$2.5 Trillion Wiped Out as Wall Street Panics Over AI Slowdown Fears

Today, Wall Street experienced one of its worst trading days in more than a year. Around $2.5 trillion in market value was wiped from the Nasdaq 100 in a single session, with AI-related companies taking some of the biggest losses.

According to Reuters, the Nasdaq Composite plunged 5.5%, its worst daily performance since April 2025. Nvidia lost more than 9%, AMD dropped over 17%, and the semiconductor index suffered one of the steepest declines in its history.

Here’s how it compares to previous crashes:

The November 5 selloff was significant in dollar terms because today’s technology companies have much larger market capitalizations than they did decades ago.

It’s crazy and an absolute bloodbath.

And this was not limited to AI stocks. The S&P 500 fell 4.4%, while the Dow Jones Industrial Average lost more than 1,800 points. Pretty much the entire market was caught in the selloff.

So, what the hell happened?

The immediate trigger appears to be growing fears that the AI industry could face tighter regulations following the Democrats’ sweeping victory in the US midterm elections.

2118383775654785407

Investors are concerned about restrictions on data-center construction, which could slow down the massive infrastructure expansion supporting today’s AI boom.

A bill that would do exactly that was introduced by Bernie Sanders and Greg Casar in September.

Under the bill:

• Developing or deploying superintelligent AI would be banned
• Advanced AI development would pause until a new federal regulator sets rules and a model review process
• A new cabinet-level AI department would monitor frontier models and oversee removing dangerous capabilities, like ignoring shutdown commands
• Companies that try to get around it would face the “corporate death penalty”
• Individuals would face up to 20 years in prison

According to WSJ, four US states just elected anti-data-center governors which means ~40% of 2027 capacity wiped out overnight.

Tech companies were already struggling to secure enough computing power to meet growing demand for their products. If construction stalled, an already fragile situation could turn dire.

Voters backed Democratic candidates, from Ohio and Michigan to Texas and Georgia, who have been critical of data center construction and pledged to slow or pause their rollout. More than a dozen towns in Ohio voted to halt or restrict new development. Similar bans passed in towns in Kansas, Florida and California — a near universal rejection across dozens of measures that signals growing bipartisan opposition.

OpenAI senior leadership has begun discussing cutting 2027 data center capacity expectations.

There are also concerns about whether companies like Microsoft, Google, Meta, and Amazon are spending too much money on AI infrastructure without generating enough returns to justify their investments.

As someone who has been following AI closely for years and now makes a living working with these tools, I find the whole situation frustrating.

I understand why people are concerned. AI companies are spending enormous amounts of money. Data centers consume electricity and water. There are legitimate questions about jobs, regulation, profitability, and how much of the promised economic growth will actually materialize.

But I also think we are losing sight of how much progress has already happened, and how much more this technology could contribute if we give it room to develop.

I am not arguing that AI companies should be allowed to do whatever they want. Far from it. I just think the benefits of continued AI development outweigh the risks of slowing everything down because of political pressure and market fears.

What actually caused the market crash?

To understand the crash, we need to look beyond the stock prices.

The AI industry has become one of the biggest drivers of investment in the US economy. Technology companies are pouring hundreds of billions of dollars into data centers, GPUs, networking equipment, electricity infrastructure, and AI research.

For years, investors have rewarded companies for announcing increasingly ambitious AI investment plans. Nvidia became one of the world’s most valuable companies largely because its hardware powers much of the AI industry.

But that enthusiasm has started running into some serious obstacles.

First, there is the question of profitability. Investors are increasingly asking how quickly these massive capital expenditures will translate into sustainable revenue and profits.

If you check isAIprofitable, you’d see that not a single tech company has break even yet.

Second, the physical infrastructure needed to run AI is becoming a political issue.

Data centers require enormous amounts of electricity, cooling, land, and access to local power grids. Many communities are worried that the rapid construction of these facilities will increase household utility bills, consume local resources, and provide relatively few permanent jobs.

These concerns played a major role in the November 3 midterm elections.

The Washington Post reported that voters in at least 29 jurisdictions approved measures restricting data-center development. This was especially concerning for investors because Texas has become one of the most attractive locations for AI infrastructure investment.

Restrictions on new data centers could delay the deployment of GPUs, reduce demand for semiconductor equipment, and make it harder for companies to generate returns on investments they have already committed to.

But the restrictions do not apply to currently permitted datacenters in Texas. They will be allowed to finish construction that covers all of the 2027 and likely well into 2028.

The market was effectively pricing in the possibility that AI infrastructure growth might slow much faster than previously expected.

But I think there is a problem with treating every political development as proof that the entire AI industry is heading toward a collapse.

The AI slowdown narrative is getting out of hand

One thing I have noticed about AI discussions online is how quickly people jump from one extreme to another.

When a new model breaks a benchmark or a major technology company announces another billion-dollar investment, everyone starts talking about an unstoppable AI revolution.

Then the market crashes, and suddenly the conversation turns into AI being overhyped, useless, or nothing more than an expensive bubble.

Both extremes are wrong.

There is certainly hype surrounding AI. I have written about AI tools long enough to know that many products are simply existing software with a chatbot. Some startups have questionable business models, and many companies use AI as a marketing term without delivering much practical value.

But dismissing the entire industry because some companies are overvalued is also ridiculous. A falling stock price does not mean the underlying technology suddenly stopped working.

Nvidia dropping 9% does not make its GPUs less capable. AMD losing 17% does not mean companies no longer need computing infrastructure. And a slowdown in data-center construction does not mean millions of people will suddenly stop using AI assistants.

What the market is questioning is how much future growth should already be reflected in today’s stock prices.

That is a legitimate financial question. It is not the same as questioning the usefulness of AI itself.

We have seen something similar with earlier technological developments.

The dot-com crash destroyed enormous amounts of market value, and many internet companies disappeared. But the internet itself continued growing and eventually became essential to the global economy.

I am not saying AI will follow exactly the same path. History does not guarantee that every new technology will succeed commercially.

My point is that speculative investments can fail while the technology behind them continues improving.

I think there is a real possibility that some AI companies will lose badly over the next few years, even as AI becomes more useful and widely adopted.

Those two outcomes are not contradictory.

People have legitimate reasons to oppose data centers

Before going further, I want to address one argument from the public backlash that I think deserves serious attention.

People are worried about the impact of AI infrastructure on their communities, and they have every right to be.

Personally, I am not in favor of the planned datacenter in the Philippines.

In This Economy Pax Silica What Philippines Trump Card

Imagine living in a small town where a technology company announces plans to build a massive data center. The project promises investment and economic growth, but residents are also told that it could consume substantial amounts of electricity and water.

Naturally, people will ask who benefits from that arrangement.

Will the project create enough local jobs? Will the company pay for the necessary electricity infrastructure? Will household utility bills increase? What happens if the company eventually abandons the facility?

These are not unreasonable questions. Read some really good points of the disadvantages brought about by datacenters.

Pax Silica Brings Promise But At What Cost

According to the International Energy Agency, data centers consumed approximately 485 terawatt-hours of electricity worldwide in 2025. That figure could nearly double to 950 terawatt-hours by 2030.

And although data centers would still represent only around 3% of global electricity consumption by 2030, their impact can be much greater in communities where large facilities are concentrated.

So, should we build datacenters in space instead?

There is also evidence that public opposition is much more widespread than some AI companies expected.

An NBC News and SSRS exit poll found that 76% of voters opposed new data-center construction in their own communities, according to Reuters’ reporting on the Texas election.

That is a remarkable number.

It shows that opposition to AI infrastructure is not simply coming from a small group of people who dislike technology.

Companies should also be transparent about water consumption, environmental effects, employment opportunities, and the long-term benefits promised to the surrounding community.

I support stricter requirements in those areas.

Where I disagree is the idea that restricting or banning data-center development should automatically be considered the best solution.

Slowing AI development comes with costs, too

One argument I find frustrating is the assumption that slowing down AI development is inherently good because it reduces risk.

Of course, a slower pace of construction can reduce immediate pressure on electricity grids and other local resources.

But we also need to consider what we give up by delaying the development and deployment of AI systems.

AI is already helping people write software, conduct research, analyze information, translate languages, and automate repetitive work.

And those are just the applications we can observe today.

There are also potentially enormous benefits in medicine, scientific research, education, manufacturing, and other industries where AI adoption is still developing.

I spend a significant amount of time testing AI models and writing about how people can use them. Over the past few years, I have seen coding assistants progress from generating short snippets to helping developers build entire applications.

Research assistants are getting better at handling complex questions, processing long documents, and working across different sources. A few months ago, Sir Demis Hassabis launched Isomorphic Labs which is an AI company that aims to solve all diseases.

https://www.isomorphiclabs.com

All these could change all our lives, how we work, and how we live.

The Stanford 2026 AI Index Report found that 88% of surveyed organizations reported using AI in at least one business function in 2025.

The same report reviewed research showing productivity improvements of approximately 14% to 15% in customer support, 26% in software development, and 50% in marketing output.

These figures do not mean every worker becomes that much more productive simply by using a chatbot. The results vary by task, experience, implementation, and how the benefits are measured.

Still, they provide substantial evidence that AI is already creating economic value.

The report also estimated that generative AI tools delivered around $172 billion in annual consumer value in the United States by early 2026.

That figure measures estimated benefits to users, not revenue collected by AI companies. A technology can provide enormous value to society before companies figure out the best way to monetize it.

The question should be how to make the technology more accessible, affordable, and useful while managing the costs associated with its development.

Simply making it harder to build the necessary infrastructure does not solve that problem.

But are AI companies spending too much money?

Now, this is where I think the critics have one of their strongest arguments.

The amount of money being invested in AI infrastructure is extraordinary.

According to Reuters, analysts at J.P. Morgan projected approximately $900 billion in global AI investment for 2026 and $5.5 trillion by 2030.

And let’s not forget OpenAI’s $500B project Stargate which Elon Musk criticized.

Those are staggering numbers, and investors have good reasons to question how all of that spending will generate sufficient returns.

Building AI infrastructure is expensive. GPUs are not cheap, data centers take years to construct, and the hardware needs to be replaced or upgraded as newer generations become available.

The costs do not end when construction is completed.

Companies still need to pay for electricity, maintenance, networking, cooling, engineering, and ongoing model development.

And because AI capabilities are improving so quickly, there is a risk that some expensive infrastructure investments could become less economically attractive before companies recover their costs.

It is also possible that AI becomes dramatically cheaper to run, which would benefit users but put pressure on providers that invested heavily based on expectations of high prices.

This is one reason I do not agree with the argument that every AI stock is automatically a good long-term investment.

Some companies will likely overbuild. Others may struggle to find enough paying customers. Smaller infrastructure providers that rely heavily on debt could face serious financial problems if demand grows more slowly than expected.

We should also distinguish between the companies building AI infrastructure and those using it to sell profitable services.

For example, Alphabet’s third-quarter 2026 earnings showed consolidated revenue of $127.6 billion, up 25% year over year. Google Cloud revenue more than doubled to $30.7 billion, supported by enterprise AI infrastructure and services, alongside its other cloud products.

But it does not prove every dollar spent on AI infrastructure will produce an attractive return.

Alphabet itself expects enormous ongoing capital expenditures, and other technology companies face similar financial pressures.

The market is right to demand greater accountability.

What happens next?

I do not think anyone can confidently predict how quickly AI-related stocks will recover from this crash. There are too many moving parts.

Investors will be watching how new political leaders approach data-center regulation, particularly in states that have attracted significant infrastructure investment.

They will also be paying close attention to the next rounds of earnings reports from major technology companies.

The most important question will be how much revenue and profit those companies can generate relative to their enormous AI investments.

And the answer will be different for each company.

Some AI infrastructure providers may struggle with debt and delayed construction. Others could benefit from continued demand for computing resources, even if the overall pace of expansion slows.

Companies that provide AI applications may also perform differently from those selling hardware or data-center capacity.

I think we will see investors become more selective about the AI businesses they are willing to finance.

There is one additional possibility.

If access to computing infrastructure becomes more constrained, companies may have stronger incentives to make AI models more efficient rather than relying entirely on ever-larger computing budgets.

We have already seen that model efficiency can improve alongside capability.

Of course, better efficiency is not a complete replacement for new infrastructure. Demand for AI computing could continue growing even as individual models become cheaper to run.

But I suspect the next phase of AI development will place much greater emphasis on delivering useful capabilities at a reasonable cost.

That could ultimately benefit users and businesses more than the current obsession with announcing bigger infrastructure spending plans.

Final Thoughts

I have mixed feelings about this market crash.

I understand the concerns surrounding AI investment. Some companies are spending extraordinary amounts of money, and there are legit concerns about how quickly those investments will pay off.

I also think communities deserve stronger protections against rising electricity costs, excessive water consumption, and irresponsible infrastructure development.

But I am frustrated by how quickly the conversation has turned into another argument that AI development needs to slow down or even stop.

There are researchers using AI to help solve difficult problems, developers building applications faster, small businesses automating repetitive work, and individuals learning skills that previously required expensive courses or specialized assistance.

I do not think those benefits should be dismissed simply because investors are suddenly worried about infrastructure spending.

We should hold companies accountable, protect communities, enforce sensible regulations, and demand stronger evidence of commercial returns.

At the same time, we should recognize that AI has already created substantial value, and there is still a lot of room for improvement.

I am not interested in defending every company that puts AI in its name. I am interested in seeing how far this technology can go and how many people can benefit from it.

That is why, despite the risks and the uncertainty, I still believe the benefits of continued AI development outweigh the costs of slowing it down.

The industry needs accountability, not panic.