Data Centers: Deep Tech’s De-risking Sandbox

As artificial intelligence ( AI ) drives unprecedented energy demands, the data center industry is on the verge of a transformative era. The power consumption of data centers is expected to increase significantly as advanced AI models consume up to 1, 000 times more power than their predecessors ( see Sustainable Metal Cloud’s estimates below ). This quick expansion prompts industry leaders to look for creative ways to decarbonize these energy-intensive hubs, which pose challenges and opportunities for sustainability and administrative efficiency. &nbsp,

Uneven Energy Demands: A U. S. Case Study&nbsp,

As data center power consumption grows, its impact is unevenly distributed. States with considerable data center infrastructure, such as Virginia and Texas, are experiencing accelerated energy demand. As regions struggle to meet growing needs, Lawrence Berkeley National Lab identifies grid interconnection queues, with thermal and energy storage projects delayed. However, fresh power supply remains inadequate, creating an immediate need for revolutionary energy solutions. &nbsp,

Challenges at the Forefront&nbsp,

Speed of Deployment: Technologies that make it possible to quickly deploy a data center in response to urgent AI needs.
1. Energy Optimization: Reducing operating energy consumption for both cost and environmental sustainability.
2. Long-Term Reliability: ensuring data centers can integrate clear baseload power while minimizing risks from outside forces, such as severe weather. &nbsp,

Accelerating Deployment: Modular Data Centers and Localized Energy&nbsp,

With flexible data centers and by locating facilities close to renewable energy sources, innovators are tackling the issue of rapid deployment. &nbsp,

Flexnode specializes in pre-fabricated liquid-cooled flexible data centers. These high-density centers were created to be quickly deployed in challenging environments while reducing water usage. Flexnode even collaborates with the Department of Energy’s ARPA-E Coolerchips program to advance cooling technologies.

Crusoe Energy leverages “digital flare mitigation”, converting wasted healthy gas into energy for AI-driven data centers. Crusoe minimizes transmission losses and maximizes power efficiency by locating data centers close to generation sites. Just last week, Crusoe raised a$ 600M Growth Equity round to advance its low-carbon AI cloud, including financing from Peter Thiel’s Founders Fund. &nbsp, &nbsp,

Cooling Innovation: Tackling a 40 % Energy Burden&nbsp,

Cooling accounts for 40 % of a data center’s energy use, making it a crucial focus for decarbonization. The demand for high-density server racks in conventional air-cooling systems is causing the development of wet cooling solutions. &nbsp,

 

LiquidStack develops both immersion ( single and two-phase options ) and direct-to-chip cooling systems that reduce energy use by up to 95 %. These systems are ideal for high-performance Artificial applications, minimizing both capital and operating costs.

Submer provides single-phase immersion cooling systems, completely submerging servers in non-conductive liquids to optimize heat dissipation.

Jetcool focuses on microconvective direct-to-chip cooling, using specific fluid jets to manage hotspots in AI systems. Their technology is very effective for ultra-dense server environments. &nbsp,

Driving Clear Baseload to Market: Enhanced Geothermal&nbsp, &nbsp,

Geothermal and nuclear power are the mainstays of decarbonizing data centers, and alternative energy sources are necessary. &nbsp,

Fervo Energy uses oil and gas hydraulic fracturing techniques to extract heat from “hot clean rock” using hydraulic fracturing techniques. Partnering with Google, Fervo powers a Nevada facility, supplying fresh energy for Google data centers.

Sage Geosystems is one of the first geopressured thermal systems to combine heat and pressure to produce regular power. By 2027, they intend to supply 150 MW of energy through their partnership with Meta. &nbsp,

On-site Power Management&nbsp,

Beyond cooling, tweaking data center energy usage goes beyond cooling. How energy is managed is being changed by sophisticated power electronics and AI-enabled tools. &nbsp,

Phaidra develops AI-driven software to optimize cooling and energy consumption, achieving up to 30 % energy reductions in pilot programs.

Daanaa develops power transaction units to streamline electricity conversions, reducing energy losses, and reducing heat generation. These systems integrate with renewable, storage, and transmission infrastructures for maximum efficiency. &nbsp,

&nbsp, Grid Resilience Tech Can Tackle Data Center Downtime&nbsp,

The risk of downtime as a result of power disruptions is high in both a financial and service delivery perspective as AI becomes more important in the world economy. Grid hardening technologies are being developed to address issues with power transmission and distribution, which will probably result in a pull-through effect from the expansion of the data center market. &nbsp, &nbsp,

TS Conductor creates carbon hybrid conductors to increase capacity and resilience, preventing infrared sag during high-heat events.

Veir introduces high-temperature superconductors to increase capacity while preventing severe weather by cooling wires privately. &nbsp,

Transformers are essential to maintaining secure energy distribution for data centers, especially as power demand increases rapidly with AI-driven operations. As data centers become more energy-dense, modern motor technologies are emerging to optimize power flow and prevent inefficiencies. &nbsp,

Highly versatile, allowing users to isolate loads and prevent faults, Amperesand is developing solid-state transformers that use semiconductors to convert and control power. They have the ability to be a potent solution for grid resilience against disasters and extreme weather on a global scale. Non-grid customers who have higher levels of power use variability benefit from efficiency and resilience.

IONATE has re-invented transformers ‘ internal processes without re-inventing the form factor, giving them a novel approach but still a drop-in. The company uses an AI layer to control electric power flow and to modulate flow in real-time and with a higher degree of detail. The end result is the addition of 20 % more pole- and wire-connected distributed energy resources and 33 % more grid capacity. &nbsp,

Looking Ahead&nbsp,

Hyperscalers are accelerating innovation and investment in fresh technologies in the data center industry at a crucial time. The ecosystem is rapidly evolving to meet AI’s energy demands effectively, from flexible designs and advanced cooling systems to renewable energy integration. Technologies de-risked in data centers, such as innovative thermal and SMRs, may have far-reaching impacts across industries, from grid optimization to electric vehicles. &nbsp,

We anticipate that the risk of technology de-risking across the continuum of AI infrastructure and data centers will be of high interest in 2025. &nbsp, Keep an eye out for potential analyses on these topics. &nbsp, &nbsp,