300 Square Feet, 500/500, and the First Rack

By November 5, 2022, Helixrack had a room, a business internet circuit, and one working rack. None of them were large. Together, they were enough to test whether customer-owned servers could be operated with clear boundaries and repeatable handling.

The first space was a roughly 300-square-foot ground-floor room in Elizabeth. It had a concrete floor, brick walls, limited lighting, and none of the later facility’s cooling or backup-power systems. The point was not to make the room look finished. The point was to make the first installation controlled and inspectable.

Preparing the room

Power work came before equipment. The facility record describes new grounded branch circuits, conduit, a small distribution panel, and a basic load plan. The original room had very little usable electrical infrastructure, so the first safe capacity was intentionally narrow.

Those details matter because an empty receptacle is not the same thing as available server capacity. Circuit rating, continuous-load limits, cable routing, power-supply redundancy, and heat removal all reduce what can be used safely. Helixrack treated the first room as a bounded test environment rather than advertising every theoretical ampere.

The first network edge

A 500/500 Mbps business fiber service was installed before the first rack went live. The facility received a provider-assigned IPv4 /28. That block belonged to the facility edge; it was not divided into customer /27s. The early design assigned a single public /32 to each customer on a dedicated VLAN, with routing and policy enforced at the shared edge.

The distinction was important. A public address was part of a controlled service configuration, not an invitation to place unrelated systems on one flat network. The small address pool also forced Helixrack to document allocations from the beginning.

Initial switching and routing used repaired or repurposed equipment. This was economical, but it was not meant to excuse unknown condition. Hardware entered service only after inspection and testing, and configurations were kept simple enough to review. The 500/500 circuit later remained useful even after the company added a larger network edge.

One two-post rack

The first rack was a used 42U two-post relay frame. The inventory places a Supermicro compute system, two Dell PowerEdge R620 servers, a Cisco Catalyst 3750G switch, a switched 30A PDU, and a 1500VA UPS in or alongside the initial setup. Patch leads and cable management came from the same accumulated stock of serviceable parts.

The 1500VA UPS should not be confused with the larger double-conversion system installed in 2023. In the mature power design, that small unit protects network and control equipment only. It was never evidence that the original room could sustain the full server load through an outage.

One item in the inventory reflected the market that motivated the company. NVIDIA had announced that its A100 data-center GPU was in full production in May 2020, so A100-based systems were period-correct by late 2022. That external announcement establishes availability, not the provenance or condition of Helixrack’s unit; purchase and serial records must do that.

What “live” meant

On November 5, “live” meant the rack could be powered, monitored, and reached over the business circuit. It did not mean the room had production cooling, generator coverage, multiple rows, or the later Helixrack identity. There were no customer servers in a scaled facility because there was not yet a scaled facility.

The value of the first rack was narrower and more useful: it turned the founding thesis into an operating checklist. Receive equipment. Record ownership. Inspect it. Burn it in. Assign power and network deliberately. Keep the customer configuration isolated. Know who responds when something changes.

That checklist, rather than the size of the room, was the first piece of infrastructure Helixrack could carry forward.