We’re wrapping up our telecom year-in-review series with perhaps the hottest topics of them all: data centers, cloud, and AI.
TeleGeography experts Jon Hjembo and Patrick Christian join the show to address core questions about recent data center and AI booms—specifically, what it means for the year ahead.
If you haven’t yet, be sure to listen to our first two episodes on transport trends, pricing, and enterprise networks.
Listen to the full episode below.
Subscribe to access all of our episodes:
Apple | Amazon | Spotify | Stitcher | TuneIn | Podbean | RSS | YouTube
Key Takeaways
The Data Center Pipeline has Exploded, but It’s U.S.-Centric
The volume of data center projects has reached unprecedented levels, with the immediate pipeline tracking approximately 650 sites—more than double the standard historical volume.
While commercial data center builds are spread globally, projects specifically dedicated to hyperscale and AI are aggressively concentrated in the United States. Currently, about 35% of all new builds are in the U.S., but that figure jumps to over 60% when isolating AI and hyperscale facilities.
This concentration is partly driven by the availability of chips; geopolitical constraints and export controls on Nvidia chips are forcing operators to scale up in the U.S. or partner nations like Saudi Arabia rather than distributing infrastructure evenly worldwide.
Power Constraints are Forcing a Geographic Migration
The primary bottleneck for new infrastructure is no longer just proximity to network hubs (like Ashburn, Virginia) but the availability of power.
- Grid Strain: In Northern Virginia, the world’s largest data center market, reserve power capacity is projected to drop from 4% to 1% in the near future, raising the risk of brownouts and driving wholesale electricity prices up by an estimated 11 times by 2027.
- New Hubs: To bypass these constraints and “NIMBY” resistance, developers are moving to non-traditional markets like West Texas and Columbus, Ohio. These locations are chosen specifically for their access to power generation and land, rather than latency or population density. In Columbus, planned data centers account for roughly 13% of all planned power.
Cloud Providers are “Retracting” to the Core
Contradicting the trend of the last five years, cloud region launches have slowed significantly. After peaking at over 40 new region launches in 2019, the number dropped to just 13 in 2024 and 16 in 2025.
- Strategic Shift: Hyperscalers are diverting capital away from expanding their geographic footprint into developing markets (such as Africa and the Middle East) and refocusing it on building massive AI compute clusters in developed core markets like the U.S. and Europe.
- Network Impact: Consequently, the “AI boom” has not yet resulted in a massive spike in long-haul or cross-border bandwidth demand. Instead, traffic growth is currently concentrated within campuses and metro dark fiber networks where the training clusters reside.
The Economy Exhibits “Bubble-Like” Circularity
There are concerns regarding the financial sustainability of this boom, characterized by “circular investment” strategies.
A small group of companies (Microsoft, OpenAI, Nvidia, Oracle) are heavily investing in one another to guarantee pipelines and perceived demand. For instance, a tech giant might invest in a startup, which then uses that capital to buy chips from a partner or rent cloud space back from the investor.
This dynamic makes it difficult to gauge true organic demand. Further, the industry is seeing aggressive debt financing, such as Meta’s $27 billion debt financing for a campus in Louisiana, signaling a shift in how these massive capital projects are funded.
The current AI infrastructure boom is like a gold rush where prospectors have stopped buying maps to new territories (halting cloud expansion to new regions) and are instead using all their money to build massive, power-hungry refineries right next to the biggest power plants they can find (AI clusters in West Texas/Ohio). Meanwhile, banks funding the expedition are actually the prospectors lending money to each other, creating a closed loop that makes the operation look busier than it might actually be.
More Cloud and Data Center Insight
More Cloud Convos
Revisit some of our favorite cloud and data center conversations on TeleGeography Explains the Internet:
View the Market Connectivity Score
-1.png?width=600&height=285&name=Mock_Market_Connectivity_Score%20(1)-1.png)
Review TeleGeography’s biannual rankings of the most connected and fastest-growing cities, scored on a scale of 0-100 in each category.
TeleGeography’s Data Center Research Service
The Data Center Research Service features multiple interactive tools and datasets that enable users to pinpoint specific data center sites.
This includes a search interface that filters by location, operator, networks connected, and IX; our famous map search interface; and profiles covering major metro areas, operators, specific sites, and connected networks.
Take a quick tour. ⬇️