Applied Digital의 전력 자회사가 조용히 IPO를 신청한 가운데, 회사의 부채 부담이 322% 급증한 매출을 가리고 있다.
Building server farms for artificial intelligence was long treated as a specialized real-estate play: put up the halls, lease them to hyperscalers, collect steady rent. That framing no longer holds. The binding constraint across the industry has shifted from concrete to raw electrical capacity measured in gigawatts, and Applied Digital is demonstrating how far the logic of energy self-sufficiency now reaches.
Reuters reported that Base Electron, the power-plant company incubated by Applied Digital to supply its AI campuses under dedicated arrangements, has confidentially filed for a U.S. listing. The paperwork discloses neither share count nor price range, but the strategic direction is clear: anyone selling AI compute must eventually own the power plants behind it.
How concrete that entanglement already is became clear from an agreement dated October 4. For the flagship Polaris Forge 3 project, Applied Digital signed a 15-year power purchase contract with Base Electron covering capacity from a planned roughly 1,200-megawatt natural gas facility in Center, North Dakota. The contract carries a firm condition: should Base Electron fail to close financing by March 31, 2027, the arrangement can be terminated.
Expansion is proceeding at considerable speed. A further build-out phase at the Polaris Forge 1 site went live on October 2, adding 75 megawatts and lifting operational capacity at the fully leased campus to 250 megawatts, against a contracted eventual total of 400 megawatts. The company is also expanding internationally. On Tuesday it disclosed access to as much as 1 gigawatt of potential power capacity in Finland from 2028, its first venture outside the United States, with talks underway with hyperscale customers.
That aggressive buildout carries a financial cost. In first-quarter fiscal 2027 results released Wednesday, revenue surged 322% year over year to $341.9 million, while adjusted EBITDA came in at $64.4 million. Net loss attributable to common stockholders, however, reached $221.0 million. The shortfall stemmed largely from higher operating costs, increased interest expenses and fair-value adjustments. A positive result on an adjusted operating basis pales next to the financing burden, and with rising Treasury yields making capital more expensive across the sector, companies funding billion-dollar construction projects can quickly find themselves in choppy waters.
The balance sheet tells its own story. As of August 31, 2026, liquid and restricted cash of $3.7 billion stood against $6.4 billion in debt.
Analysts are divided. Needham's John Todaro trimmed his price target to $70 on Thursday while keeping a buy rating. Morgan Stanley raised its target to $41 on Friday and stayed neutral. Both firms cited shifting credit conditions for AI infrastructure and the value of the site pipeline.
That gap between the two houses captures the sector's central dilemma. Bulls point to the long-term worth of contracted capacity; skeptics worry about an insatiable appetite for capital. Refinancing the costly builds is the decisive yardstick. Absent dependable cash flows, even the grandest blueprint stays exposed to market turbulence. The international push adds opportunity alongside fresh operational uncertainty in unfamiliar jurisdictions, where large projects require local acceptance and do not always proceed without friction.
The stock closed Friday at EUR 21.16, 24% below its 200-day moving average of EUR 27.86 and 51% beneath its 52-week high of EUR 43.48. The scale of that discount reflects lingering doubt over how the coming investments will be funded.
If the confidential listing of power subsidiary Base Electron is meant to channel fresh capital into the company, it underscores a simple truth: without billions in outside funding, even the fastest growth story of the AI era quickly hits its limits. Demand for compute remains enormous. The open question is who ultimately foots the bill for the power plants.