JPMorgan recently increased its 2026-2027 spending forecasts for the massive data centers run by America's top tech companies.
This adjustment isn't based on a hunch; it's a direct response to a powerful combination of factors. First and foremost, the tech giants themselves—Alphabet (Google), Amazon, Microsoft, and Meta (Facebook)—have publicly announced enormous increases in their capital expenditure, or Capex, plans. During their April 2026 earnings calls, they collectively signaled hundreds of billions of dollars in new spending, driven almost entirely by the race to build out AI infrastructure. This provides a solid, official foundation for JPMorgan's revised estimates.
Second, building these data centers is becoming more expensive. The cost of essential components like memory chips and optical parts is rising due to high demand and supply chain bottlenecks. For example, Microsoft noted that about $25 billion of its projected spending is due to higher component prices alone. This inflation means that even to build the same amount of infrastructure, companies have to spend more money.
Third, and perhaps most critically, is the physical limit of electricity. Data centers, especially those powering AI, consume vast amounts of power, and the U.S. power grid is struggling to keep up. This power shortage has become the main bottleneck, limiting how quickly new data centers can come online. You can have all the servers you want, but they're useless without electricity.
So, how are companies responding? Since they can't simply plug in more servers indefinitely, they are focusing on making their existing hardware as efficient as possible. This is where networking and optics come into play. By upgrading the connections between servers with faster switches and fiber optics, they can 'unlock' the full potential of the GPUs they already have. A dollar spent on better networking can yield more usable computing power than a dollar spent on a new server that can't be powered on. This explains why JPMorgan highlighted networking companies as key beneficiaries. To top it all off, massive bond sales by Meta and Amazon confirm they have the financial firepower to fund this AI 'super-cycle'.
- Glossary
- Capex (Capital Expenditure): Funds a company uses to buy, upgrade, and maintain physical assets like property, buildings, or equipment.
- Hyperscaler: A large-scale cloud computing provider that can offer services to millions of customers. Key examples are Amazon Web Services (AWS), Microsoft Azure, and Google Cloud.
- Networking and Optics: The hardware, such as switches, routers, and fiber optic components, that allows servers to communicate with each other at high speeds within and between data centers.