Four telltale signs of an AI boom or bust

VIRTUALLY every company is touting artificial intelligence (AI) today. But here’s the thing: There’s a major difference between “AI-powered” marketing speak and actual revenue improving the bottom line.

With Big Tech collectively pouring hundreds of billions into AI infrastructure, investors need a framework to separate the substance from the spin.

In essence, the question is not whether AI is headed for a boom or bust.

The real question we should be asking is whether the AI revenue generated supports the generous spending. Here are some tips to spot real AI revenue.

Sign #1: Look for specific dollar figures

Vague claims such as “AI is driving growth” or “we are seeing strong AI adoption” should raise red flags. The companies making real money from AI should not be shy in sharing hard numbers.

Meta Platforms is an example of a company that is transparent about AI revenue. In its recent earnings briefing, the social media giant disclosed that its AI-powered advertisement tools generate a US$60 billion annual run rate. That’s actual revenue flowing through its AI systems today.

Even better, Meta breaks it down further. Reels alone hit a US$50 billion run rate – remarkable for a feature that was barely making money three years ago. The Advantage+ shopping campaigns exceeded US$20 billion in annual run rate.

These are specific, verifiable figures that investors can track. When the next earnings report arrives, you can check the progression of these numbers. Compare this to companies that speak in generalities.

If management can’t put a dollar figure on their AI revenue, they either do not know or do not want you to know. Neither is a good sign.

Listen for annual run rates, growth percentages, and segment breakdowns. If a company is vague about AI revenue but specific about everything else, that is a telling sign.

Sign #2: Distinguish between AI revenue and AI-enhanced revenue

This is where it gets tricky and where many investors get confused.

At the end of 2024, Microsoft reported that its AI business exceeded a US$13 billion annual run rate. That’s impressive, but what exactly does “AI business” include?

In Microsoft’s case, it encompasses Azure AI services, GitHub Copilot (26 million users), and Microsoft 365 Copilot (used by 90 per cent of Fortune 500 companies).

These are products where AI is the core value proposition, and customers are paying specifically for AI capabilities.

Users of GitHub Copilot pay a subscription fee for AI-powered coding assistance. Microsoft 365 Copilot costs an additional US$30 per user per month on top of standard Office subscriptions. Azure AI customers pay for compute and application programming interface calls.

In each case, the AI is the product. Contrast this with AI-enhanced revenue, where AI improves an existing product but isn’t separately monetised.

Meta’s advertising falls into this category. Advertisers don’t pay extra for AI-powered targeting, but the AI makes the ads more effective, driving higher volumes and prices.

Meta’s AI systems such as Andromeda (ad retrieval), Lattice (connecting ad knowledge), and Gem (ad selection) work behind the scenes. Advertisers don’t see a line item for “AI services” on their invoices. But these tools drove a 14 per cent increase in ad impressions and a 10 per cent gain in price per ad last quarter.

Both types of revenue are valid. But investors should understand which bucket a company’s AI claims fall into. Pure AI revenue is easier to track and value. AI-enhanced revenue requires investors to draw a clear line connecting the solutions to revenue gains.

Sign #3: Check for customer proof points

Real AI revenue comes with real customers, and real customers leave evidence.

Alphabet does not provide a single AI revenue figure similar to Meta. Instead, the parent company of search giant Google offers customer proof points that tell the story.

For 2025’s third quarter, Google Cloud revenue grew by more than 33 per cent year on year, with operating margins expanding to nearly 24 per cent from around 17 per cent a year earlier.

But the more telling metrics lie in the details. Nearly 150 Cloud customers are each processing about one trillion tokens. These are not companies running small experiments. Trillion-token workloads represent production-grade AI usage at massive scale.

The deal flow confirms this. Alphabet secured more contracts worth more than US$1 billion in the first nine months of 2025 than in the previous two years combined – more evidence that enterprise customers are relying on Google’s infrastructure for their AI distribution.

Similarly, Microsoft highlighted that customers who purchased Copilot in its first month of availability have collectively increased their seat count by tenfold over 18 months. Daily Copilot usage per user has more than doubled over the past quarter.

Usage trends that accelerate over time signal genuine value creation, rather than one-off experimentation. If customers come back for more and use the product more frequently, the AI is delivering real value. Look for paid customers, expansion rates and retention metrics.

Sign #4: Watch the margin impact

The ultimate test of AI revenue quality is whether it improves profitability. This is where hype meets reality. AI infrastructure is expensive, consisting of data centres, chips, energy and talent. If a company’s AI revenue is growing, but margins are compressing, then the unit economics may not work in the long run.

Google Cloud’s operating margin expansion, for instance, is a positive signal. The trend suggests that AI services are scaling efficiently, where revenue is growing faster than costs.

For context, Amazon Web Services operates at around 35 per cent margins, suggesting that Google Cloud has room to improve further. Contrast this with companies where AI investments are diluting margins with no clear path to profitability.

Investment spending is acceptable in the early stages, but at some point, the returns need to materialise.

The question to ask: Is AI revenue accretive to margins, or is the company spending a dollar to make 80 cents?

Get smart: revenue receipts over promises

Here’s something everyone should remember: The AI sector is still in its infancy. ChatGPT is barely three years old. As it stands, many companies are still in investment mode, spending today for revenue tomorrow. That’s fine – as long as they’re honest about it.

The danger lies in confusing potential with performance. A company might have brilliant AI technology and still fail to monetise it. Another might have mediocre technology but excellent distribution and customer relationships.

As investors, our job is not to pick the best AI model. It’s to identify companies turning AI capabilities into durable revenue streams.

The framework is simple: specific dollar figures, clear revenue categorisation, customer proof points and margin impact. Companies that score well are likely building real AI businesses.

The best AI investments come from companies that can show you the money today while articulating a clear path to greater returns tomorrow.

After all, where the business goes, the stock price eventually follows.

The writer owns shares of Alphabet, Amazon, Meta Platforms, and Microsoft. He is co-founder of The Smart Investor, a website that aims to help people invest smartly by providing investor education, stock commentary and market coverage.