AI Boom Faces a Funding Crisis: Why Intel and Broadcom CEOs Are Worried

The AI boom may face a constraint that more chips and electricity cannot solve on their own: financing the infrastructure on terms that make economic sense. Intel CEO Lip-Bu Tan and Broadcom CEO Hock Tan raised that concern at a conference in Singapore on October 12. Their message was especially relevant because both still described strong near-term demand for computing infrastructure.

Let’s break down what the CEOs said, why AI funding is becoming a bigger issue and how the risk reaches company earnings and valuations. The key question is whether the businesses using all this infrastructure can generate enough cash to sustain the spending.

What did Intel and Broadcom’s CEOs say about AI funding?

The remarks came at the DBS Asia Future Forward Conference. DBS’s official agenda confirms that both executives participated in a session on financing the next phase of AI investment, moderated by DBS CEO Tan Su Shan.

Sources: CNBC’s October 12, 2026 conference coverage, accessed through its syndicated report; Reuters, October 12, 2026; DBS’s official conference agenda.

Reuters separately reported Hock Tan’s emphasis on capital while noting his positive comments about revenue generated by frontier AI labs. CNBC reported that both executives remained optimistic about near-term computing demand. The warning should therefore be understood as a financing and returns question, rather than evidence that they believe demand has already collapsed.

That distinction is the article’s starting point. An industry can have customers who want more capacity while still struggling to raise the money required to deliver it. The harder test is whether those customers can pay for that capacity throughout the financing period.

Why does the AI boom need so much outside funding?

AI infrastructure requires spending before most of its economic benefits arrive. Developers need chips, buildings, networking, power connections and operating capacity. They must then attract enough paying customers to make the investment worthwhile.

The International Energy Agency’s 2026 analysis provides useful context. It says the largest technology companies’ capital expenditure exceeded $400 billion in 2025 and projected a further 75% increase in 2026. These figures refer to those companies’ capital spending, not a separately measured total for AI alone or a final audited outcome for 2026.

The IEA also argues that data-centre investment has become too large to fund exclusively from company balance sheets. Capital markets consequently matter to the pace of expansion. That makes expected returns, investor confidence and financing conditions part of the AI supply chain.

There are several ways to raise this money. A company can use retained cash, borrow, issue shares, lease equipment or partner with an infrastructure investor. Each choice allocates the cost and risk differently, but none removes the need for the assets to produce economic value.

Debt creates interest and repayment obligations. New equity can reduce existing shareholders’ percentage ownership. Leasing reduces some immediate cash needs but leaves future payment commitments. The central concern is whether future cash generation can support the obligations created today.

What do the latest AI financing deals reveal?

Recent announcements and reporting show that chip supply and financing are becoming increasingly connected. However, the amounts below describe different structures and stages. They should not be added together as a measure of debt already raised.

Sources: Financial Times, October 5, 2026; The Wall Street Journal, October 7, 2026; Nvidia’s financing-platform announcement, August 10, 2026.

The Nvidia number is particularly easy to misread. It is a target for third-party capital across platforms over time, rather than Nvidia revenue, a single loan or proof that the full amount has already been deployed. Nvidia’s own subsequent explanation makes those distinctions explicit.

Syndication also has a specific meaning: lenders distribute portions of a financing package to other institutions. It does not necessarily mean every dollar has already been drawn by the borrower. Early financing discussions are still less definite.

There are signs that investors are becoming more selective. On October 11, the Financial Times reported that AI-related debt issuance fell to $23 billion in September, less than half the August amount, citing Morgan Stanley data. A decline in issuance can reflect timing and earlier borrowing as well as caution, so it is not proof of an industry-wide funding freeze.

Our assessment is that the risk is a tougher selection process. Well-supported projects may continue to obtain capital while proposals with weaker customers, uncertain utilisation or aggressive return assumptions face higher costs or delays. This could produce a more uneven AI expansion without ending it.

Why strong chip sales do not settle the question of AI returns

There are three separate financial checkpoints in the AI economy. A supplier’s results mainly tell us about the first one. Investors need evidence from the other two before concluding that the entire investment chain is self-sustaining.

A supplier can recognise a successful sale while the customer’s project is still establishing its business model. Both can be true at the same time. Strong revenue at the supplier is encouraging, but it does not establish the eventual return earned by the customer or lender.

The chain becomes more exposed if equipment suppliers also support customer financing. This can help useful projects proceed, but investors then need to understand the associated commitments. Exposure may depend on guarantees, investment arrangements or residual-value support rather than appearing only as ordinary corporate borrowing.

Nvidia’s explanation says its financing partners will independently assess customers, utilisation, cash flow and residual value. That is the right set of questions for investors too. The ability to redeploy equipment and retain its value matters if the original customer’s economics weaken.

How are Broadcom and Intel exposed differently?

Their CEOs are discussing a shared industry risk, but their companies have different business models. Broadcom supplies custom AI accelerators and networking alongside infrastructure software. Intel has processor, packaging and manufacturing businesses with substantial investment requirements.

The latest reported quarterly results provide useful context. The fiscal periods differ and the AI-related revenue categories below are not equivalent measures.

Sources: Broadcom’s September 2, 2026 earnings release and August 2 Form 10-Q; Intel’s July 23, 2026 earnings release and June 27 Form 10-Q.

Broadcom: Strong cash generation still leaves customer-funding risk

Broadcom’s reported cash generation is an important counterpoint to a blanket AI funding-crisis narrative. The company generated $13.7 billion of free cash flow in its latest quarter, using its definition of operating cash flow less capital expenditure. That demonstrates substantial cash production from its existing business.

The separate concern is whether customers can finance the next round of purchases. If customers postpone deployment, a supplier can face slower order conversion even while its current financial position is strong. Where it supports financing, the supplier must also manage the additional exposure created by the arrangement.

Broadcom’s Form 10-Q identifies customer concentration and related credit risks. Concentration figures involving distributors are not the same as the number of ultimate AI customers, so investors should read the disclosure carefully. The practical question is how much future growth depends on a small set of customers continuing to invest and pay.

Intel: Demand has to justify manufacturing investment

Intel’s CFO said in the Q2 release that the company was increasing investment in equipment, clean-room space and substrates to support expected growth. This makes capacity utilisation and the timing of customer demand especially relevant. Investing ahead of growth can pay off, but underused capacity can weaken returns.

Intel’s reported GAAP loss also needs context. Its filing describes a large fair-value loss related to escrowed shares associated with its US government agreement, helping explain the gap between GAAP and adjusted earnings. That accounting loss should not be read as an equivalent cash outflow from operations.

Cash commitments tell a different part of the story. Intel’s negative adjusted free cash flow included a net partner cash outflow of approximately $12.2 billion in the quarter, alongside investment and finance-lease payments. This reconciles positive operating cash generation with the negative adjusted measure. The quarter’s shortfall should not be extrapolated as an identical recurring funding requirement.

For investors, the financing question is therefore broader than a single earnings figure. It involves the cash needed for investment, the returns achieved on capacity and the terms under which further capital becomes available. Adjusted profit provides one perspective, but cash flow and the reconciliation to reported results remain necessary.

How higher funding costs can squeeze an AI project

Consider a hypothetical project costing $10 billion, funded with $7 billion of debt and $3 billion of equity. Assume it generates $1 billion a year in cash after operating expenses and maintenance capital expenditure but before financing costs and tax. The table shows what happens if financing becomes more expensive and operating cash generation weakens.

A three-percentage-point rise in the rate adds $210 million to annual interest in this example. The final scenario uses all available cash just to pay interest, leaving none for tax, principal repayment or shareholder distributions. It illustrates how an investment can become financially fragile without demand disappearing completely.

The timing of the rate change matters. Higher market rates do not automatically reprice existing fixed-rate debt. The effect would apply to new borrowing, refinancing or debt whose rate can change under its terms.

This is also a deliberately incomplete investment model. It excludes construction delays, growth capital expenditure, principal amortisation and terminal asset values. Those items would need to be included before assessing the actual return to lenders or shareholders.

Why AI stocks can fall even while earnings grow

Funding risk can affect a stock through earnings expectations and through the value investors assign to those earnings. A company might still grow, but less quickly than expected if customers delay spending. Investors may also demand a lower valuation multiple because future cash flows have become less certain.

Price-to-earnings (P/E) compares the share price with annual earnings per share. The example below uses a fictional company to separate earnings growth from a change in valuation.

The final scenario has higher earnings but a lower implied value than the starting point. That is why strong reported sales do not guarantee a positive stock return. A valuation can already assume years of rapid expansion, leaving little room for tighter financing or delayed deployment.

Actual company comparisons need more care. GAAP and adjusted earnings use different treatments. A conventional P/E is not useful when the earnings used in its denominator are negative. Intel’s large accounting adjustments make it particularly important to identify which earnings measure a quoted ratio uses.

For Nvidia and other infrastructure suppliers, the useful test is whether forecasts remain supported by customers’ funding and operating economics. The headline size of a financing initiative cannot answer that on its own.

What does the AI funding risk mean for Indian investors?

Indian investors with US technology shares or global funds should look at their combined exposure to the AI investment cycle. Several holdings can depend on the same customers’ infrastructure budgets even if they operate in different industries. Owning chips, networking and data-centre suppliers can therefore provide less economic diversification than the number of stocks suggests.

The theme also matters when evaluating Indian data-centre, power and equipment businesses. Announced project capacity should be assessed alongside financing, customer contracts, construction progress and the cash required before revenue begins. The US executives’ remarks alone do not establish that any particular Indian project is underfunded.

For Indian IT services companies, a more selective investment environment could increase demand for measurable productivity improvements while making some customers more cautious about experimental projects. That is a possible business effect rather than a forecast of company earnings. Investors should look for actual contract wins and revenue conversion before assuming either a benefit or a setback.

What should investors watch to judge whether AI funding is becoming a problem?

The first signal is whether announced projects secure binding financing on workable terms. The second is whether funded capacity becomes operational and attracts paying usage. The third is whether cash receipts grow sufficiently to support debt service and further investment.

At suppliers, watch the relationship between revenue growth, receivables and operating cash flow. At borrowers, look for refinancing needs, utilisation and payment commitments. Where financing support is provided, examine the terms and size of the exposure rather than assuming it matches the headline project amount.

The counterargument deserves equal attention. Strong customer revenue, better model efficiency and broader adoption could improve project economics and make capital easier to attract. Financing platforms can help address a real bottleneck when lenders evaluate the underlying cash flows independently and conservatively.

Our assessment is that capital will increasingly determine which AI projects proceed and how quickly. The strongest financial evidence is growth supported by recurring customer payments and cash generation. An ambitious construction plan or a large funding target is a starting point for analysis, not proof that the investment will earn an adequate return.