AI’s New Bottlenecks: Power and Capital

The race to build artificial intelligence is rapidly becoming a contest over two older, less glamorous necessities: electricity and money.

Microsoft’s agreement with Chevron and Engine No. 1 to pursue a large natural-gas-powered project for a proposed data-center campus in West Texas has become one of the clearest signs yet that the industry’s AI ambitions are reshaping both the energy system and the way investors think about Big Tech. The proposed project, tied to roughly $7 billion in investment and initial generating capacity of about 2.5 gigawatts, points to a new reality for companies rushing to add computing capacity: waiting for the grid, or for clean power alone, may no longer be fast enough.

At the same time, investors are recalibrating how they value the sector. Companies that once stood out for fortress balance sheets and abundant cash are now spending so heavily on AI infrastructure that bond yields, financing costs and debt issuance have become far more important to their outlook.

Taken together, the developments suggest that the AI boom is moving beyond a story of models, chips and software. It is becoming a story about how to secure firm power, how to finance enormous construction programs and who ultimately bears the risks.

Power, at any speed

For years, major technology companies presented their growth plans alongside climate pledges and commitments to cleaner energy. But the pressure created by AI data centers — which require vast, uninterrupted amounts of electricity around the clock — is testing those ambitions.

Microsoft’s move in Texas underscores that tension. Natural gas, long criticized by environmental advocates as a fossil-fuel dependency that complicates emissions goals, is increasingly being viewed by some technology companies as a practical answer to a more immediate problem: how to get reliable power quickly enough to keep pace with AI demand.

West Texas is an especially telling setting. The state offers abundant energy resources, available land and its own independent grid, making it a magnet for large industrial power users. The appeal of “bring your own power” or co-located generation has grown as developers confront long waits for transmission upgrades and grid interconnection.

That strain is no longer just a regional concern. On June 18, the Federal Energy Regulatory Commission told regional grid operators outside Texas to justify or revise how they handle connections for very large new loads, including data centers. The move reflected a broader recognition that existing grid rules were not designed for clusters of facilities that can demand power on the scale of a small city.

The basic challenge is structural. AI data centers require long-lived assets — land, buildings, substations, transmission links and dedicated power supply — while the chips inside them become obsolete far more quickly. That mismatch raises the stakes of every decision about location, energy source and financing.

A new bond-market sensitivity

The spending needed to support that buildout is also altering the financial profile of the technology sector.

Investors have long treated the largest tech companies as relatively insulated from swings in interest rates because of their large cash holdings and strong profitability. But the AI infrastructure push is beginning to erode that perception. Hyperscalers and other AI-linked companies are committing tens of billions of dollars to new campuses, networking equipment, semiconductors and power arrangements, while also replacing hardware on a much faster cycle.

As a result, debt markets are taking on new significance. Morgan Stanley has forecast nearly $570 billion of AI-related global debt issuance in 2026, a figure that highlights how central borrowing could become to the next phase of the boom. Recent reporting has shown investors paying closer attention to Treasury yields and credit spreads when assessing the outlook for major tech companies, whose capital spending plans are increasingly large enough to affect cash reserves and financing needs.

That marks a subtle but important shift. In an earlier era, higher rates often hurt technology shares mainly by reducing the present value of future earnings. Now, they can also affect the direct cost of building AI infrastructure. If bond yields rise, the expense of funding giant data-center programs rises with them.

For investors, the AI trade is therefore no longer just a bet on demand for chips or cloud services. It is also a bet on execution in construction, energy procurement and capital markets.

The climate tension

The turn toward natural gas also exposes a difficult contradiction at the center of the AI expansion.

Many of the same companies driving the boom have made prominent commitments to reduce emissions or eventually run on carbon-free energy. Yet the urgency of deploying AI capacity has increased the appeal of fuel sources that can provide steady, dispatchable electricity without the delays that often accompany new transmission lines, nuclear development or some renewable projects paired with storage.

That does not necessarily mean climate goals are being abandoned. Companies can still pursue carbon accounting strategies, offsets, renewable procurement and future clean-power additions. But the immediate emphasis on speed and reliability suggests that, for now, the hierarchy of needs is changing. In the competition to secure enough compute, dependable electricity is becoming the first priority.

What comes next

Several questions remain unresolved. It is not yet clear whether projects of the size proposed in Texas can be completed on schedule, or whether regulators and grid operators will make it easier for giant new loads to connect without imposing costs on other customers. Investors are also still trying to determine how much AI capital expenditure will ultimately be financed with debt rather than internal cash or equity, and whether sustained higher interest rates could more meaningfully pressure valuations across the sector.

But the direction of travel is increasingly clear. The scramble for AI dominance is spilling out of Silicon Valley and into gas fields, power markets and bond desks. The companies at the center of the boom are discovering that the next constraint on artificial intelligence may be less about imagination than about turbines, transmission lines and the price of capital.