3 E Network's 'Vera Rubin-Ready' Data Center: The Blueprint Is the Cheap Part

A Hong Kong company with a market cap of about four million dollars announced this week that it had unveiled the engineering blueprint for its AI data center in Mikkeli, Finland — a facility "designed for" NVIDIA's next-generation Vera Rubin architecture. Before that sentence does its work on you, split it in two, because the two halves have very different claims on your money.

Half the announcement is a real, small building site. The Finnish subsidiary Aurora Core Technology signed a 30-year land lease with the City of Mikkeli, the building permit became legally binding with no appeals, and by April construction had begun — earthmoving and an initial 6-megawatt phase built from modular units. For a microcap, this part is unusually verifiable; the city of Mikkeli confirmed the permit on its own website. Nothing about the site is vaporware.

The other half — the Vera Rubin blueprint — is a PDF, not a product. The release carries the full vocabulary of next-generation engineering: direct-to-chip liquid cooling, 1.6-terabit lossless networking, 48-volt DC power, HGX training zones and MGX inference zones. In substance it is a facilities design document. And here is the number the release does not contain: no total capital expenditure, no compute capacity, no racks, no GPU count. A blueprint is the cheapest statement a company can print. It commits the balance sheet to nothing.

The gap between "designed for Vera Rubin" and "running Vera Rubin" is the whole story, and it is denominated in dollars a four-million-dollar company cannot reach.

Do the per-unit math, because that is where the bluster dies

A single NVIDIA Vera Rubin NVL72 rack is a roughly $9 million piece of hardware that draws on the order of 130 to 230 kilowatts. Scale that against the project's own stated ambition — 26 megawatts over three years under its agreement with Orka Technologies, at least 6 megawatts in year one. At realistic rack power, 26 megawatts is on the order of 130 to 170 racks, or north of a billion dollars of compute silicon — before cooling, power, and the physical building add more on top. Even the 6-megawatt phase one, were it ever populated with Vera Rubin-class racks, runs to the hundreds of millions.

Now hold that up against the whole company. Market cap about $4.35 million; enterprise value about $3.4 million. Revenue in its last reported half was $3.7 million, and it swung to a small net loss from a modest profit as it redirected spending into this pivot. Its financing is a trail of single-digit raises: a $5 million IPO in early 2025, a $460,000 convertible installment, up to $1.3 million in notes and warrants this spring, $1.38 million more in June. Roughly ten million dollars of total capital, against a project that realistically needs one to two orders of magnitude more to hit its advertised spec.

None of this proves the company is a fraud. What it establishes is an evidence boundary. The release's silence on capex, financing, a GPU allocation, an anchor customer, or a signed tenant is missing evidence, not proof of failure. The history — a 25-for-1 reverse split this spring just to keep the share price above the Nasdaq minimum, and dilutive micro-raises to keep the listing alive — is negative evidence about the near-term capacity to fund a from-scratch hyperscale buildout.

Read "designed for Vera Rubin" as ambition, not ownership

There is a wide, expensive river between laying out a facility that could someday hold liquid-cooled racks and actually being granted scarce NVIDIA supply and writing a nine-figure check for it. NVIDIA ships Vera Rubin to hyperscalers and large partners first; small operators must fund both the silicon and the demand for it. The company that designs a data center is not the company that gets to fill it.

For a retail investor, the honest takeaway is a sizing problem, not a valuation puzzle. This is a roughly four-million-dollar nano-cap whose story borrows the most exciting technology name in the market and whose actual business — a few million dollars of software-services revenue — could not fund one Vera Rubin rack on its own. When a press release is rich in engineering detail and empty of dollars, the engineering is doing narrative work. The part that would change the case is exactly the part a blueprint cannot supply: a financing commitment and a customer.