A correction in AI valuations would not make useful AI less useful, but it would expose investments that were never tied to measurable value. Boards should make every AI investment stand on its own business case, measured in money against a baseline, and reduce dependence on any single supplier whose finances could change.
In a recent interview, Alvin Graylin discussed, among other things, whether the AI industry is in a bubble. We do not forecast markets. This article sets out what a board should do so that its own AI investments hold their value whichever way the market goes.
Is the AI bubble question relevant to a business that only uses AI?
Yes, for two reasons. The first is that market sentiment shapes internal decisions. When AI is fashionable, investments get approved on the strength of the story. When sentiment turns, the same investments can be cut regardless of their merit. Either way, decisions are being made on mood rather than evidence.
The second is that your suppliers are exposed. Financial authorities have taken the risk seriously. In October 2025 the Bank of England’s Financial Policy Committee judged that “the risk of a sharp market correction has increased”, noting that equity valuations appeared stretched and that some technology companies were trading at valuation ratios which implied high future earnings growth (Bank of England, Financial Policy Committee Record, 8 October 2025). A correction could change the pricing, investment or survival of AI vendors you rely on.
What makes an AI investment resilient to a correction?
The same things that make any investment sound. An AI investment holds its value if it improves a result the business already tracks, such as revenue, margin, cost to serve, cycle time or risk, and if that improvement is measured against a baseline taken before the change. It is fragile if its case rests on strategic positioning alone, on hours saved that nobody redeploys, or on a vendor’s roadmap.
| Resilient investment | Fragile investment |
|---|---|
| Tied to a named value lever on the P&L | Justified by “we need to be doing AI” |
| Baseline measured before the change | Benefits estimated after the event |
| Owner who signs for the benefit | Owned by a committee or a project team |
| Can switch supplier with acceptable effort | Locked into one vendor’s product and pricing |
| Staged funding released on evidence | Large upfront commitment on a multi-year promise |
How should the board protect against supplier risk?
Treat the financial health of key AI suppliers as part of supplier due diligence. Many AI vendors are young, heavily funded and not yet profitable. If funding tightens, they may raise prices, cut services, be acquired or close. The board should expect management to know which AI suppliers are critical, to understand their financial position as far as it can be known, and to hold exit terms that let the business move its data and configuration elsewhere.
Watch long commitments in particular. Multi-year contracts signed at the top of a market can look expensive later. Staged commitments, with break points linked to delivered value, keep options open and give the business leverage at each renewal.
Should the board cut AI spending now, just in case?
No. Cutting across the board would remove the investments that are working along with those that are not. The disciplined response is to review the portfolio and sort it: continue what is delivering measured value, fix what is promising but unmeasured, and stop what cannot show a credible route to value. This is the discipline a finance function brings to any capital programme, and it is why we argue the CFO has a central role in AI, as set out in The CFO’s AI Mandate.
A market correction can even help a disciplined buyer. Prices can fall, talent can become easier to hire and weaker competitors may retreat. Organisations with a clear view of where AI creates value will be positioned to take advantage. Those that have only followed the market will be left defending spend they cannot explain.
Timing matters too. A portfolio review is easiest to run before sentiment turns, while there is still time to renegotiate, stage commitments and build fallbacks. Once a correction arrives, suppliers and internal sponsors are both under pressure, and decisions tend to be rushed. Boards that ask for the review now, as routine, will make better choices than those that wait for the market to force the question.
What should the board ask management?
- For each significant AI investment, what result does it improve, and what was the baseline?
- Which investments would we still fund if AI were no longer fashionable?
- Which AI suppliers are critical to us, and what would we do if one raised prices sharply or failed?
- How much of our AI spend is committed beyond the next twelve months, and on what terms can we exit?
- Who signs for the benefit of each investment, by name?
Where to start
Start with a portfolio review of your AI spend, sorted into continue, fix and stop. Our briefing on how to measure AI ROI and why productivity numbers mislead sets out the five-part measure a CFO will accept: a named value lever, a baseline, early measures, outcome measures and an owner, with a board scorecard. The step on the AIPEX roadmap that fits here is Step 3, Plan: the Executive AI Strategy Day (£1,499 + VAT) produces priorities, named owners and a benefits scorecard in one working day.
Frequently asked questions
Would an AI market correction affect businesses that only use AI?
Yes. Sentiment shapes internal decisions, so good investments can be cut and weak ones approved on mood. Suppliers are exposed too: many AI vendors are young and not yet profitable, so a correction could change their pricing, services or survival.
What makes an AI investment resilient?
A named value lever on the P&L, a baseline measured before the change, an owner who signs for the benefit, the ability to switch supplier with acceptable effort, and staged funding released on evidence rather than a large upfront commitment.
Should boards cut AI spending in case of a bubble?
No. Review the portfolio instead: continue what delivers measured value, fix what is promising but unmeasured, and stop what shows no credible route to value. See The CFO's AI Mandate →
Correct as of 28 September 2026. This content reflects the law, official guidance and deadlines as published on that date. The EU AI Act and its guidance, standards and deadlines are still being amended and clarified. If you are reading this after 28 September 2026, check for later amendments, European Commission guidance and harmonised standards before relying on it. It is general information, not legal advice, and AIPEX accepts no liability for decisions taken on it without specific professional advice.