From a global AWS outage and massive AI cluster roll outs to energy partnerships, local development battles, and supercomputers, it’s been another eventful couple of weeks in the data center sector. Here’s a roundup of the latest news stories shaping the industry.
Global AWS Outage
A major Amazon Web Services outage in the US‑EAST‑1 region on Monday served as another reminder of how dependent the web has become on a single cloud giant. Platforms from Snapchat, Signal, and Fortnite to Canva, Airtable, and Ring Cameras were knocked offline as Amazon cited “significant error rates for requests made to the DynamoDB endpoint.” One expert told CNN the financial impact of the outage could total hundreds of billions of dollars. Services largely recovered by the end of the day, but the incident underscores that hyperscalers aren’t immune to large-scale disruption – and that regional redundancy is no longer optional.
AI Infrastructure Scaling and Spending
OpenAI is continuing its ambitious expansion, this time partnering with Broadcom to co-develop custom AI accelerators and Ethernet solutions, aiming to deploy 10 GW of chips by 2029. Meanwhile, Microsoft has rolled out its first “supercomputer-scale” GB300 NVL72 Azure cluster, linking 4,608 GPUs for 92.1 exaFLOPS of FP4 inference, marking a major step in hyperscale AI infrastructure. The cluster is set to power OpenAI workloads. OpenAI and Oracle also announced that the Stargate campus in Abilene, Texas will eventually house 450,000 Nvidia GB200 GPUs and consume 1.2 GW of power.
Internationally, Google is making major commitments in AI infrastructure. The company has committed $15 billion to build an AI hub in Andhra Pradesh, India, in partnership with AdaniConneX and Airtel. In Belgium, meanwhile, Google is investing €5 billion over the next two years to expand its AI infrastructure, one of the largest single tech investments in the country’s history.
Expansion Meets Local Resistance
At a more local level, Google has also been busy. The tech giant is committing $9 billion to expand its South Carolina campuses and has acquired 950 acres in Monroe County, Georgia for potential future data center buildout. Lastly, in Linn County, Iowa, Google is looking to build up to six data centers near a decommissioned nuclear reactor that it hopes to help restart. The company is funding a water usage study to assess potential impacts on local Linn County resources as part of its broader $7 billion Iowa expansion.
Meanwhile, not every project is advancing smoothly: Microsoft has canceled its Caledonia, Wisconsin campus due to community opposition (Fox6), and local data center pushback made headlines in Maine, Michigan, North Carolina, Kentucky, and Virginia. The dichotomy between hyperscale ambitions and local scrutiny remains a defining tension.
Carbon Capture, Natural Gas, and Grid Strain
In energy news, Prometheus Hyperscale says it’s working with carbon capture technology firms to make its 1.5 GW AI campus near Casper, Wyoming completely carbon negative, despite it being powered primarily by natural gas. And speaking of natural gas, Fermi America has secured a natural gas pipeline interconnection for its planned 11 GW Amarillo, Texas campus, while VoltaGrid agreed to supply Oracle with 2.3 GW of modular natural gas power for its AI data centers in Texas.
At the same time, Apple added over 650 MW of renewable capacity across Europe, mobilizing more than $600 million in financing. And Bloom Energy and Brookfield have signed a $5 billion partnership to deploy fuel cell technology at AI data centers globally.
S&P Global forecasts U.S. data center grid demand will rise 22 % in 2025 and nearly triple by 2030. Lastly, renewables overtook coal as the biggest source of electricity worldwide in the first half of 2025, according to a report from global energy think tank Ember. The biggest increases were seen in China, where solar generation grew by 55%.
The Season of Supercomputers
In innovation news, this recent stretch has been all about supercomputers. MIT Lincoln Laboratory launched TX-GAIN, a 13.39 petaflop Nvidia H100-powered supercomputer for biodefense, materials research, and national security projects. MIT claims it’s the most powerful AI supercomputer at any US university. New York University, meanwhile, unveiled Torch, a 10.79 petaflop AI-optimized HPC system with liquid cooling for improved efficiency, claiming it’s the most powerful high-performance computing system in the state of New York. Oracle also announced plans for its Zettascale10 cloud AI supercomputer, aiming for a massive 16 zettaflops across multi-gigawatt clusters.