THE AI PARADOX : Yes it's a bubble. And Yes, it will still change everything

Sash Mohapatra

Sash spent 20 years at Microsoft guiding enterprise clients through the cloud revolution and the rise of AI. Now, as the founder of The Rift, he’s on a mission to enhance human potential by helping people develop practical, future-ready AI skills. He writes from a place of deep curiosity, exploring what it means to stay human as machines reshape the world around us.

December 3, 2025

I remember the sound of a dial-up modem at 2 AM. That screeching handshake with the future - 56k of pure possibility. I was seventeen, teaching myself HTML in Notepad, building websites that maybe fifty people would ever see, chatting with someone in Germany about nothing important except that we could. The internet was magic to me. Pure, uncomplicated magic.

I had no idea that while I was discovering I could put my thoughts on a screen that anyone in the world could read, the adults were busy destroying five trillion dollars in market value. Pets.com, Webvan, hundreds of companies with .com in their names-gone. I barely noticed. The internet was still magic. Still changing everything.

Twenty-five years later, I'm watching AI unfold with a strange double vision. I still feel that teenage wonder-I use these tools every day, I see what they enable, I watch high-agency people build things they couldn't have built before. But now I understand the economics. I read the earnings reports, I track the valuations, I see the patterns forming.

And this time, the patterns scare me in ways the dot-com bubble never did.

Subscribe now to unlock the full article and gain unlimited access to all premium content.

And this time, the patterns scare me in ways the dot-com bubble never did.

The Valuation Bubble Is Real (And That's the Least Scary Part)

Let's start with what's obvious: the numbers have lost touch with reality.

OpenAI went from a $1 billion valuation in 2019 to $500 billion by November 2025-a 500x increase in six years. For twelve months straight, OpenAI's valuation increased by roughly $29 billion every month-almost $1 billion per day. Anthropic nearly tripled its valuation from $61.5 billion to $183 billion in just six months.

These aren't numbers tethered to profitability. OpenAI expects to lose $9 billion in 2025 on approximately $13 billion in revenue. They spend approximately $1.69 for every dollar of revenue they generate.

Nvidia-actually making money-became the first publicly traded company to exceed $5 trillion in market value in October 2025. The company added $1 trillion in market cap in just four months.

Is this a bubble? Yes. Obviously, yes.

Amazon, Microsoft, Alphabet, and Meta are collectively spending approximately $382 billion on AI infrastructure in 2025-up from around $233 billion combined in 2024. According to tech analyst Ed Zitron, these companies have invested roughly $560 billion in AI infrastructure over two years while generating just $35 billion in AI-related revenue combined. That's a 16:1 spending-to-revenue ratio.

I watch these numbers and I think about teenage-me, blissfully unaware that the internet companies were hemorrhaging cash. Adult-me knows this matters. Adult-me has pattern recognition.

But here's what keeps me up at night: this isn't just a valuation bubble anymore. It's become a dependency we can't escape.

The Part That Actually Scares Me: We've Bet Everything

In 1999, if Amazon failed, life went on. The internet would have survived. Other companies would have emerged. It was a private sector bet that went sideways.

In 2025, we've done something far more dangerous: we've made global economic growth structurally dependent on AI succeeding. Not eventually-now.

The numbers are stark: Investment in information processing equipment and software is only 4% of GDP, but it was responsible for 92% of U.S. GDP growth in the first half of 2025. GDP excluding these categories grew at a 0.1% annual rate in H1. Point-one percent.

Let that sink in. Without AI, U.S. GDP growth would have been 0.8% instead of 1.6%-potentially recession territory. Deutsche Bank reports that private business investment that doesn't include AI has stayed mostly flat since 2019. As one Bank of America economist put it: "It's the only source of investment right now."

We're not talking about speculative venture capital anymore. We're talking about the entire American economy running on the assumption that AI data centers will keep getting built, keep getting funded, keep driving growth. JP Morgan notes that AI-related capital expenditures contributed 1.1% to GDP growth, outpacing the U.S. consumer as an engine of expansion for the first time ever.

I remember when consumer spending was two-thirds of GDP and the reliable engine of growth. Now? We're betting on server farms in Texas.

When Nations Go All-In

And it's not just the U.S. This has become a nation-state competition, with countries betting their economic futures on AI supremacy.

Four days after Trump took office in January 2025, he stood in the White House with Sam Altman, Masayoshi Son, and Larry Ellison to announce Stargate. The project plans to invest $500 billion in AI infrastructure by 2029, with an immediate deployment of $100 billion. Trump called it "the largest AI infrastructure project in history." He indicated he would use emergency declarations to expedite development, particularly regarding energy infrastructure.

By September, Stargate had announced nearly 7 gigawatts of planned capacity and over $400 billion in investment over the next three years. This isn't a corporate initiative with government cheerleading-the White House has signaled support through executive orders, regulatory frameworks, and technology partnerships, making it part of its broader AI policy agenda.

Meanwhile, China saw the starting gun and sprinted ahead.

In 2017, China's State Council released its Artificial Intelligence Development Plan, outlining its ambition to build a 1 trillion yuan AI-powered economy by 2030. The document described AI as a "focus of international competition". When DeepSeek's R1 model shocked Silicon Valley in January 2025, the Bank of China responded three days later with an AI Industry Development Action Plan, pledging 1 trillion yuan ($137 billion) over the next five years.

China treats AI as a tool of national strategic importance, embedded within broader initiatives like "Made in China 2025" and "Scientific Innovation 2030". Over 50 state-backed Chinese companies are developing large-scale AI models. State-owned enterprises have been instructed by government authorities to integrate AI into logistics, HR, and manufacturing systems.

This isn't the dot-com era anymore. This is two superpowers in an AI arms race, each betting that dominance in artificial intelligence determines economic and military supremacy for the next century. As Chinese President Xi Jinping has stated: "Whoever can grasp the opportunities of new economic development such as big data and artificial intelligence will..." The quote trails off, but the implication is clear: whoever wins AI, wins everything.

When I was a teenager playing with HTML, governments weren't betting national security on the internet. Now they're building their entire economic strategies around AI-before we even know if the productivity gains will materialize at scale.

That's what scares me. Not the bubble. The dependency.

But Here's Why I'm Still Optimistic

And yet, despite all of this, despite the hubris and the concentration and the geopolitical risk, I still feel that teenage wonder when I use these tools.

Because the technology is genuinely transformative. Not "might be someday" transformative. Transformative right now, if you're paying attention.

Anthropic's research, analyzing 100,000 real conversations with Claude, found that AI reduces task completion time by approximately 80%. Tasks like curriculum development that would take 4.5 hours can be completed in 11 minutes. Document writing sees 87% time savings. Financial analysis tasks save 80% of time.

GitHub's controlled study found that developers using Copilot completed tasks 55% faster than those without. McKinsey estimates AI could reduce software engineering costs by 20-45%.

I'm living this. I built Pollzy, a Mentimeter clone-in two hours. Two hours. Back in 1999, it was just not possible for me to build this. MVP development costs have collapsed from $50,000+ to under $5,000 using AI-first development stacks-an 80-90% reduction.

This isn't vaporware. This isn't promises. This is happening now, to anyone who chooses to use it.

Goldman Sachs projects that generative AI could drive a $7 trillion increase in global GDP (roughly 7%) and lift productivity growth by 1.5 percentage points annually over a 10-year period. The Penn Wharton Budget Model estimates that AI will increase productivity and GDP by 1.5% by 2035, nearly 3% by 2055, and 3.7% by 2075.

The deflationary impact is real. The U.S. Bureau of Labor Statistics documented that computer and software product prices fell by roughly 74% from 1997 to 2022. AI is accelerating this trend. The International Monetary Fund noted that "AI is our best chance at relaxing the supply-side constraints that have contributed to slowing growth, new inflationary pressures, and rising costs of capital". McKinsey's analysis suggests AI efficiency gains could translate into an estimated 0.5 to 0.7 percentage-point annual drag on CPI.

I see this empowering high-agency people to do things they couldn't do before. I see science advancing. I see genuine productivity gains-not projections, but measurable improvements happening now.

So yes, I'm optimistic about the technology itself. The teenage wonder isn't gone. It's just complicated now by adult understanding.

The Uncomfortable Middle Ground

Here's what I'm sitting with: the valuations are insane, and a correction is coming. Companies will fail. Investors will lose fortunes. MIT research shows that 95% of generative AI pilots at companies are failing. Gartner predicts that 30% of generative AI projects will be abandoned due to poor data quality and unclear value. Even Alphabet CEO Sundar Pichai acknowledges "elements of irrationality" in current AI investing.

The parallels to 2000 are real. The Nasdaq peaked at 5,048 on March 10, 2000, then fell 77% by October 2002, wiping out more than $5 trillion in market value. Pets.com, Webvan, eToys-gone. Cisco lost 90% of its value within a year.

But here's what often gets forgotten: telecommunications companies had laid more than 80 million miles of fiber optic cable across the U.S. When the bubble burst, much of that capacity sat unused. But it formed the backbone for everything that followed-streaming video, cloud computing, social media, e-commerce.

Amazon's stock plummeted from $107 to $7 per share during the crash. But unlike Pets.com, Amazon had a viable business model. Today it's one of the world's most valuable companies.

The infrastructure will survive. The $400+ billion flowing into AI data centers annually will leave behind computational capacity that entrepreneurs will exploit for decades. The correction won't erase that.

But here's what's different-and what keeps me up at night: we've made this correction potentially catastrophic by building economic dependency before the transformation fully arrives.

Financial comparison between Dot-com boom and the AI era

The productivity gains are real, but they're not showing up in broad economic growth yet. The GDP numbers are from building data centers, not from AI making the economy 10x more efficient. We're in this uncomfortable middle period where:

Global growth depends on AI investment continuing
But the revenue to justify that investment hasn't materialized
And nation-states can't back out because they've made it a matter of strategic survival
So everyone has to keep believing, keep investing, keep building

I think about regular businesses trying to navigate this. Do you invest in AI transformation now, knowing a correction is coming? Do you wait, knowing that might leave you behind? How do you make strategic decisions when the entire sector is simultaneously experiencing a valuation bubble and genuine technological transformation?

What Happens Next?

I wish I had answers. Teenage-me would have ignored all of this and kept building. Adult-me can't ignore the macro implications.

A few things I believe:

The correction will come. Valuations will reset. Companies will fail. The gap between $560 billion in investment and $35 billion in revenue cannot persist indefinitely. Some AI startups will go to zero. Some investors will lose everything. The MIT finding that 95% of enterprise AI pilots are failing suggests the reckoning might come sooner than the bulls expect.

The technology will survive and thrive. Just as fiber optic cables survived to enable YouTube and Netflix, the AI infrastructure being built today will power applications we haven't imagined yet. The productivity gains are real. The deflationary impact is measurable. High-agency people are building remarkable things.

The biggest winners might surprise us. Cisco was the dot-com era's Nvidia-the "picks and shovels" play that seemed safest. Yet Cisco's stock still hasn't recovered to its 2000 peak twenty-five years later. The companies that ultimately captured the internet's value-Google, Amazon, Facebook-barely existed during the bubble or emerged from its ashes. The ultimate AI winners may similarly surprise us.

The dependency is the real risk. This is what's genuinely new. We've built a structure where global economic growth requires AI investment to continue at unprecedented levels-before the productivity gains have shown up in GDP. If the correction comes hard enough, fast enough, we might find out that we've created a house of cards that takes the broader economy down with it.

I watch the concentration of wealth and resources-92% of U.S. GDP growth coming from 4% of the economy. I watch nation-states betting their futures on AI supremacy, driving an arms race in defense AI, state-funded models, infrastructure buildout. I see the genuine transformation happening for individuals and small teams, the empowerment of high-agency people, the advance of science.

And I hold both truths simultaneously: this is a bubble, and this will change everything.

The question isn't whether AI transforms the world-I believe it will, just as I believed the internet would back in 1999. The question is whether our economies can survive the transformation we've committed to before that transformation fully arrives.

I'm optimistic about the technology. I'm scared about the economics we've built on top of it.

And unlike when I was sixteen, I now understand that sometimes the economics matter just as much as the magic.

‍

Sources

‍

And this time, the patterns scare me in ways the dot-com bubble never did.

The Valuation Bubble Is Real (And That's the Least Scary Part)

Let's start with what's obvious: the numbers have lost touch with reality.

Nvidia-actually making money-became the first publicly traded company to exceed $5 trillion in market value in October 2025. The company added $1 trillion in market cap in just four months.

Is this a bubble? Yes. Obviously, yes.

I watch these numbers and I think about teenage-me, blissfully unaware that the internet companies were hemorrhaging cash. Adult-me knows this matters. Adult-me has pattern recognition.

But here's what keeps me up at night: this isn't just a valuation bubble anymore. It's become a dependency we can't escape.

The Part That Actually Scares Me: We've Bet Everything

In 1999, if Amazon failed, life went on. The internet would have survived. Other companies would have emerged. It was a private sector bet that went sideways.

In 2025, we've done something far more dangerous: we've made global economic growth structurally dependent on AI succeeding. Not eventually-now.

I remember when consumer spending was two-thirds of GDP and the reliable engine of growth. Now? We're betting on server farms in Texas.

When Nations Go All-In

And it's not just the U.S. This has become a nation-state competition, with countries betting their economic futures on AI supremacy.

Meanwhile, China saw the starting gun and sprinted ahead.

That's what scares me. Not the bubble. The dependency.

But Here's Why I'm Still Optimistic

And yet, despite all of this, despite the hubris and the concentration and the geopolitical risk, I still feel that teenage wonder when I use these tools.

Because the technology is genuinely transformative. Not "might be someday" transformative. Transformative right now, if you're paying attention.

GitHub's controlled study found that developers using Copilot completed tasks 55% faster than those without. McKinsey estimates AI could reduce software engineering costs by 20-45%.

This isn't vaporware. This isn't promises. This is happening now, to anyone who chooses to use it.

I see this empowering high-agency people to do things they couldn't do before. I see science advancing. I see genuine productivity gains-not projections, but measurable improvements happening now.

So yes, I'm optimistic about the technology itself. The teenage wonder isn't gone. It's just complicated now by adult understanding.

The Uncomfortable Middle Ground

Amazon's stock plummeted from $107 to $7 per share during the crash. But unlike Pets.com, Amazon had a viable business model. Today it's one of the world's most valuable companies.

But here's what's different-and what keeps me up at night: we've made this correction potentially catastrophic by building economic dependency before the transformation fully arrives.

Global growth depends on AI investment continuing
But the revenue to justify that investment hasn't materialized
And nation-states can't back out because they've made it a matter of strategic survival
So everyone has to keep believing, keep investing, keep building

What Happens Next?

I wish I had answers. Teenage-me would have ignored all of this and kept building. Adult-me can't ignore the macro implications.

A few things I believe:

And I hold both truths simultaneously: this is a bubble, and this will change everything.

I'm optimistic about the technology. I'm scared about the economics we've built on top of it.

And unlike when I was sixteen, I now understand that sometimes the economics matter just as much as the magic.

‍

Sources

‍

More Explorations

See All