"Beyond the AI Hype" clearly struck a nerve. In the last 2 weeks, since it was published on 3 July, the single most common question landing in my inbox has not been whether artificial intelligence will transform industries. Investors seem convinced that it will. The more pressing question has been far more practical: how do we value an AI company?

Is discounted cash flow (DCF) still relevant? Should real options replace traditional valuation models? Is CAPE, the market's favorite valuation temperature gauge, actually a reliable signal for when an AI correction arrives?

This piece is an attempt to answer that question as honestly as I can - why DCF still matters, where it fails, how real options and intangibles change the analysis, and what investors should do about concentration risk, not with a single formula.

The question has only grown more pressing in the weeks since. The past week alone has been defined by high-stakes positioning and tactical profit-taking ahead of the Q2 2026 tech earnings season. Call it the “show me the money” phase.

NVIDIA felt compelled to publicly reassure investors that its “roadmap is intact” after a single report of a server delay briefly rattled Asian tech shares. Meanwhile, a handful of names further down the AI supply chain have quietly posted the triple-digit gains: memory and sensor suppliers up well over 100 percent year-to-date, one of them by more than 500 percent. None of this settles whether today's AI prices are right.

Every AI Company Has Two Prices. One is the market price flashing on the trading screen every second. The other is its intrinsic value—the economic value of all future cash flows the business is expected to generate. Bull markets often blur the distinction between the two. Technological revolutions blur it even further.

Artificial intelligence has created one of the largest valuation challenges modern finance has encountered since the internet revolution. Investors are not merely pricing today's revenues. They are pricing possibilities that may take years to materialize. That is why traditional valuation methods appear inadequate—not because they are obsolete, but because they are frequently misused.

The worth of the orchard depends on how much fruit it will produce over time and what that future harvest is worth today after accounting for uncertainty and the time value of money. Businesses are no different. Forecast future cash flows. Discount them appropriately. Everything else is mathematics.

The weakness of DCF has never been the model itself. It lies in the assumptions fed into it. Two honest inputs, cash flows and a discount rate, decide everything else. The model's Achilles heel is the terminal value, the single assumption covering every year beyond the explicit forecast, which commonly makes up 60–80 percent of total value. For high-growth companies, nearly three-quarters of the estimated value often comes from a single assumption—the terminal value. A small change in long-term growth expectations can dramatically alter today's valuation.

History offers enough reminders. During the dot-com boom, many companies were valued based on perpetual growth assumptions that could never realistically be achieved. The spreadsheet appeared rigorous; the assumptions were not. Perhaps the most useful exercise for investors today is not building another discounted cash flow model. It is running one backward. Instead of asking what a company is worth, ask what growth rate today's market price already assumes. If the answer requires artificial intelligence to permanently generate economic growth beyond anything achieved by previous general-purpose technologies, investors should recognise that they are paying not merely for optimism, but for perfection.

DCF has one blind spot, it can price only a possibility, not a probability. Oil companies routinely pay for drilling rights on land they may never touch, buying the right to decide later, once more is known. That right has value even if the well is never spudded, and financial economists have spent forty years building the mathematics — real options theory — to price it properly. A hyperscaler's data center that can pivot between training and inference and has publicly available utilization data is a disciplined option. By contrast, capital expenditure justified only because competitors are spending aggressively creates no comparable option value. Both look identical on a balance sheet as capex. Only one is actually buying optionality.

Another reason AI companies appear difficult to value is that their most valuable assets rarely appear on balance sheets. Algorithms, training data, engineering talent, developer ecosystems, distribution platforms, and network effects. These intangible assets now make up roughly 90 percent of S&P 500 market capitalization, up from just 17 percent in 1975.  For an AI company, they are nearly the entire investment case. This makes valuation more demanding, not less.

Valuation professionals have three ways to price the invisible: what it would cost to recreate the asset, what comparable assets have sold for, or what incremental cash flow the asset itself generates on its own. The pharmaceutical industry has spent decades getting the third method right because it has clean, verifiable inputs, including clinical trial success rates, patent lives, and a long paper trail of licensing deals against which any single valuation can be tested. Cost spent is not the same as value created, and no industry is relearning that lesson more expensively than the one building AI today.

Which companies will still dominate ten years from now? History shows that technological leadership is surprisingly fragile. Competitive advantage is not created by technology alone. It is created by economic moats that become stronger as customers stay longer, developers invest more deeply, and switching becomes increasingly expensive. A network effect makes a product more valuable as more people use it, directly, as in an early telephone network, or indirectly, as in an app ecosystem. Microsoft Windows is the clean proof: four decades of compounding developer lock-in that survived the internet, mobile, and the cloud, because switching costs grew heavier every year, the loop kept turning. MySpace is the dark corollary: a genuine network effect with no structural switching cost, wiped out within roughly two years once Facebook offered a cleaner product.

Even outstanding businesses can produce disappointing investment returns when purchased at excessive valuations. In this context, Robert Shiller's Cyclically Adjusted Price-to-Earnings ratio, better known as CAPE, acts as a good valuation compass. It smooths a decade of earnings to strip out the business cycle.

Shiller's own CAPE-based warnings, published before the dot-com peak, are among the most validated calls in financial history. The work correctly anticipated the dangers of the technology bubble, but not the exact moment sentiment would reverse. Valuation rarely predicts timing. It predicts vulnerability. That distinction matters enormously today.

Artificial intelligence now represents one of the largest thematic concentrations in global equity markets. Whether investors own AI stocks directly or indirectly through broad market indices, many portfolios have become significantly more exposed than they realise. Concentration risk deserves as much attention as valuation.

My dear readers, remember that a DCF model is only ever as honest as the person who types the number into the terminal-growth cell. The spreadsheet may produce a precise valuation, but it is the quality of the assumptions—not the sophistication of the model—that ultimately determines whether an investment creates lasting wealth.

Before adding another AI-related investment, understand the true concentration risk your portfolio already carries. Look beyond your direct holdings and account for the AI exposure embedded in thematic funds, ETFs, and broad market indices. The distinction between successful investing and successful storytelling has never been more important. Numbers still matter.

Best Regards

 
Dr. Tarunika Jain Agrawal Ph.D. 

Alpha Research

 
The information contained in this communication is intended solely for the use of the addressee and others authorized to receive it.It may contain confidential/legally privileged information. If you are not the intended recipient you are hereby notified that any disclosure, copying, distribution or taking any action in reliance on the contents of this information is strictly prohibited and may be unlawful. If you have received this communication in error, please notify us immediately by phone, fax or email and delete it from your system. Alpha Capital is neither liable for the proper and complete transmission of the information contained in this communication nor for any delay in its receipt