AI spending under bubble risk: a decision guide for business leaders

What should business leaders decide first about AI spending?

Decide the size of the commitment before you choose the tool. A budget that fits what the company can carry if AI prices fall keeps every later choice flexible. Start with 3 numbers: the total AI budget for the next 12 months, the share already committed by contract, and the share that can still change. The third number is your room to manoeuvre, and it shows how much risk you hold today.

Then name an owner for each budget line. One accountable manager per project turns a vague spend into a decision with a name and a date. The EQOS Asia AI study finds that 31% of failures come from no alignment with the strategy (EQOS Asia AI study, April 2026), so each owner also confirms the link between the line and a business goal.

Why do AI projects fail, and what does that mean for AI spending?

The study lists the causes of failure by frequency: insufficient or biased data 42%, no alignment with the strategy 31%, team resistance 24%, ROI not demonstrated 22%, and ethical issues or bias 18% (EQOS Asia AI study, April 2026). Projects often cite several causes at once. Each cause maps to a budget line you control: data quality, strategy fit, change management, and measurement.

The study also records that 47% of executives took a major decision based on a hallucination. A spending plan therefore includes review time for outputs, next to licences and compute. Budget for checks the way you budget for the tool: a model that answers fast and wrongly costs more than a slower process that a manager can verify.

How does vendor lock-in change the risk of AI spending?

Lock-in turns a flexible purchase into a fixed obligation. Multi-year commitments, proprietary data formats, and fine-tuned models that run on one platform all raise the price of changing course. If a supplier raises prices, changes terms, or exits the market, the switching cost lands on your budget. Reading each contract for term length, price reset rules, data export rights, and exit help gives you the facts before you sign.

The page on AI vendor lock-in lists the clauses to request: shorter terms, price caps, data portability, and a clear exit plan. Each clause converts a supplier risk into a negotiated term. Run this review on your top 3 suppliers first, since they carry most of your AI spend and most of your exposure.

How do stage gates keep an AI budget under control?

A stage gate releases money in steps. Each step has a goal, a metric, and a date, and the next step starts when the metric is met. This structure matches spending to evidence: a project that proves its value earns more budget, while a project that stalls stops at a small cost. The study shows the need, since 22% of failures cite ROI not demonstrated.

Typical gates are a scoped test, a working pilot, and a production release. The page on stage gates for AI budgets sets the criteria for each gate and the review rhythm. Quarterly reviews fit most companies, and they give you a regular moment to reread the bubble indicators and adjust the plan.

How large should an AI pilot be if the AI bubble may burst?

Size the pilot to the lesson you need, and let the budget follow. A pilot answers one question: does this use case create value in our data and our workflow? A small team, a clear dataset, and a 90-day window answer that question. Larger pilots add cost before they add evidence, and they raise the sunk cost if the market turns.

The guide to sizing an AI pilot gives a method: a cap tied to what you can lose, a success metric agreed in advance, and an exit rule. Successful pilots then scale in stages. Pilots that fall short still pay back, because they teach your team which data and processes need work first.

Which 4 actions raise the odds of successful AI spending?

The study recommends 4 actions for a company deploying AI: measure the data signal before deploying, require explainability, audit the alignment between AI strategy and the field, and fund a retraining plan worth 2% to 5% of total AI investment (EQOS Asia AI study, April 2026). Together they lift the market score from 0.337 to 0.401.

The predicted failure rate falls from 94% to 82%, and the study estimates $80 billion of recoverable value per year for the world market. These figures describe the market, and your company can sit above or below it. The Λ simulator estimates your own score in about 10 minutes, free and anonymous.

How can EQOS Asia score your AI spending decision?

EQOS Asia scores a decision on 9 dimensions and returns a reproducible result. A Reading costs US$358 and arrives in 24 to 48 hours, a Decision costs US$910 and arrives in 48 to 72 hours, and a Study at US$1,369 adds a 12 to 24 month projection. The EQOS Asia pricing page, as of October 2026, lists all 8 tiers.

Bring one concrete decision, such as a platform contract or a 12-month AI budget, and a 30-minute pre-diagnostic shows which tier fits. You leave with a score and a ranked set of options that your board can read in one page.

Frequently asked questions

About the author

Stéphane Jambu

President, EQOS Forecast Asia

Stéphane Jambu leads EQOS Forecast Asia and works as a consultant in e-commerce, search visibility, and AI. LinkedIn profile