AI Bubble or Turning Point: What the Summer 2024 Market Correction Means for Companies

The share-price falls of summer 2024 say little about the benefit of AI inside your own company, and whether an initiative stands is decided by its own business case. Decision-makers should therefore stress-test their AI portfolio: does each initiative hold if supplier prices rise, a supplier drops out or expectations fall? Anyone who has answered these questions can stay calm through price swings in either direction and release investment in stages.

This article places the market correction of the first week of August in context, separates the question of how technology companies are valued from the question of what AI delivers in operations, and provides a stress test for your own portfolio. It is written for managing directors and heads of department who are being asked by boards, shareholders or colleagues whether their AI investments can still be justified after the price falls. The statements on markets are a classification and not investment advice.

What happened on the markets in summer 2024?

On 5 August 2024 Japan's Nikkei 225 closed down 12.4 per cent, its largest one-day loss since the crash of October 1987. In the preceding week the US markets had already given ground. Reporting named weak US labour-market data, recession worries and the interest-rate rise by the Bank of Japan as triggers, the last of which accelerated the unwinding of debt-funded yen positions. Highly valued technology shares also fell in this phase. The following day the Nikkei recovered sharply.

A sober conclusion follows for companies. Prices reacted to interest rates, economic data and investor positioning, and there was no single cause in AI technology. Anyone who infers from the prices that AI does not work reads a market movement as a technical finding. Anyone who infers from the recovery that the worries are over makes the same mistake in the other direction.

Is AI a bubble?

The question of a bubble concerns the price of assets, and what matters for companies is the separate question of what the technology delivers. The technology bubble around the turn of the millennium is the textbook case. The Nasdaq Composite reached its high in March 2000 and lost about 78 per cent by October 2002. The internet did not disappear as a result, usage kept growing, and companies that moved their processes onto the new technology in that period benefited for years. The economist Carlota Perez describes this pattern in “Technological Revolutions and Financial Capital” (2002): a phase of installation, in which capital pours into new technology and prices overshoot, is followed by a correction and then by a phase of broad application, in which the economic benefit arises.

Assessments of what AI delivers differ widely, and both sides have solid arguments. A field study of more than 5,000 customer support agents by Brynjolfsson, Li and Raymond, published in 2023 as an NBER working paper, found on average 14 per cent more issues resolved per hour, and considerably more for novices. The economist Daron Acemoglu estimated in spring 2024, also at the NBER, that AI will raise total factor productivity of the whole economy by no more than about 0.66 per cent over ten years, because only a share of tasks is affected. The two results do not contradict each other. They measure different levels: the effect in a clearly defined activity and the effect for the whole economy. For a company the first level counts, because it can be tested in its own process.

What consequences can price swings have for companies?

Consequences arise mainly through the suppliers from which a company obtains AI. There are three routes: suppliers change prices and terms, a supplier discontinues a product or is acquired, and access to capital becomes scarcer for young suppliers. A company that ties its processes to a single supplier bears this risk in full. A company that keeps models and tools interchangeable takes price changes in its stride.

A second consequence concerns the mood inside the company. After price falls, shareholders and boards often ask for AI budgets to be frozen. After price gains they ask for the opposite. Both reflexes share the same flaw: they couple internal decisions to external prices. Initiatives with a proven benefit should continue, and initiatives without a business case should be stopped anyway. What budgets should look like beforehand is shown in the article on AI investment without a plan, and how announcements can be separated from benefit in everyday work in the article on the first half of 2024 in AI.

How do you run a stress test for the AI portfolio?

A stress test checks every running and planned initiative against four scenarios and assigns it a decision. The scenarios are: the supplier's prices rise sharply, the supplier discontinues the product, the expected benefit falls to half, and budgets are cut by a third. For each scenario you answer whether the initiative still has a positive business case and what is to be done. The scenarios and their thresholds (double the price, half the benefit, a third less budget) are assumptions of this article and not market figures. They can be adapted to your own portfolio.

The test can be run in a half-day workshop if the business cases exist. If they do not, that is the first finding. An initiative without a proven benefit fails a stress test, because even under normal conditions it rests on assumptions. The basis of a business case is a clean measurement of the baseline before the start.

How do you release AI investment in stages?

You release it in stages, each ending in a decision backed by measured figures. The first stage is a prototype in the real process with real data that shows the benefit on one metric. The second stage introduces the application in one area, with owners and adjusted processes. The third stage extends it to further areas. Before each stage leadership decides on the basis of the metric whether to go on. This staging protects against both errors: against the activism that starts everything at once out of fear of missing out, and against the freeze that stops even initiatives with a proven benefit after a price slide. The article on the AI hype cycle deals with hype and progress, and the article on AI washing with products that promise more than they deliver.

AI investment with Niehues Consulting

Niehues Consulting is a consultancy for AI management and organisational design from a delivery perspective, based in Düsseldorf. Founder Marin Niehues brings AI into operations, on projects from VC-funded scale-ups to DAX 40 groups and in over 20 countries on six continents. In the service area of AI management the work runs from business analysis and requirements through the business case to the selection and sequencing of applications and the steering of the portfolio.

The approach follows the Niehues Delivery Model. Three of its test questions matter for the question of bubble or turning point: What is the baseline, and how do we measure success? Is the business case confirmed with real values? Which metric shows after 90 days that the solution has landed? The entry point is a free initial consultation of 30 minutes, which you can arrange via contact. Scope and terms of an engagement are discussed there.