毕马威调查显示仅26%企业完全掌握AI支出情况

This article shares key takeaways from newly disclosed data on global corporate AI spending management, as outlined below:

1. Unpublished 2026 KPMG survey data shows that AI spending transparency among global enterprises remains extremely low today. Only 26% of companies have full visibility into their AI spending details, 50% have limited oversight over their AI expenditures, and the remaining 22% have no clear insight into how much they spend on AI, only learning their actual costs after receiving invoices.

2. Most current AI services operate on a token-based pricing model, where tokens are the basic unit large language models use to process text, and costs are calculated based on actual token consumption. This pay-after-use model easily creates unexpected spending pressure. There are already documented cases of companies burning through their full annual AI and cloud budgets within a few months, and more enterprises are expected to face this issue in 2026.

This article outlines key actionable insights for brands building out AI capabilities, as follows:

1. When rolling out AI applications, brands must prioritize AI spending oversight first. Across all industries, only 26% of companies achieve full AI spending transparency, leaving the majority exposed to the risk of uncontrolled spending. Brands need to build out an AI spending governance framework in advance.

2. As a new category of corporate investment, AI spending tends to grow exponentially. Combined with the existing token-based, pay-after-use pricing structure, unexpected budget overruns are highly common. When building annual financial budgets, brands should reserve sufficient budget flexibility to avoid sudden overruns disrupting overall operations.

3. Drawing on the pattern of widespread cloud spending cuts following the post-pandemic cloud boom, brands should not rush into AI investment following the crowd. Instead, they should pace their investments: test use cases first, then scale gradually, avoiding waste from overinvestment upfront that will later force costly budget cuts.

This article shares risk warnings and actionable guidance for sellers adopting AI, as outlined below:

1. Risk warning: AI spending oversight is inherently difficult today, and most sellers cannot accurately forecast their total AI costs upfront. The token-based, pay-after-use model makes sudden overruns very likely. There are already cases where token consumption rose 6x within months, burning through a full year’s token and cloud budget, and more sellers will encounter this problem in 2026.

2. Action step: Sellers should prioritize reviewing AI provider invoices this quarter, audit AI usage and reconcile spending proactively, to avoid being caught off guard by large unexpected bills that hurt cash flow.

3. Strategic guidance: Following the development pattern of cloud services after the pandemic, AI investment is currently in the early stage of rapid growth. We recommend sellers build an AI spending oversight framework first, then scale AI investment gradually, rather than ramping up budgets blindly to follow industry trends.

This article outlines key guidance for factories pursuing digital transformation and adopting AI technology, as follows:

1. AI is a new category of management resource unlike traditional capital investment, and its spending grows exponentially. When advancing AI transformation and adopting AI services, factories must first build an AI spending governance system to address the lack of spending transparency. Across the industry, only 26% of factories have full visibility into their AI spending, leaving most with critical oversight gaps.

2. The dominant AI pricing model today is token-based, pay-after-use billing, which creates a high risk of unexpected overruns. When partnering with AI service providers, factories should clarify billing terms in advance, require providers to offer real-time usage tracking, and reconcile usage regularly to avoid discovering overruns only when the invoice arrives.

3. With multiple companies already burning through their full annual AI and cloud budgets in just a few months, factories should not rush into AI adoption blindly. Drawing on post-pandemic cloud spending correction lessons, factories should run small-scale tests first, expand gradually, and reserve sufficient budget flexibility.

This article breaks down industry pain points and growth opportunities for AI-related service providers, as outlined below:

1. Enterprise clients universally struggle with AI spending control today: only 26% of companies have full visibility into their AI spending, 50% have limited oversight capabilities, and 22% have no clear insight into their AI spending details at all. This widespread pain point represents a major new business growth opportunity for providers.

2. The existing token-based, pay-after-use model creates significant financial pressure for corporate clients, with multiple cases already documented of clients burning through full annual budgets in just a few months. Providers can develop targeted supporting tools for spending transparency and budget control, and optimize billing models to meet client demand.

3. AI spending is currently in a phase of exponential growth. Following the development trajectory of cloud services during the pandemic, enterprise demand for AI spending management will rise rapidly in coming years. Providers that build out these solutions early can capture early market share.

This article outlines key takeaways for platform operators offering AI services, as follows:

1. Enterprise clients currently have an urgent need for transparent, controllable AI spending. Under the existing industry model, only 26% of companies can fully track their AI spending, leaving most exposed to the risk of major budget overruns. AI platforms need to upgrade their service capabilities to address this pain point.

2. The existing token-based, pay-after-use model easily triggers unexpected overruns. There are already cases where clients saw token usage jump 6x in months and exhausted their full annual budgets. Platforms can optimize billing mechanisms, add features such as real-time usage alerts and budget cap notifications to improve client experience.

3. Drawing on the pattern of widespread enterprise cloud spending cuts after the post-pandemic cloud boom, platforms should standardize billing transparency early and help clients control spending. This will help avoid extreme industry boom-bust cycles, stabilize business growth, and mitigate the risk of widespread client order cuts down the line.

This article outlines new issues and trends in AI industry development for industry researchers, as follows:

1. Newly disclosed KPMG survey data shows that a new common industry problem has emerged as AI scales to enterprise adoption: unregulated AI spending. Only 26% of enterprises have full visibility into their AI spending, and more than 70% suffer from varying degrees of oversight gaps. This is an emerging research topic as the AI industry enters its mass adoption phase.

2. The dominant business model for AI services today is token-based, pay-after-use billing. As a new resource requiring dedicated management, AI spending grows exponentially, and multiple cases of early annual budget exhaustion have already been documented. More enterprises will expose this problem in 2026, providing new research samples for studies of AI commercialization pathways.

3. The analogy drawn between current AI spending chaos and the post-pandemic cloud boom gives researchers new comparative material to explore the cyclical patterns of new technology industrialization and enterprise investment adjustment, enabling further research into common characteristics of new technology industry development.