The number everyone will quote
Big companies anticipate that their spending on AI will increase from 4.5% of their costs last year to 5.8% next year. Since the total amount of the global enterprise market is approximately $30 trillion, expenditure will amount to around $1.7 trillion. On first consideration, this seems like a positive development because demand is at last large enough to justify the existing infrastructure.
A better approach is to think about how the money is distributed. It’s impossible for companies to keep on increasing their budgets. While some of their AI spending will be new, some will result from cost savings achieved through improved productivity, replacing workers, reducing cloud waste, or spending less on outdated technology. Therefore, the AI budget isn’t solely financed by new money – it also involves taking from others’ allocations.
The hidden transfer
It is important to note that not every individual who may suffer as a result is in a strong financial position. The credit data indicates that the majority of software loans are in the area of application software, and most of these borrowers have low credit ratings, for example B- or worse. Although the figures are only approximate, the main point is still clear: many well-established software companies are exposed to the risk of being disrupted as well as the risk of being unable to refinance their debt.
A software company that is profitable can afford to have slower growth and still be able to invest in new products. However, a company which has a great deal of debt and is facing increasing interest costs has much less flexibility. Once customers begin to use fewer tools, request AI-based pricing, or reallocate their budgets towards infrastructure, even a small decrease in revenue can rapidly become a financial difficulty, well before the product is no longer available.
The fact that the figure of $1.6 trillion does not end the debate
By 2026 the AI sector could have a total revenue of about $1.6 trillion and this makes the investment cycle appear more sustainable than it did six months earlier. The fact is, however, that not every dollar of revenue is completely separate from the rest.
A large cloud provider receives revenue when a more recent cloud company rents space from it. The company that provides the AI model earns revenue when the newer cloud company pays to use the model. An application generates revenue when it sells its results to a business customer. All of these are actual transactions, but if you sum them up, you could end up counting the same end-user demand more than once. The total revenue in the ecosystem is therefore not equivalent to new money flowing in from outside sources.
The four gates we would use
Start by looking at revenue generated from external customers, not merely at the internal AI billing. Secondly, compare the amount of computing power being used with the profit margin taking into account depreciation, since revenue can still be recorded for unused computing power even though it may result in a loss. Thirdly, keep an eye on renewal rates and the decreasing number of users in older software, particularly for companies that have a high level of debt. Fourthly, verify that free cash flow increases after having spent on new equipment, not just prior to such spending.
Looking at it in this way results in a more sensible and less impressive conclusion: AI can be genuine and widely applied, yet still fail to deliver good returns for companies in highly competitive markets. On the other hand, it can provide excellent opportunities for businesses that take market share from those burdened with heavy debt. The most important factor is the way money flows, not merely the large revenue figures.
What would prove us wrong
The argument appears less convincing when the majority of AI expenditures are indeed new spending, when the renewal rates for existing software remain constant, when the credit risk associated with application software does not increase, and when infrastructure usage grows quickly enough to deliver a return after allowing for depreciation. Conversely, the case becomes more compelling if companies finance their AI by reducing payments to vendors, whereas weaker software companies are forced to refinance at higher interest rates.
The question of whether the AI boom is real is still being debated. We think the next major question will be this: who is actually funding it, who is being paid more than once, and who is quietly losing the amount of money that had kept their finances in order.
The amount that is being spent on AI is now beginning to seem more plausible. However, the real issue is which of the present sources of income will need to decrease in order to finance it?
Here is the question we have for you: Of the existing budget items, which one do you think AI will replace first – software licenses, outsourced labour, cloud waste, or another item?
MoatPeak Team | moatpeak.com