AI Is Quietly Moving Wall Street’s Buyback Machine Downstream

Big Tech is putting an unprecedented amount of capital into artificial intelligence, but there is a change that is being overlooked on the other side of those transactions: the companies that are receiving the money are becoming the new investors for corporations on Wall Street.

Perhaps one of the most important figures in the current artificial intelligence boom has nothing whatsoever to do with the number of GPUs shipped, the capacity of data centres, or Nvidia’s market valuation.

It is $100 billion.

As stated in Neuberger Berman’s most recent analysis of S&P 500 share buybacks, the companies that it categorises as AI capital-expenditure receivers carried out stock buybacks amounting to roughly $100 billion over the 12-month period ending June 2026, which represents an increase of 12% compared with the previous period. In contrast, those companies classified as AI-capex spenders cut their buyback programmes by 32% to reach about $85 billion.

This results in a rather odd flow of capital within the equity market since the hyperscalers are investing greater amounts in chips, servers, networking equipment, cooling systems and data centres, which means they have relatively less cash left over for their usual share buyback schemes. At the same time, the companies receiving those payments are seeing their cash flows increase rapidly and a number of them are giving the money back to their own shareholders.

The AI boom might therefore be achieving something even greater than boosting technology earnings: it is causing Wall Street’s corporate demand for equities to shift from the companies that are buying AI infrastructure to those that are selling it.

Five Beaten-Down Stocks Still Riding the AI Boom

Nvidia is the best example of how rapidly such capital can move through the system. As stated in Nvidia’s fiscal second-quarter 2027 results, revenue for the quarter ended July 26 amounted to $96.2 billion, a rise of 106% on the previous year. Revenue from the data centre sector alone was $89 billion, GAAP net income rose to $59.7 billion and free cash flow reached about $21.3 billion.

The capital return followed. In that quarter Nvidia sent back about $26 billion to its shareholders by means of dividends and share repurchases. According to its SEC filing, the company spent $19.7 billion on the purchase of 94 million shares during the quarter and $39.8 billion on the acquisition of 203 million shares in the first six months of the fiscal year. At the end of July Nvidia still had around $99.3 billion left under its share repurchase authorization.

The economic chain is something worth looking into. When companies such as Microsoft, Amazon, Meta and the other AI developers spend money on computing infrastructure, Nvidia receives part of that expenditure as revenue. This revenue turns into operating income and free cash flow. Nvidia then takes part of that cash and buys up its own shares. Capital which had originally been spent by another company eventually flows back into the equity market as corporate demand.

Broadcom has shown that Nvidia is by no means an isolated example. The figures from Broadcom’s most recent quarter can be found here, with the company recording $29.6 billion in revenue and $13.7 billion in free cash flow during its third fiscal quarter. Free cash flow amounted to an exceptional 46% of revenue. Additionally, earlier this year Broadcom’s board approved a new share-repurchase programme of $10 billion which will run until December 2026.

That is the reason why the AI-capex cycle is important beyond just driving revenue growth. Suppliers who are able to turn hyperscaler spending into large amounts of free cash flow have another option: they can reinvest it, acquire other businesses, pay dividends, reduce their debt or buy back their own equity.

Bank of England Scrutinizes the Debt Behind Britain’s AI Investment Boom

The opposite side of the transaction is becoming more and more dependent on capital.

Previously, the technology giants funding the AI expansion had appeared to be considerably less asset-intensive when compared to traditional industrial companies. Software could be scaled at a low cost. Digital advertising also needed only a small amount of physical capital. Cloud computing altered this situation, but generative AI is taking it much further.

Vanguard has estimated that between 2020 and 2024 Alphabet, Amazon, Meta, Microsoft and Oracle together issued on average about $35 billion in debt each year. This figure rose to around $93 billion in 2025, and by July 31, 2026, the group had already issued roughly $132 billion. (Neuberger Berman)

This alters the equation concerning the allocation of capital.

A dollar spent on a new data centre cannot at the same time be used to buy back shares unless the company obtains that capital by borrowing or from some other source of financing. When the scale reaches a sufficient level, AI thus ceases to be merely an earnings story and becomes a balance sheet story covering leverage, interest expenses, free cash flow and finally return on invested capital.

The change is already apparent in the overall data on share buybacks. Neuberger Berman states that S&P 500 companies carried out a record $1.10 trillion worth of stock repurchases over the 12-month period ending in June. However, behind that record figure, the kinds of companies carrying out the repurchases have changed: those who spend on AI have reduced their buying while those who receive AI and financial companies have become the main sources of repurchase activity.

For investors who have a background in finance, this difference is important since buybacks have an impact that goes beyond merely affecting investor sentiment. In the case where repurchases lead to a lower number of diluted shares, the same level of net income is spread over a smaller number of shares, which in turn increases earnings per share. As a result, per-share figures can be improved even if there is no corresponding growth in total profits.

It doesn’t follow that each dollar spent on share repurchases results in value for shareholders. If the company buys back stock that is overpriced, this can lead to a loss of value; the impact of stock-based compensation may counter the decrease in the number of shares, and by returning too much capital the business might give up worthwhile opportunities for reinvestment. The quality of a share buyback is determined by the price, the source of the funds, and the other options that management could have taken with the money.

Why Bond Yields May Matter More Than Stock Earnings Right Now

There is also a powerful opposing argument to the view that AI suppliers are mainly motivated by financial engineering.

Nvidia is an excellent example.

The diluted weighted-average number of shares fell from about 24.532 billion for the previous year’s quarter to 24.285 billion in the most recent quarter, a decrease of only around 1%. Nevertheless, GAAP diluted earnings per share rose by 128% from the previous year.

The main reason for the increase in EPS was therefore higher profits, not a reduction in the number of shares.

Jensen Huang, CEO of Nvidia, described the company’s view of the cycle by stating, “AI has reached its inflection point. It’s doing useful work.” According to Nvidia’s fiscal second-quarter 2027 results, clearly Huang’s attitude is based on the fact that his company is benefiting in the most obvious way from the upswing, but the actual financial results make it hard to ignore the optimistic counterpoint.

There is also a significant misconception concerning passive investing. Just because the value of a stock has increased does not mean that an existing S&P 500 index fund will continuously buy more shares simply since its index weight has gone up; the shares already owned will instead increase in value.

Yet the make-up of the index is important. When new money joins capitalization-weighted funds it is distributed in accordance with the current index weights, while new companies that are admitted result in real portfolio rebalancing needs. This year, S&P Dow Jones Indices added Vertiv, Lumentum and Coherent to the S&P 500, with Marvell Technology and Flex being added later. A number of these companies are situated right within the infrastructure chain and benefit from AI investment.

This creates a more subtle feedback effect than just stating that passive funds are driving up the prices of AI stocks. When there is strong demand for artificial intelligence, the fundamentals of the companies that supply it improve. Improved fundamentals can lead to higher profits, greater free cash flow, and a higher market capitalization. In some cases, companies return part of that cash in the form of share repurchases. As a result, successful companies may end up being included as constituents in larger index funds or may even enter major indexes altogether, which in turn increases their exposure to future passive investments.

The most important question is what occurs when the first link in that chain slows down.

If the hyperscalers later manage to slow down the rate at which they are spending on AI, suppliers might then see a decrease in the growth of their revenues, narrower incremental margins and less free cash flow; at the same time, their capacity to carry out share repurchases could decline since investors are also reducing the valuation multiples they are willing to pay.

Which is why the following stage of the AI trade should be evaluated based on more than just revenue growth.

Investors ought to keep an eye on hyperscaler capex growth, supplier free-cash-flow conversion, the actual number of diluted shares, corporate borrowing costs and evidence of return on invested AI capital since these figures will show if the current cycle is becoming self-sustaining or is merely growing more reliant on ever-larger investment commitments.

At present, the corporate bid for the shares has not vanished.

It has dropped down the AI supply chain.

The question that Wall Street will eventually have to face is whether the profits reaching the bottom of that chain can on continue to grow at a faster rate than the capital being invested at the top.