Are ChatGPT and Claude the AOL and CompuServe of the 2020’s?

The Companies That Introduce Us to a Revolutionary Technology Aren’t Necessarily the Ones That Survive It

What if ChatGPT and Claude are destined to become the AOL and CompuServe of our generation? Not because artificial intelligence is going away, but because the way we’re currently using it may be little more than a temporary stop on the road to something much bigger. And if history is any guide, the companies spending hundreds of billions of dollars to dominate AI today may not be the ones collecting the profits tomorrow.

For those of us who remember the early internet, this should sound familiar. In the 1990s, America Online and CompuServe were the gateways to an entirely new world. Millions of Americans believed that signing onto AOL was the same thing as going onto the internet. AOL made getting online so simple that even your technologically challenged relatives could figure it out. And who could forget that cheerful little voice announcing, “You’ve got mail!”

AOL mailed out so many free trial CDs that you could practically use them to shingle your roof. CompuServe was the somewhat more sophisticated alternative, favored by businesspeople and computer enthusiasts who considered themselves a cut above the AOL crowd. Between them, they dominated how ordinary Americans experienced the early internet. Both appeared to have an almost unbeatable position in one of the greatest technological revolutions in history.

And here’s an astonishing coincidence. In December 1996, AOL introduced its unlimited-access plan for $19.95 per month. Thirty years later, ChatGPT Plus charges $20 a month. Claude Pro? Also $20. Three decades of technological progress, trillions of dollars in investment, and the monthly admission price has increased by a nickel. After accounting for thirty years of inflation, these companies are actually charging substantially less in real dollars than AOL did. I don’t know whether to call that technological progress or a warning about the economics of the entire business. Perhaps it’s both.

Then the internet grew up. People discovered they didn’t need AOL or CompuServe to get online. They could connect directly through their cable or telephone companies, open a browser, and go wherever they pleased. Suddenly, the special services, chat rooms, news feeds, and email systems that once made these companies indispensable weren’t so special anymore. The internet hadn’t failed. Quite the opposite. It had become so successful that the companies that introduced people to it were no longer essential.

Which brings us to ChatGPT and Claude.

Today, millions of people open ChatGPT, type a question into a little box, and wait for an answer. Claude operates much the same way, although it has developed a strong following among writers, programmers, and corporate users. Both are remarkable achievements, and I’ve used AI extensively in my own businesses, writing, research, and legal work. The technology has already changed the way I operate. But are we looking at the finished product, or are we watching the equivalent of AOL handing out free CDs?

Think about how most people use artificial intelligence. You open ChatGPT, explain what you’re trying to accomplish, upload documents, answer questions, and wait for the results. Then you copy those results into whatever program you’re actually using. It’s certainly better than doing everything yourself, but it’s still a remarkably clumsy way to use something supposedly capable of transforming the world. We’ve built an extraordinarily powerful technology and made people interact with it through a glorified text-message window.

Of course, this is already changing. AI companies are connecting their systems to email, calendars, documents, and business applications. But that development actually reinforces my argument. The more artificial intelligence becomes part of the programs we already use, the less reason there is to visit a separate AI website.

Consider Microsoft Excel. Today, someone analyzing a complicated spreadsheet might upload it to ChatGPT, ask a series of questions, and transfer the results back into Excel. But what happens when Excel itself can perform the analysis, identify problems, suggest corrections, and make the changes? Why would anyone want to leave the program, visit another website, and spend fifteen minutes explaining a problem that the software already understands?

The same principle applies to nearly every business application. Lawyers shouldn’t have to explain their cases to AI systems that could already be connected to their documents and legal research. Stock analysts shouldn’t have to manually upload financial statements that could be analyzed automatically. Business owners shouldn’t have to ask an AI assistant which customers need follow-up calls when their existing software already contains the information. Eventually, the intelligence becomes part of the work rather than a separate destination.

And that’s where the problems begin for OpenAI and Anthropic.

The Corporate Customer Isn’t Going to Save Everybody

Both companies are aggressively pursuing corporate customers, and the reasoning is obvious. Consumers may pay twenty dollars a month for an AI subscription, but corporations can purchase thousands of licenses, sign long-term contracts, and spend millions integrating artificial intelligence into their operations. It’s the kind of recurring revenue that makes venture capitalists salivate and investment bankers start calculating their commissions.

There’s just one problem. Corporate customers are likely to confront the same economic issues that eventually confront consumers, only on a much larger scale. A corporation doesn’t want thousands of employees spending their days chatting with artificial intelligence. It wants the work done faster, more accurately, and at a lower cost, preferably without employees having to think about which AI program is doing it.

Imagine a company paying millions annually for enterprise AI access, only to discover that its existing accounting, legal, inventory, and customer-management software can perform those same functions directly. Why maintain thousands of separate subscriptions when comparable intelligence is available inside the programs the company already uses? Why pay a premium for one outside service when another provider offers similar results for half the price?

The corporate market may ultimately be even less forgiving than the consumer market. Individuals develop habits and emotional attachments to products. Corporations have purchasing departments, chief financial officers, and executives whose bonuses depend on reducing expenses. Once the novelty wears off, those executives will start asking an extremely unpleasant question: Why are we paying so much for something everybody else can provide?

Now, corporations are spending real money on AI, and the growth has been extraordinary. Anthropic’s November 2026 SEC filing shows revenue exceeding $33 billion during the first nine months of the year, compared with approximately $2.5 billion during the same period in 2025. That’s staggering growth, driven heavily by business customers. But it tells us very little about what those customers will be willing to pay five years from now, particularly if competing systems become substantially cheaper.

OpenAI and Anthropic could become enormously successful suppliers of AI technology embedded in other companies’ software. But that’s a different proposition from owning the customer relationship and collecting premium subscription fees indefinitely. The companies may succeed in bringing AI into corporate America while discovering that doing so is nowhere near as profitable as their investors currently expect.

Losing Billions While Building the Future

This brings us to the truly astonishing part of the AI revolution: the amount of money these companies are spending and the losses they’re accumulating.

According to audited financial statements reviewed by Ars Technica and the Financial Times, OpenAI generated approximately $13.1 billion in revenue during 2025. Impressive, until you discover that the company reported an operating loss of approximately $20.9 billion that same year. It lost substantially more operating its business than it collected from customers. And according to documents obtained by The Information, OpenAI burned another $3.7 billion in cash during the first quarter of 2026 alone.

Anthropic wasn’t exactly operating a lemonade stand either. Its SEC filing shows that the company generated approximately $4.6 billion in revenue during 2025 while recording an operating loss of roughly $8.1 billion. Combined, OpenAI and Anthropic produced approximately $29 billion in operating losses in a single year. That’s an extraordinary amount of money to lose selling technology that is supposed to make the rest of the world more efficient.

To Anthropic’s credit, its finances improved dramatically in 2026. Its operating loss for the first nine months narrowed to approximately $1.3 billion despite enormous revenue growth, demonstrating that AI can become more profitable as usage expands. But the same SEC filing disclosed approximately $895 billion in future purchase and lease commitments, largely connected to computing capacity and infrastructure. Those obligations extend over multiple years, and some are conditional or subject to termination rights, but the number is still breathtaking.

OpenAI faces its own enormous funding requirements. In September, the Financial Times reported that a company presentation projected approximately $278 billion in cumulative negative free cash flow between 2026 and 2030, although that figure was subsequently disputed by someone identifying himself as an OpenAI finance executive. Regardless of the precise forecast, the current business requires extraordinary continuing investments in computing infrastructure, research, and customer acquisition.

The important distinction is that revenues are growing rapidly, but profitability remains uncertain, and the financial commitments are massive. These companies have raised extraordinary amounts of capital, so they’re not necessarily facing an immediate financial crisis. But continuously financing huge operating losses and new infrastructure commitments with investors’ money is not a sustainable long-term strategy. Eventually, the business must produce enough cash to support itself.

And here is the great irony of the AI revolution: These companies may succeed in making intelligence so cheap and widely available that they destroy the very pricing power their investors are counting on.

The Technology Could Win While Investors Lose

We’ve seen this before. During the late 1990s, telecommunications companies spent billions building fiber-optic networks to accommodate the internet’s expected growth. Their forecasts about demand were absolutely correct. Internet traffic exploded, and those networks eventually became indispensable. Unfortunately, competing networks and improving technology drove prices down, and numerous companies went bankrupt despite having correctly predicted the future.

Being right about a technology and making money from it are two entirely different things.

The same problem could be developing in artificial intelligence. Microsoft, Google, Amazon, Meta, OpenAI, Anthropic, and others are committing enormous sums to data centers, computer chips, electrical infrastructure, and specialized facilities. Nvidia has become a major beneficiary of this spending spree. But what happens if the cost of delivering useful AI continues falling while the companies providing it remain saddled with enormous infrastructure obligations?

Remember what happened to computers. Machines that once filled entire rooms eventually fit on desktops, then laptops, and finally into our pockets. Computing power that once cost a fortune became so inexpensive that we barely think about it. The same process is already underway with AI, where smaller and more efficient models can perform an increasing number of useful tasks without requiring the most expensive systems.

An inexpensive AI system doesn’t have to be as powerful as the latest model from OpenAI to handle accounting, sort emails, analyze routine documents, or manage customer inquiries. It only has to be good enough to do the job. And if customers can accomplish 95 percent of what they need at a fraction of the price, the economics of the most expensive providers become increasingly difficult.

There’s a reason I keep returning to this historical comparison. The average person seems to think history began the day they were born. Wall Street often operates under the same assumption. Every generation discovers a revolutionary technology, convinces itself that the old economic rules no longer apply, and proceeds to bid up the companies associated with it to extraordinary valuations. Then the bills come due, competition arrives, and everyone acts surprised.

The railroad industry transformed America, but many railroad investors lost fortunes. The automobile revolution changed the world, but hundreds of manufacturers disappeared. The internet created trillions of dollars in wealth while leaving behind a remarkable collection of failed companies whose executives were absolutely convinced they had the future cornered.

Why should artificial intelligence be any different?

Remember What Happened to AOL

AOL didn’t lose its dominant position because the internet became less valuable. It lost because internet access became easier, cheaper, and available everywhere. The company had mistaken its early position as the gatekeeper for permanent control of the technology. CompuServe suffered much the same fate, and Yahoo eventually discovered that being one of the most recognizable names on the internet didn’t guarantee permanent dominance either.

Some people still use their @aol.com email addresses today. There’s nothing inherently wrong with that, but it’s a reminder of how thoroughly the technological world can change while people cling to familiar habits. There was a time when AOL seemed practically indispensable. Today, the very idea of paying a company simply to provide a special doorway to the internet seems almost absurd.

Thirty years from now, perhaps our grandchildren will find it equally amusing that we once visited a separate website, typed elaborate instructions into a box, and waited for a computer to tell us what to do. Artificial intelligence may become as ordinary and invisible as electricity. You’ll use it constantly without giving much thought to which company supplies it, just as you don’t spend your morning wondering which power plant is running your coffee maker.

Does that mean OpenAI and Anthropic are doomed? Certainly not. Both could become enormously successful technology suppliers, and their sophisticated models may remain essential for demanding work that cheaper alternatives cannot handle. But that doesn’t guarantee that their present subscription models will dominate forever, or that their future profits will justify today’s extraordinary investments.

In fact, the most valuable AI businesses may turn out to be companies that simply use artificial intelligence to improve existing products and eliminate expensive work. They won’t necessarily be the companies building the largest models or spending the most money. They may be the businesses that figure out how to put inexpensive intelligence to work while their competitors are still arguing about whose chatbot is smartest.

The real question is whether today’s AI giants can establish lasting, profitable businesses before much of what they sell becomes cheaper and more widely available. If they succeed, investors may be rewarded handsomely. If they fail, artificial intelligence could still transform civilization while leaving some of its biggest financial backers wondering what happened to their money.

Thirty years ago, millions of Americans believed they were experiencing the future when they logged onto America Online. They weren’t entirely wrong. They were simply seeing the beginning of something much bigger than AOL itself.

Today, millions of people believe they’re experiencing the future when they open ChatGPT or Claude. Once again, they may be seeing only the beginning.

And in the meantime, I’m still waiting for my free ChatGPT CD-ROM to arrive in the mail.

Source: https://www.financialsurvivalnetwork.com/p/are-chatgpt-and-claude-the-aol-and