We are in the middle of an AI gold rush. Companies are pouring billions into artificial intelligence, hoping to unearth a fortune in efficiency and innovation. Yet, if you ask most chief financial officers to put a precise value on their AI efforts, you are likely to be met with a blank stare and some nervous shuffling of papers. The truth is, we are trying to measure the 21st century’s most transformative technology with 20th-century accounting tools. It’s not working.
This isn’t just an academic puzzle; it’s a boardroom crisis unfolding in slow motion. The stock market rewards predictable, quarterly earnings. AI, on the other hand, delivers its value in disruptive, unpredictable, and often unquantifiable ways. This creates a massive disconnect—a valuation paradox—that threatens to penalise the very companies making the boldest bets on the future. How do we solve it?
The Great AI Valuation Muddle
So, what exactly is AI business valuation? In simple terms, it’s the attempt to figure out what a company’s AI capabilities are actually worth. This includes its technology, its data, the new business models it enables, and its enhanced operational efficiency. In today’s market, a company’s ability to demonstrate a clear strategy and value from its AI investments is becoming a critical factor for investors, partners, and its own leadership.
We see companies left, right, and centre trumpeting their AI adoption. But when it comes to the numbers on a balance sheet, the story gets decidedly murky. The tools we use—like discounted cash flow or price-to-earnings ratios—were designed for an era of factories and physical assets. They crumble when faced with the fluid, evolving nature of artificial intelligence.
The Numbers Don’t Lie, But They Don’t Tell the Whole Story
Trying to apply traditional accounting metrics to AI is like trying to measure the volume of the internet with a measuring jug. It fundamentally misunderstands what it’s measuring. The value isn’t in the individual servers or lines of code, but in the network of capabilities they create.
The evidence for this disconnect is stark. A recent, and rather telling, McKinsey survey highlighted in a Macao News report found that while 39% of business leaders said AI had some effect on their corporate earnings, a tiny 5% could directly link that spending to the bottom line. Think about that. Billions are being invested, but only a handful of executives can confidently point to where the money went and what it produced.
This isn’t because AI is failing. It’s because we’re looking for the results in all the wrong places and with all the wrong instruments. We are expecting a simple, linear payoff from a technology that is anything but.
Redefining ROI in an AI World
This brings us to the thorny issue of AI ROI measurement. For decades, Return on Investment has been a simple calculation: how much profit did you make from a given expenditure? For most conventional tech projects, like a new CRM system, companies expect to see that return within 7 to 12 months.
With AI, that timeline is stretched dramatically. A Deloitte poll indicates that most businesses don’t expect a positive ROI on their AI projects for two to four years. Why the long wait? Because implementing AI isn’t like installing a new piece of software. It’s more like planting an orchard.
You don’t plant an apple seed on Monday and expect to be selling pies by Friday. You have to prepare the soil, water the sapling, and protect it for years before you see a single apple. But once that tree matures, it produces fruit for decades. Traditional IT projects are more like a vegetable patch—fast to grow, quick to harvest, but you have to replant it every year. AI is the orchard; it’s a long-term tech investment that requires patience and a different way of measuring success.
The Invisible Fortune: Accounting for Intangibles
At the heart of the valuation problem is how we handle intangible assets. In the AI era, a company’s most valuable possessions aren’t on the factory floor; they’re locked in its data centres and its employees’ heads. These are things like:
– Proprietary algorithms: The unique recipes that power a company’s AI.
– Vast datasets: The fuel that trains and refines those algorithms.
– Transformed workflows: The completely new, hyper-efficient ways of working that AI enables.
Current intangible assets accounting rules are laughably out of date. They were built for patents and brand copyrights, not for self-improving machine learning models. Under these rules, the billions spent on collecting data and training a foundational model are often treated as an operational expense—money that’s simply ‘spent’—rather than an asset that grows in value over time. It’s financial malpractice by way of outdated convention.
Companies need a new playbook for accounting that recognises these digital assets for what they are: the primary drivers of future growth. This requires a radical shift towards greater transparency, where firms proactively explain the strategic value of their AI investments, even if the financial returns aren’t immediately obvious.
Embrace the J-Curve: Why It Gets Worse Before It Gets Better
One of the most crucial concepts for any leader in this space to understand is the “J-curve” of AI adoption. The theory goes like this: when you first introduce a powerful new technology like AI, productivity doesn’t immediately go up. In fact, it often goes down.
Why? Because people need time to adapt. Workflows have to be completely redesigned. Old habits have to be broken. There is a period of chaos and adjustment before the real benefits begin to materialise. This initial dip, followed by a sharp and exponential rise in productivity, forms the shape of a ‘J’.
Leaders who panic during that initial dip and pull funding will never reach the upward curve. This is where a long-term tech investment strategy becomes non-negotiable.
We’re seeing fascinating strategic plays in this area globally. Analysis from Jefferies, for instance, points out that Chinese developers are rapidly closing the performance gap with US rivals despite spending significantly less—an estimated $124 billion in capital expenditure compared to America’s $700 billion between 2023 and 2025. They are doing this by cleverly building on powerful open-source models like Meta’s Llama. Alibaba’s open-source model, Qwen, has already outpaced Llama in downloads, showing that brute-force spending isn’t the only path to success. The smart play is often about integration and adaptation, not just invention.
This doesn’t mean direct monetisation is impossible. Look at OpenAI, which is projected to generate a staggering $13 billion from its 35 million paid ChatGPT subscribers. But for most companies, the value of AI won’t come from selling a model; it will come from fundamentally rewiring their business from the inside out.
A New Valuation Playbook
So, where does this leave us? The world of AI business valuation is messy, complicated, and desperately in need of a rethink. The old metrics are failing us, creating a dangerous blind spot for investors and executives alike.
We must move away from short-term, cost-based accounting and towards a more strategic, long-term view. This means:
– Accepting longer ROI horizons and educating stakeholders on the J-curve effect.
– Developing new methods for intangible assets accounting that capture the true value of data and algorithms.
– Focussing AI ROI measurement on strategic capabilities and operational transformations, not just immediate profit.
The companies that thrive in the coming decade will be those that master this new valuation game. They will be the ones with the courage to invest for the long term and the vision to articulate that strategy clearly, even when the numbers on the quarterly report don’t tell the whole story.
The question for every business leader today is a simple one: is your organisation looking at AI through the right lens? Or are you still trying to measure an orchard with the tools you’d use for a vegetable patch?
