The narrative around AI investment is undergoing a fundamental shift. What began as a story about technological breakthroughs, then evolved into a saga of massive capital expenditure, is now transforming into a tale of financing. Greg Jensen of Bridgewater Associates recently summed up the current predicament succinctly: "We are entering the critical moment for capital."

Morgan Stanley projects that total spending on AI infrastructure could reach $3.2 trillion by 2028, with roughly $1.75 trillion of that needing to be raised through credit markets. The sources of funding are expanding, moving from traditional investment-grade bonds to include leveraged loans, private credit, and securitized products. Capital is available, but it will not come cheaply.

The bond market is already reacting. Data from Apollo Global Management shows that AI-related bond issuance now accounts for 40% of long-duration supply. The credit spreads for hyperscale cloud computing companies have widened significantly this year, while the broader investment-grade market has remained nearly unchanged. "New Bond King" Jeffrey Gundlach has issued a stark warning, suggesting that using GPUs as collateral for long-term bonds is akin to "using bananas to back a 30-year ABS deal."

Cash flow is moving in the wrong direction. While capital expenditure forecasts are being repeatedly revised upwards, projections for free cash flow are steadily declining. Morgan Stanley has sharply reduced its 2027 free cash flow estimates for major hyperscalers. The most striking case is Oracle Corporation (NYSE: ORCL), where the 2027 free cash flow forecast has been pushed close to negative $40 billion.

The scale of spending commitments is equally alarming. The procurement commitments of hyperscalers have exploded to $982 billion. Morgan Stanley points out that a significant portion of real capital expenditure now resides off-balance-sheet. This means that simply looking at balance sheets will severely underestimate the actual financial pressure.

The credit market is already beginning to price in this risk, with investors demanding higher risk premiums. According to Morgan Stanley data, the credit spreads for hyperscalers have "widened significantly this year, far outpacing the general investment-grade market." Spreads for high-quality hyperscalers have expanded by about 25 basis points, and for average hyperscalers by about 22 basis points, while the overall investment-grade market spread has seen zero movement.

Oracle Corporation (NYSE: ORCL) is the most watched case. While the credit default swaps (CDS) for other major hyperscalers generally trade in the 30 to 80 basis point range, Oracle's CDS has surged from around 40 basis points in mid-2025 to a peak of nearly 190 basis points in April 2026, and it currently hovers around 180 basis points. In contrast, Meta Platforms Inc (NASDAQ: META) has a more moderate CDS, having risen slightly to about 75 basis points. The credit market has clearly identified the company it is most worried about.

Balance sheets for now appear strong, but the real problem lies in the future. Morgan Stanley's Q1 2026 data shows that hyperscalers have a total leverage ratio of just 1.3 times, a net leverage ratio of 0.5 times, a cash-to-debt ratio of 128%, and a median credit rating of AA-. By comparison, the overall non-financial investment-grade universe has a total leverage ratio of 2.4 times and a rating of BBB. The issue, however, is the future. Morgan Stanley has sharply increased its 2027 cloud capital expenditure growth forecast from 14% to 29%, and hyperscalers are "reaffirming their confidence in investment returns." As spending forecasts double, the funding gap inevitably widens.

Jeffrey Gundlach, the prominent bond investor, has offered a sharp critique of financing schemes that use AI assets as collateral. Commenting on a plan for a $50 billion fund consortium, he warned that the plan "is likely to fail the test of time." He questioned on social media, "Using assets with an unknown lifespan as collateral for long-term debt? Why not do a 30-year ABS deal backed by bananas in a warehouse?" He added, "These are brand-new engineering bananas with an unknown lifespan." Gundlach's analogy cuts to the heart of the issue: GPUs depreciate rapidly, and their technology iteration cycle is far shorter than the life of the debt. When using such assets as collateral for long-term financing, the actual value of the collateral is highly uncertain.