WHY WE STARTED ELYRIAN
In 1920, about one in three American homes had electricity. The demand existed long before the grid did, and it took utilities two decades to close the gap.
AI is in a similar position. Compute is being ordered faster than power can be delivered to run it, and most of that compute will be spent on inference rather than training.
Elyrian builds the sites that close the gap.
WHAT WE BUILD
Elyrian develops inference data centers under 100 MW.
Instead of designing each site from scratch, we build to a frozen design: standard power, electrical and building modules, added in 10 MW increments. A site is assembled from repeated blocks rather than engineered as a one-off.
Repetition changes how a project runs. The supply chain is the same every time. The engineering is done once and reused. At this scale a site sits inside a community without reshaping local utility rates, and it stays below the thresholds that pull federal regulators into the process. We target twelve months from start to live power, against the four to five years a bespoke campus typically takes.
HOW WE WORK WITH CUSTOMERS
We develop the site, bring the power, electrical and building modules on site, take them through to completion, and commission them. We then maintain the site for fifteen years.
OUR INVESTORS
We’re fortunate to be backed by Underscore VC, NewBuild Venture Capital, Afore Capital and Outcast Ventures, alongside operators including Srinivas Narayanan, former CTO of B2B at OpenAI and VP at Meta, and Manik Gupta, former CPO at Uber.
CONTACT US
If any of the above resonates, we’d love to chat.