Anthropic Seeks Funding. AI Must Prove Its Worth.

Image generated with Codex under licence.

Every week, another AI tool hits the market, promising to outperform everything that came before it. A smarter model. A more autonomous agent. Yet another subscription that somehow feels absolutely essential.

Meanwhile, Anthropic is preparing to go public, raising a question that goes well beyond the company itself: how much better does AI need to get before it can justify what we're paying for it?

What investors are putting in. What businesses are spending. What it's costing society. And, of course, what those of us collecting AI subscriptions are paying in the hope of getting more done.

The Relentless Pursuit of Never-Ending Growth

On October 26, Anthropic publicly filed for its IPO, although the share price and number of shares have yet to be determined. The figures help explain the excitement: the company generated $16.286 billion in revenue during the first half of 2026, up from $1.163 billion in the same period a year earlier.

The losses, however, tell a more complicated story. The reported $80.748 billion net loss (an eye-watering figure by any standard) includes $72.388 billion in financial instrument valuation adjustments — the kind of accounting treatment that can make headline numbers rather misleading. Meanwhile, the operating loss actually narrowed from $2.135 billion to $693 million. That's a genuine improvement in the underlying business, something the headlines tend to overlook. Even so, the company itself warns that substantial investment will still be required.

In short, the headline numbers are staggering, whether we're talking about growth or losses. But look beyond them and there's evidence of a business moving in the right direction, assuming there's no excessive financial engineering distorting the picture. Whatever the case, this is an extraordinarily capital-intensive industry, and Anthropic makes no secret of it.

So the real question isn't simply when the company will turn a profit.

It's how much money it will take to get there. And how much more it will need to keep spending once it does.

Because in this race, taking the lead doesn't mean you can stop running. Staying ahead means investing relentlessly, and doing so at an extraordinary pace.

The Stock Market Will Put a Price on Patience

Going public could give Anthropic a huge financial boost. But it will also mean having its promises measured against its performance, quarter after quarter.

Every three months, the company will need to explain how fast revenue is growing, whether costs are coming under control and how much more investment the next stage will demand.

And here's the crucial distinction: a company can deliver excellent results and still disappoint shareholders. If its valuation already assumes exceptional performance, even impressive growth may fall short of expectations.

In other words, stock markets don't run on numbers alone. Perception and, above all, expectations play an enormous role. That's precisely what the two winners of the 2026 Nobel Prize in Economics, Robert E. Hall and Michael Woodford, have explored in their research:

Economic activity and financial markets are shaped by expectations about the future, not just by present conditions.

The market can applaud extraordinary growth one quarter and dismiss similar results as disappointing the next.

SpaceX offers an interesting comparison: its early public financial results showed strong growth alongside pressure on its share price, as investors weighed concerns over spending and other factors. Impressive growth alone wasn't enough to put those doubts to rest.

Anthropic will face a similar challenge: keeping its business growing while living up to market expectations in an industry where both technology and investor sentiment are constantly shifting.

An Influx of Capital Could Reshape the Competitive Landscape

A successful IPO would give Anthropic the resources to expand its infrastructure, recruit top talent and accelerate product development.

For everyone competing against it, the bar would rise considerably.

We can reasonably expect greater consolidation among companies developing frontier AI models. But that doesn't mean smaller businesses are necessarily doomed.

There's a world of difference between building an AI model and building a solution that genuinely solves a problem.

A specialised company may bring deep industry expertise, proprietary data, seamless integration with existing workflows or relationships built on years of customer trust. Those advantages can be valuable even when the underlying AI model comes from someone else.

The businesses in a more vulnerable position are those charging for a feature that could become a standard part of a general-purpose AI assistant tomorrow.

For many AI companies and tools, the real question will be: why should customers keep paying for them when their main AI provider offers something similar?

There's a Limit to What Users Will Pay, Too

This is where the conversation moves beyond financial markets.

One subscription for writing. Another for coding. Another for generating images. Another for video. And then something new comes along, promising to do several of those things better.

And money isn't the only cost. We're also paying with our time: learning new interfaces, comparing outputs, moving information between platforms and deciding whether switching is worth the effort.

A tool can be technically superior without making a meaningful difference to our work. Sometimes, the improvement simply isn't worth the disruption.

The number of AI tools may keep growing, but our time, attention and budgets are finite.

That's why I expect users to become far more selective. We'll hold on to the tools that deliver tangible results and drop those that merely duplicate existing features or make our workflows more complicated.

The goal isn't to have the latest tool or collect dozens of subscriptions. It's to know which tools actually make sense for our work and, above all, how to get the very best out of them.

I mention this because there's a growing obsession, particularly on social media, with endless rankings of supposedly groundbreaking AI tools. Self-proclaimed influencers would have us believe that a single short prompt is all it takes to achieve spectacular results. And if you're not using the latest release, you're apparently missing out on a world where AI does all your work while you relax under a palm tree.

As I've said many times, AI is incredibly useful and makes a real difference to our daily work. But it still needs direction, correction and supervision. That means knowing what we're trying to achieve, understanding the limitations and accepting that getting good results takes both time and money.

Will a Handful of Platforms End Up Dominating Everything?

Possibly. But I think it's far too early to assume that's how things will play out. Right now, it's hard to imagine a market controlled entirely by just two or three companies.

After all, native AI providers aren't the only players in this market. We also have the tools we already use every day — Office, Workspace, Canva, Adobe, Artlist and many others. These companies have a significant advantage: their customers are already working within their platforms, managing files and following established workflows. In that context, built-in AI can be more useful than yet another standalone application, even when the latter offers more advanced features.

But this is where Agentic AI changes the picture. Integrations are opening up new possibilities, allowing major AI platforms such as Anthropic's Claude Code and OpenAI's Codex to interact with other applications and act as a central command centre for assigning, managing and supervising tasks. The number of available integrations keeps growing, although they're far from perfect. Adobe is one example: you can access Firefly through Codex, but via a browser-based interface rather than a native app. It's not quite as powerful, but it's still a significant step forward.

What does seem increasingly likely is a market built around a few major model providers, a number of general-purpose assistants, a broad range of specialised solutions and an extensive ecosystem of integrations. Along the way, we'll probably see plenty of companies disappear, mergers and acquisitions reshape the industry, and established players lose or gain ground.

Do We Really Need to Chase Every New AI Breakthrough? Is AI FOMO Taking Over?

Obviously, for the companies developing AI models, continuous innovation isn't optional. It's essential to staying in the game.

For those of us using these tools, the calculation should be very different. Does this new release allow me to do something I couldn't do before? Does it genuinely save time? Is the improvement significant enough to justify another subscription?

Anthropic's upcoming IPO will bring some uncomfortable but necessary questions into sharper focus: how much capital does the AI industry really need, and how much value is it delivering in return?

An IPO won't prove that we're in a bubble, nor will it establish that the entire industry has a sustainable business model. A revolutionary technology, genuinely promising businesses and excessive valuations can all exist at the same time.

But it might finally help shift the conversation.