarterly results.
On average, software stocks even won back around 94 percent of their losses, according to Hargreaves Lansdown. The laggards are precisely the companies whose products investors fear AI will replace. That is the key difference with now: the chipmakers falling this week are still seeing strong revenue growth. The doubt concerns the pace of the buildout, not the demand itself.
Conclusion: wait and see
Even the most celebrated market analysts can’t reliably predict future market values. Estimating the value of the AI boom is a monumental task, let alone predicting who the big winners will be. AI labs are posting fast-growing revenue, but are investing far more in compute, hoping for growth without precedent. The infrastructure around them rests on the assumption that the hardware, networks and applications of this technology will generate enough value.
However promising that sounds, the US elections were yet another reminder of the real world. Where do you put your data centers? Where do you get the energy? How much can you ask of local residents in terms of nuisance and energy prices, and should you even be asking? The answers will come through political debate, public discussion and, eventually, regulation that seems inevitable. That brake could also put an end to the unprecedented rise in valuations, with potentially serious consequences for sentiment around the wider global economy.
There are some markers to watch, though. Nvidia reports its quarterly results on November 17, and Hinojosa takes office in January. According to J.P. Morgan, it’s mainly capacity planned for 2028 that is at risk, since three quarters of what is due to come online in 2027 is already under construction. In that light, a correction is only logical. Whether it stops at this one blow, we’ll only find out in the coming months.