JPMorgan says $5.5T AI capex through 2030 is profitable, but “for now.”

JPMorgan’s midyear outlook has one very specific claim that matters: the global AI capex explosion is not just happening, it is increasingly profitable, “for now.” The key number behind the argument is $5.5 trillion in total global AI-related capital expenditures through 2030, up from $5.1 trillion. JPMorgan frames this as a broadening capex cycle that is anchoring growth expectations, with the “AI upstream” buildout driving demand, including data centers, chips, and supporting infrastructure.

So what does “profitable” mean in this context? JPMorgan points to hyperscalers as the main drivers and expects their capital expenditures to reach $650 billion in 2026 and exceed $1.1 trillion in 2027. It also projects operating cash flow to surpass $900 billion by 2027. The bank’s argument is that the business model is still turning cash, even as the industry spends aggressively to add capacity.

To understand why JPMorgan thinks the cycle can keep running, start with where the money is concentrated. The report says “AI upstream” investment is still heavily concentrated in the U.S., which accounts for about 85% of AI and machine learning venture capital. That concentration matters because it means supply chains, financing, and buildout momentum are clustering geographically. JPMorgan expects spillover benefits in China, South Korea, and Taiwan, tying those regions to their roles in semiconductor supply chains. In other words, the cash is mainly pouring into one part of the world, but the manufacturing and component ecosystems create downstream effects elsewhere.

Now zoom in on the part that tends to break bubbles: the financing mechanics. JPMorgan raised its estimate for debt financing tied to the AI buildout to $4.1 trillion, citing higher loan-to-cost ratios. The report describes loan-to-cost ratios averaging above 85%, with some exceeding 90%. That range signals that, for many projects, debt is being underwritten with a favorable view of asset values and credit conditions. JPMorgan also points out that equity markets are sometimes valuing expansion surprisingly richly. For example, it cites a $15 million-per-megawatt investment translating into a $25 million increase in market capitalization. Whether you interpret that as efficient pricing or exuberance, it is still a real signal that investor expectations are supporting the funding environment.

But JPMorgan is not ignoring the obvious risk that usually turns a “profitable for now” thesis into a later reckoning: the question of whether AI demand will grow fast enough to justify the capacity being built. The report acknowledges that scale of spending can raise concerns about matching real consumption. Even with cloud providers such as Amazon, Google, and Microsoft reporting rising AI revenue, investors remain divided on how long it will take for returns to match investment levels. That is the tension. Debt and cash flow can look fine during a ramp, but the economic payoff depends on adoption keeping pace with buildouts.

JPMorgan’s view of profitability, however, is supported by a broader picture of balance-sheet strategy and market structure. Recent equity issuance is reinforcing balance sheets, and JPMorgan suggests that elevated leverage can reflect a strategic choice: companies are financing projects with debt while conditions are favorable, preserving flexibility to deleverage later. That is a crucial nuance for boards and CFOs. It implies the risk is not just “how much capex,” but “how capex is financed” and whether the firm can stay nimble if credit conditions change.

Which brings us to the credit market, the actual enabler behind the spending surge. JPMorgan forecasts that high-grade corporate debt will account for more than $2.1 trillion in data center financing over five years. For 2026, it anticipates $150 billion in U.S. hyperscaler issuance and another $100 billion equivalent abroad. Additional financing of about $170 billion is expected from data center and chip issuers outside the core high-grade market, though alternative channels are described as relatively small. Translation: for now, the cycle is getting a steady runway from credit markets that are still willing to fund the upstream AI buildout.

For executives watching from the sidelines, the strategic stake is straightforward. This is a highly concentrated investment cycle, meaning any slowdown could have an outsized impact on industry growth because the spending is driven by a small group of companies. JPMorgan’s conclusion is essentially that AI-driven capital spending is scaling rapidly and the economics are holding, for now. The market still has one job left: keep tracking whether adoption grows at a pace that can absorb the trillions being invested. If it does, the profitability case strengthens. If it does not, the same debt and cash flow mechanics that fuel the ramp can amplify the downside.