In its Annual Energy Outlook 2026, the U.S. Energy Information Administration projects electricity consumed by data center servers will increase across the commercial building stock – more in standalone data centers than in all other data center rooms combined. By 2050, server consumption will reach between 446 billion BkWh and 818 billion BkWh, with servers accounting for up to 33% of all U.S. commercial sector electricity use by mid-century. 

Acting as a bridge between utility grids and power-hungry server infrastructure will be transformers, but not your traditional iron and copper transformers, as they won’t handle the jump in per-rack densities from 10 kW to120 kW (or to what some say could be as high as 600 kW).

Now that the industry codified 800V DC as the universal standard for next-generation AI data halls, an era of solid state transformers will be ushered in for modern data centers, which will have to handle trillion-parameter generative AI workloads, as well as heavy LLM training, and real-time AI reasoning.

The NVIDIA superchips that are on the horizon will bring with them crushing workloads and physical bottlenecks, so data center operators will need precision power control. Researchers like Grand View Research say that need will drive the SST market to grow from about $169.4 million now to about  $935.8 million by 2030, a CAGR of 32.0%.

Rich Miller, who RCRTech interviewed on AITechTalk, recently wrote that SSTs powered by silicon carbide or gallium nitride devices could reduce data center electrical equipment footprints by as much as 80%, with advanced power electronics and software-defined logic that enable switching at ultra-high frequencies, as well as bidirectional routing of power, and almost instantaneous conversion of utility-level power to direct server DC.

Despite those advantages, SSTs are still in the nascent phase, with commercial deployments just beginning. There are still serious pricing and efficiency issues to address, which you can read about in today’s story, “Solid state transformers for ultra-high-density workloads.”

AI scale-out networking: Ciena’s Helen Xenos talks to RCRTech about AI networking and the coexistence of coherent pluggables, performance transponders, full-spectrum systems, co-packaged optics, and liquid cooling approaches.

NTT, Citigroup: NTT Global Data Centers seeks to raise $1 billion in a deal that reflects a broader trend, where developers establish dedicated investment vehicles for early-stage projects – separating pre-revenue assets into standalone entities.

View More News

Salem Township $1.2B land deal: 4-3 Glen Brook Group is applying for zoning approval of a 2nd data center project in Salem Township in PA, where 194 landowners are selling a collective 4,000 acres, on which 39 buildings Potential power will go up. 3.5 to 4 GWs.

SpaceX $60B Cursor deal: SpaceX has entered the AI coding race with a $60 billion acquisition of Cursor. The all-stock transaction is expected to bolster SpaceX efforts to compete with Anthropic and OpenAI, which offer popular coding tools.

Underwriting data centers: First Street examined 97 global data center markets and found that 80% of capacity is at elevated risk to climate hazards like flooding, fire, heat and drought. Underwriting using historical records no longer works.

Microsoft IN data center: LaPorte city leaders joined Microsoft officials for the groundbreaking of a 489-acre data center on the city’s south side. The first building will be completed spring of 2029, with another by mid-spring, and a third by winter.

Brewer in semi manufacturing: Brewer Science will acquire the semiconductor chemicals business line of Heraeus Epurio, bolstering its leadership in ultrapure electronic chemicals used in semiconductor manufacturing.

RCR Events