Episode Summary
Executive Summary: David Roberts interviews Verus CEO Nelson Abramson about a new data-center design that aims to be quieter, use far less water, and draw more compute from the same grid connection. Abramson explains Verus’s integrated approach: medium-voltage power distribution, large LFP batteries for backup/grid flexibility, and priority-aware load management. The conversation also probes community opposition, land-use politics, noise/water claims, and whether the model can scale without diesel backup or lost reliability.
Main Topics: Verus’s integrated data-center redesign (Priority: 5/5): Abramson argues data centers should be treated as energy assets, not just industrial real estate, and that combining power, storage, cooling, and controls into one system yields better performance than bolt-on upgrades. Medium-voltage power plane and rack-level flexibility (Priority: 5/5): The company’s power architecture keeps electricity at medium voltage longer, reduces copper and conversion losses, and allows power to be shared across the whole facility instead of being locked into fixed buckets. Battery backup and grid flexibility (Priority: 5/5): Verus uses large LFP batteries rather than diesel gensets for primary backup and demand-response capability, while keeping small local UPS-style batteries for sub-second ride-through. Priority-aware compute capping (Priority: 4/5): If battery storage is depleted during long grid events, the tenant can throttle lower-priority compute workloads while preserving more critical processes, similar to what cloud operators already do internally. Noise, water, and community impacts (Priority: 4/5): Abramson claims closed-loop air cooling, sound walls, large slow fans, and adiabatic cooling make the sites much quieter and far less water-intensive than conventional data centers, but Roberts presses for verification. Political opposition and land-use conflict (Priority: 4/5): The interview explores why data centers face strong backlash despite industrial zoning and promised jobs/tax revenue, especially in Michigan and Oregon where Verus projects have faced delays and controversy. Ride-through standards and future innovation (Priority: 3/5): Both discuss mandatory ride-through for grid stability, possible loss of competitive advantage if all operators must comply, and next-generation ideas like DC distribution, better communication pathways, and new storage tech.
Key Arguments: Verus’s core thesis is that a data center is really an energy conversion system; redesigning it holistically can extract more compute from the same utility interconnection. The medium-voltage power plane reduces losses, copper use, and capacity fragmentation, enabling roughly one-third more racks in the same facility footprint/capacity envelope. Large LFP batteries are better suited than diesel generators for sustained outages and grid services; they also make the facility quieter and cleaner. Small server-level batteries remain necessary for microsecond/sub-second ride-through because large batteries are not optimized for that function yet. Priority-aware capping is not a hard requirement for tenants because most large operators already have workload prioritization, batching, and scheduler controls. Closed-loop air cooling with carefully designed fans, berms, and sound baffles should keep noise around library-like levels at the property line. Water use is mostly limited to occasional adiabatic cooling on the hottest days plus ordinary domestic/landscaping needs; the system is intended to use 95%–99% less water than conventional evaporative data centers. Public opposition is real but should be addressed with concrete performance claims, community benefits, and industrial-zoned siting rather than by abandoning data-center development altogether. Ride-through is necessary for grid stability at very large loads, and Verus supports standards that require it across the industry. Future innovation likely lies in deeper integration among data centers, utilities, and compute schedulers, plus new storage, cooling, and possibly high-voltage DC distribution.
Data Points: Career length in data centers: ~30 years total; 20 years at Google - Abramson describes his background before founding Verus First data center size: <1 MW - He recalls working in a converted Hechinger’s hardware store Next-generation rack load: 1 MW+ per rack - Used to illustrate how much power density has changed Typical data-center growth: 50x–100x larger - Abramson says data centers have scaled dramatically over decades Target compute utilization: Up to 77% of provisioned utility capacity - Verus claim for modeled design vs conventional design Conventional design utilization: About 49% - Comparison baseline in Verus’s modeling Water savings: 95%–99% less water - Claim for air-cooled/closed-loop system versus traditional evaporative cooling Battery runtime: 4 hours nominal - Designed for worst-case demand response/outage planning day Demand-response window: 2–4 hour strips - Abramson says this matches utility/RTO market design Battery chemistry: LFP - Used for Detroit project stationary storage Community benefits package: $10 million - Michigan project package cited in discussion Infrastructure and resiliency fund: $5 million - Part of the $10 million package Public safety fund: $2 million - Part of the community benefits package Education and workforce fund: $3 million - Part of the community benefits package Permanent jobs: 210 - Projected operational jobs in Michigan Property-tax revenue: ~$300 million over 17 years - Michigan project estimate Salem property-tax base increase: Roughly one-third - Abramson’s estimate for the Oregon project Water use estimate: Under 12 million gallons annually - One cited Oregon figure for the Salem project Water comparison: Like three Cheesecake Factories - Abramson’s analogy for 12 million gallons Local land history: Industrial since 1957 - Michigan site has long been designated industrial land Sound target: Library-like at property line - Modeling claim for the Michigan site
Pivotal Quotes: "The most efficient data center is the one you don't have to build." — Nelson Abramson: Explaining why higher utilization from the same interconnection is a core design goal "We're not particularly purely ideological about the particular battery chemistry." — Nelson Abramson: Discussing why Verus chose LFP batteries for the Detroit project "Data centers underpin the technology that we all take advantage of every day." — Nelson Abramson: Defending the social value of data centers amid public backlash
Implications: If Verus’s model works, future data centers could be quieter, far less water-intensive, and more grid-responsive. But the episode also shows that technical improvements alone may not resolve community distrust, siting politics, or the industry’s broader legitimacy problem.