Enovum and the Arithmetic of the Montreal Data Center

A data center is often described as real estate with power. That description is too simplistic for Enovum Data Centers Corp. A more precise definition would be: contractable electricity, cooling discipline, network access, and operational trust built into a leased or retrofitted building. In Montreal, that package has a particular arithmetic. Electricity is cheaper and cleaner than in most competing North American metropolitan areas, the winter climate improves cooling economics, and latency to Canadian and northeastern US users is acceptable for many enterprise and inference workloads. But those advantages are not permanent rents. They are inputs that must be priced, reserved, permitted, cooled, and sold before larger operators, hyperscale campuses, cloud platforms, or utility policies erode the margin.

Enovum is not a hyperscale giant. It is not the deepest interconnection platform in Montreal. It is not simply a hosting reseller with a data center page. It is a Canadian data center operator with visible facilities in Montreal, a GloboTech lineage, official network resource registrations, pre-acquisition audited financial statements, and a post-2024 role within WhiteFiber, the AI infrastructure business controlled by Bit Digital. The economic significance of the company lies in its position amid a market shift: old enterprise colocation relied on cabinets, cages, cross-connects, and remote hands; new AI/HPC colocation is built on high-density power, liquid cooling, rapid deployment, custom builds for the customer, and the ability to convert an industrial building into energized compute before the window closes.

This article's central claim is precise. Enovum's strategic value in Canada does not lie in owning the largest data center portfolio, nor in out-interconnecting Cologix, nor in having an especially defensible brand. Its value lies in having operated, sold, and expanded Montreal data center capacity at the exact moment when Hydro-Québec power, city-proximate AI infrastructure, and enterprise colocation have become scarce enough to hold platform value. The weakness is symmetric: if incremental power becomes expensive, delayed, or politically constrained, Enovum would revert to being a small, competent operator within a capital-hungry AI infrastructure group, rather than a scarce Canadian platform.

The company behind the legal name

The public identity begins with a simple government registry entry. Canada's Investment Canada Act page lists Enovum Data Centers Corp. in Montreal, Quebec, and describes it as a company that "designs, builds and operates data centers." This confirms Canadian operational geography and basic business activity, but does not reveal ownership conditions, national security commitments, site economics, or customer contracts. (Innovation, Science and Economic Development Canada)

The audited financial statements filed with WhiteFiber's SEC documents are more precise. They state that Enovum Data Centers Corp. was incorporated under the Canada Business Corporations Act on July 13, 2023, and that it designs, develops, and operates high-performance computing data centers offering hosting and colocation services. The same filing notes that Enovum's revenue comes from customers' use of physical space, electricity, and cooling within the data center facilities.

That incorporation date should not be mistaken for operational history. Enovum's market identity predates the 2023 legal date. Its 2021 launch announcement stated that the team had been involved in data centers since 1999 through GloboTech Communications, a Montreal hosting provider. The same announcement described Enovum's first phase in Montreal as a transparent colocation facility powered by renewable hydroelectricity, with a customer portal offering real-time utilization data and advanced analytics. (News wire service)

Local Quebec press completed the founding narrative. TVA Nouvelles described Enovum-MTL-1 as a CAD 200 million data center project launched without government subsidy, linked it to Pierre-Luc Quimper's background at GloboTech, and quoted the initial positioning against larger, foreign-capital-backed competitors in Quebec. The article described the project as "100% private," stated that the first phase was already leased, and framed Enovum as a Quebec-owned rival to Vantage, eStruxture, and Cologix. (TVA Nouvelles)

Ownership changed in 2024. Bit Digital announced it had acquired Enovum for approximately CAD 62.8 million (about USD 46 million), with CAD 56 million in cash and approximately 1.62 million equivalent shares issued to key management to reinvest a portion of their stake. Bit Digital described Enovum as an owner, operator, and developer of HPC data centers, with a fully operational and fully leased 4 MW Tier-3 data center in Montreal. (Bit Digital)

This creates three identities, all economically relevant. The first is the official Canadian corporation. The second is the Montreal/GloboTech operational lineage. The third is the WhiteFiber/Bit Digital AI infrastructure platform. A customer buying a rack or a high-density deployment cares mostly about the second identity: whether the operator can keep power, cooling, network, and remote hands running. An investor or strategic buyer cares mostly about the third: whether Enovum can become a repeatable way to capture AI infrastructure margin. A regulator cares about the first: who owns and operates Canadian data center infrastructure, and under what power-allocation regime.

What Enovum actually sells

Enovum sells colocation, but "colocation" hides the economics. The audited financial statements define the revenue model as recurring revenue from cabinet space, electricity, and connectivity, including cross-connects, plus non-recurring installation services related to initial customer deployments. Contracts typically range from one to five years, billed monthly, with revenue recognized over the contract term. Service-level failures can reduce revenue through credits or cash payments.

In practice, the customer buys the right to place equipment in a controlled environment and offload various operational risks: power redundancy, UPS and generator backup, cooling, fire suppression, physical security, cable management, carrier access, remote hands, environmental monitoring, and incident response. Enovum's MTL1 technical page describes 70,000 built square feet, rack/cage/private data suite options, diverse meet-me rooms, 24/7/365 remote hands, 4 MW ready plus 20 MW expansion capacity, 2N UPS and generators, 48 hours of generator fuel, N+1 cooling, biometric access, and power and cooling monitoring for customers. (Enovum Data Centers)

The sales pitch is service-oriented. Enovum markets itself as a vertically integrated data center builder and operator that can construct custom-fitted spaces to customer requirements, provide transparent real-time utilization data, and offer carrier-neutral access to major local providers via redundant dark fiber. The home page also highlights the Montreal location, hydroelectricity, free-cooling potential, and scalable high-density rack infrastructure. (Enovum Data Centers)

This is a different proposition from generic wholesale space. A hyperscale buyer may want 20, 50, or 100 MW with standardized designs, aggressive pricing, and predictable expansion rights. A small business may want a few racks and predictable monthly billing. An AI infrastructure customer may want 50 kW to 150 kW per rack, direct-to-chip liquid cooling, rapid construction, and a contract that treats electricity as the product, not a footnote. Enovum's opportunity sits between these categories: large and technical enough to host high-density workloads, but still small and flexible enough to sell custom colocation rather than just standardized megawatt blocks.

The early "transparent colocation" pitch matters because data center customers often distrust opaque billing. Electricity overages, remote-hands charges, setup fees, and cross-connect costs can turn a nominally cheap colocation quote into a substantially different monthly bill. Enovum's launch material specifically emphasized transparent infrastructure and a customer portal. The audited financials later show connectivity services, including cross-connects, as part of revenue. This is not necessarily a contradiction: "transparent" does not mean "free," and an initial launch offer can differ from later normalized pricing. But it highlights the central margin tension. Cross-connects and connectivity are often high-margin colocation revenue lines; forgoing or simplifying them can help win customers but reduces one of the classic carrier-hotel profit centers. (Enovum Data Centers)

The scarce asset is not the building

Montreal has many industrial buildings. It does not have unlimited, ready-to-use data center power. That is the main economic fact underlying Enovum.

Quebec's appeal is clear. A note from the Quebec National Assembly library indicates the province had 67 data centers in July 2024, up from 39 in 2019, with about 130 MW of Hydro-Québec peak demand contracted by data centers, excluding crypto use. The same note cites a 2021 comparison where a large electricity consumer in Montreal paid 5.24 cents/kWh, versus 9.57 cents/kWh in Ottawa, 10.64 cents/kWh in Seattle, and 20.22 cents/kWh in Boston. It also notes the province's cold climate and hydroelectricity as structural draws. (PREMIÈRE LECTURE)

These advantages explain why a Montreal colocation operator can be relevant despite not being in Northern Virginia or Silicon Valley. A workload that needs to serve Canadian users, satisfy Canadian or Quebecois location preferences, or operate on low-carbon electricity can accept Montreal as a strong middle-ground solution: cheaper and cleaner than many US metropolitan areas, closer to users than remote hydro or wind sites, and with sufficient connectivity for enterprise and AI inference use cases.

But cheap electricity has become politically scarce. In February 2026, Hydro-Québec proposed a new rate for large data centers and blockchain users, stating that data center electricity consumption could grow sevenfold by 2035, to over 1,000 MW. For new data center customers with demand above 5 MW, the proposed rate is 13 cents/kWh on average, about double the current large-consumer rate, and projects over 5 MW will remain subject to a selection process. Existing data center customers would receive a five-year transition rate. (Hydro News)

This is not a minor rate detail. At a PUE of 1.3, one megawatt of continuous IT load requires roughly 1.3 MW at the facility level. Over a year, that equals 11.388 million kWh. At 5.24 cents/kWh, the electricity cost is about CAD 597,000 per MW-year. At 13 cents/kWh, it is about CAD 1.48 million per MW-year. The difference is about CAD 884,000 per MW-year. For a 5 MW AI deployment, that is about CAD 4.4 million per year before demand charges, contractual pass-throughs, taxes, or site-specific rate treatment.

That sensitivity explains the commercial value of capacity that is already energized or has advanced approval. A building without usable electricity is optionality. A building with approved electricity is inventory. A building with approved electricity, cooling, carrier access, and a signed AI customer is strategic infrastructure.

Hydro-Québec's regular rate materials reinforce this point. The LG rate applies to customers with a minimum annual contract billing demand of at least 5,000 kW, unless they qualify as primarily industrial; the rate tables separately list demand and energy components, meaning the effective cost depends on both energy consumption and reserved demand. (Hydro-Québec) (Hydro-Québec) A data center operator is therefore not just buying electrons. It is asking the utility to reserve capacity around the clock, in a province trying to allocate hydroelectricity among electrification, industrial policy, decarbonization, and AI.

MTL1: small enough to understand, full enough to matter

MTL1 is the operational proof. Bit Digital's acquisition announcement stated that Enovum operated a 4 MW Tier-3 data center in Montreal, powered by renewable hydroelectricity, leased through 2036 with two five-year extension options, fully leased to more than a dozen colocation customers, and expected to generate approximately CAD 10 million in revenue in 2025. (Bit Digital)

WhiteFiber's 2025 annual report later described MTL1 as a 4 MW gross Tier-3 HPC data center in Montreal that was fully operational and fully leased at the time of acquisition. Its investor materials described MTL1 with 3 MW of IT load, fully occupied by 14 customers, with an average customer tenure of 2.5 years. The difference between the 4 MW gross and the 3 MW IT is important: public claims about data center capacity often conflate gross power, IT load, installed load, sold load, and expansion potential. Commercial assessment must follow the exact denominator. (Securities and Exchange Commission (SEC))

The pre-acquisition financials show a small but real fixed-asset business. For the period ended September 30, 2024, Enovum reported USD 3.36 million in colocation revenue, USD 1.26 million in cost of colocation revenue, USD 789,000 in depreciation and amortization, USD 1.02 million in general and administrative expenses, USD 285,000 in operating income, USD 323,000 in finance costs, and USD 290,000 in net income after a tax benefit. Total liabilities were USD 17.21 million against total assets of USD 19.29 million.

Site-level margin was attractive before corporate expenses and depreciation. Cost of revenue represented about 38% of revenue, implying gross contribution of around 62% before depreciation. But operating income was only about 8.5% of revenue, and finance costs exceeded operating income. That is the pattern of a young colocation business: high utilization can generate site-level margin, but capital, depreciation, lease liabilities, and overhead absorb much of the spread.

The disclosure of remaining performance obligations provides another view of customer durability. As of September 30, 2024, Enovum had approximately USD 12.33 million in remaining performance obligations, including USD 6.43 million expected in 2025 and USD 3.59 million in 2026. This is enough future revenue to demonstrate real contracts exist, but not enough to justify the CAD 62.8 million acquisition price as a simple multiple of existing contracted revenue. The buyer paid for operational capability plus expansion capacity, not just the current cash flow of MTL1.

Customer concentration is visible but not fully quantifiable. The audited financials state that about 85% of trade receivables as of September 30, 2024 came from three customers. Receivables concentration is not the same as annual revenue concentration; a few late invoices can skew a point-in-time figure. Still, the signal is clear: a small, full facility can have concentrated commercial exposure even when it advertises a diversified customer base.

The WhiteFiber pivot

Bit Digital bought Enovum to vertically integrate. Its acquisition announcement states that Enovum would allow Bit Digital to capture margin that would otherwise go to third-party data centers, place its own GPUs, add colocation and on-demand compute services to its GPU offering, and improve time-to-market for customers. Management specifically argued that sites adjacent to metropolitan areas are important for inference because latency is a primary concern. (Bit Digital)

WhiteFiber's annual report lays out the same strategy in more formal language. The company describes itself as an AI infrastructure provider that owns HPC data centers and offers cloud GPU services. It states that AI and machine learning workloads demand higher power density, advanced cooling, and robust bandwidth; it also states that operating its own data centers can reduce reliance on third-party providers and capture additional margins. (Securities and Exchange Commission (SEC))

This changes how Enovum should be valued. Before Bit Digital, Enovum was a Montreal colocation operator with expansion optionality. After Bit Digital and WhiteFiber, it became a Canadian operating platform within a broader AI infrastructure stack. The economic question is no longer simply "how many racks can MTL1 sell?" It is "can the Enovum team repeatedly identify, secure, retrofit, energize, and fill city-proximate facilities for AI workloads before power, capex, or customers go elsewhere?"

WhiteFiber's site selection criteria are explicit: proximity to metropolitan areas, partial infrastructure, retrofits rather than new development, existing underutilized electricity, low-latency inference needs, and a targeted build time of about six months from start of construction. The filing also states the company prioritizes sustainable or secured electricity and looks for sites where power can ramp over time. (Securities and Exchange Commission (SEC))

The incentive structure supports this interpretation. WhiteFiber's 2025 annual report indicates that Billy Krassakopoulos, President and CEO of Enovum, has restricted stock units (RSUs) tied to Enovum's "growth EBITDA" from new data centers, equal to 4% of incremental growth EBITDA, up to a maximum of CAD 4 million in RSUs. That compensation design makes the business objective unambiguous: grow EBITDA from new centers, not just operate MTL1 conservatively. (Securities and Exchange Commission (SEC))

MTL2: The building is not the business

MTL2 is the warning asset. In December 2024, Bit Digital announced the acquisition of a 160,000-square-foot former encapsulation manufacturing facility in Pointe-Claire for CAD 33.5 million. The company expected to invest about CAD 27.6 million to convert it into a 5 MW Tier-3 data center expansion project, with direct-to-chip liquid cooling capabilities and rack densities up to 150 kW. (Bit Digital)

On paper, that is exactly the retrofit strategy: buy an industrial building near Montreal, convert it into AI-ready colocation, leverage hydroelectric power and cold climate, and bring capacity to market quickly. However, in WhiteFiber's subsequent 10-K, MTL2 was not presented as a simple execution story. The filing states that the company acquired the site and expected to develop it to Tier-3 standards with an initial 5 MW gross load, but "prioritized other builds and preserved capital for more pressing projects." A risk factors section reiterates that MTL2 fell behind other builds and warns that changes in the business model may not succeed. (Securities and Exchange Commission (SEC)) (Securities and Exchange Commission (SEC))

That disclosure is more useful than a success press release. It shows the bottleneck. Data center expansion is not just "find a building, add generators, and sell racks." The operator must sequence electricity availability, customer demand, capex, financing, equipment lead times, construction labor, cooling design, and regulatory approvals. A 160,000-square-foot building can be a liability if the best customer-backed returns are elsewhere.

MTL2 also shows why the announced project pipeline must be discounted. Bit Digital's acquisition announcement described an Enovum proprietary development pipeline of 288 MW, including 93 MW under letters of intent with property owners. WhiteFiber later described a much larger gross pipeline of 1,500 MW under management review, cautioning that timelines and capacities are subject to change. A project pipeline is not capacity. A letter of intent is not electricity. Management review is not an energized data floor. (Bit Digital) (Securities and Exchange Commission (SEC))

MTL3: The AI test

MTL3 is the opposite of MTL2: a more convincing proof because it links the site, the customer, and the revenue. In April 2025, Bit Digital said it had secured rights to a site in Saint-Jérôme to be developed for Cerebras under a 5 MW colocation agreement. The site is about 202,000 square feet on 7.7 acres, structured as a 20-year lease-to-own operation with a fixed-price purchase option, and was expected to require roughly CAD 55 million in development capex. The same announcement said future expansion would be subject to utility approvals. (Bit Digital)

WhiteFiber's annual report states that MTL3 is being developed as a 7 MW gross Tier-3 data center that will support Cerebras with 5 MW of IT load, and that future expansion is subject to utility approvals. It also says that MTL3 was retrofitted and went live in November 2025, and that billing to Cerebras began on November 1, 2025, at a rate of CAD 1.4 million per month under a five-year contract. (Securities and Exchange Commission (SEC))

This is the clearest commercial evidence for the AI colocation thesis. CAD 1.4 million per month for 5 MW of IT load equals CAD 16.8 million per year, or CAD 3.36 million per MW-year of IT load. The public filing does not disclose the exact electricity pass-throughs, customer equipment obligations, cooling scope, capex allocation, margins, or termination rights. But the revenue density is enough to explain why WhiteFiber wants to go beyond traditional cabinet colocation.

Cerebras is not an ordinary enterprise cabinet customer. It announced plans for six new AI data centers across North America and Europe, including Montreal, and said the Enovum facility in Montreal would bring Cerebras wafer-scale inference to Canada. Cerebras framed the Montreal deployment as relevant for Canadian enterprises, government, and research institutions requiring high-speed inference. (Cerebras)

The phrase "subject to utility approval" remains the decisive caveat. WhiteFiber's Q1 2026 earnings release indicated the company had completed the purchase of MTL3 in May 2026 after exercising its purchase option, expecting to reduce lease payments by approximately CAD 3.1 million per year, and reported remaining performance obligations of about USD 921 million for colocation services, mainly from NC1. Public call coverage also reported that management said it had applied for more than triple the available power at MTL3, but timing depended on utility review and approval. (PR Newswire) (TradingView)

That is the Montreal arithmetic in a single site: a successful retrofit, an identified AI customer, strong revenue per MW, and expansion value controlled by the utility.

Network evidence: real, enough, not dominant

Data center companies often overstate network capability. The useful test is whether the operator has verifiable network resources, public exchange presence, and facility registrations. Enovum passes that test, but the evidence does not support treating it as Montreal's dominant interconnection platform.

PeeringDB lists AS23116 of Enovum Data Centers Corp., also known as EnovumDC, with the company website, an AS-ENOVUM route set, an open peering policy, contact information, a looking-glass URL, and an operational 100G connection at CANIX Montreal. It also lists interconnection facilities such as Cologix MTL3, Enovum Data Centers MTL02 in Pointe-Claire, and Enovum Datacenters MTL01 in Montreal. (PeeringDB)

The ARIN Whois record shows AS23116 registered to Enovum Data Centers Corp. on July 16, 2025. Hurricane Electric's BGP Toolkit shows AS23116 as Enovum Data Centers Corp., country of origin Canada, with one internet exchange, two originated IPv4 prefixes, zero originated IPv6 prefixes, and nine observed IPv4 peers, including Hurricane Electric, Cogent, NetActuate, EBOX, i3D.net, OVH, and others. (Whois) (BGP Tools)

The MTL01 facility page on PeeringDB places the facility at 3195 suite D, chemin Bedford, Montreal, with Enovum, GloboTech Communications, and another Canadian ASN as networks present. It also marks Enovum's ownership status as tenant. This is consistent with the lease-based economics visible in the financial statements and the GloboTech operational lineage. (PeeringDB)

This evidence demonstrates that Enovum is not just a marketing front. It has an autonomous system, public peering metadata, NOC contacts, an exchange connection, and identifiable facility records. It also demonstrates the limits of the network story. A small set of originated prefixes and observed peers is not the footprint of a carrier hotel. For many Enovum customers, that may be enough. For customers whose architecture depends on dozens of carriers, direct cloud on-ramps, dense content networks, and a deep interconnection marketplace, Cologix has a structurally stronger position.

Montreal latency is useful, not magical

The latency argument for Enovum depends on workload type. Large model training can often tolerate remote sites if power is cheap and network backhaul is adequate. Inference, enterprise SaaS, financial applications, content delivery, gaming infrastructure, VFX workflows, private AI, and hybrid cloud architectures care more about proximity to users, engineers, and networks. Bit Digital made this argument directly when it said that data center assets adjacent to metropolitan areas have value for inference models where latency is a primary concern. (Bit Digital)

Montreal is a plausible enterprise and inference colocation market. It is close to Canadian enterprise buyers, Northeast US traffic, universities, AI talent, and public-sector demand. It offers renewable hydroelectric power and a cold climate. But it is not Northern Virginia. It does not have the same cloud density, customer concentration, or interconnection depth as Ashburn, and a small operator like Enovum cannot manufacture that ecosystem on its own.

The business implication is that Enovum should not be regarded as a universal data center platform with the best location. It is a "good-enough latency plus better power/cooling/customization" platform. For workloads that need a few reliable carriers, a Canadian location, high power density, and operational support, that package can outperform a generic hyperscale region or a distant wholesale campus. For workloads that require maximum cloud on-ramp density or global peering depth, Enovum will often be a secondary choice.

The competitive reference makes this visible. Cologix claims its Montreal platform spans about one million square feet across 12 data centers, including the main Montreal carrier hotel, with access to more than 100 network service providers and direct on-ramp connections to AWS Direct Connect, Google Cloud, IBM Cloud, Microsoft Azure ExpressRoute, and Oracle FastConnect. It also identifies its 1250 René-Lévesque facility as the downtown Montreal carrier hotel. (Cologix)

Enovum's network value is therefore not "better than Cologix." It is "sufficient network access bundled with power, cooling, and customization." This is a smaller but commercially coherent claim.

Who depends on Enovum

Public customer names are limited. Pre-acquisition financials disclose revenue mechanics and receivables concentration but not customer identities. WhiteFiber's annual report says its data center customers include enterprise customers in healthcare, financial, and various technology sectors, as well as cloud GPU customers offering on-demand GPU access for AI, VFX rendering, and scientific computing. It also says MTL1 and MTL3 served 15 customers as of December 31, 2025, and that no single customer represented more than 50% of data center revenue in 2025 or 2024. (Securities and Exchange Commission (SEC))

That customer count matters. MTL1 looks like a multi-customer retail/enterprise/high-density colocation facility. MTL3 looks like an AI deployment with an anchor customer dominated by Cerebras. These are different risk profiles. Multi-customer enterprise colocation reduces single-customer exposure, but requires more sales, support, cross-connect operations, and contract management. Single-anchor AI colocation can produce high revenue density, but creates dependency on the deployment timeline, credit quality, technical requirements, and renewal behavior of one customer.

Cerebras is the visible customer that changes the interpretation. A 5 MW AI inference deployment is large compared to Enovum's original 4 MW MTL1 facility. It signals that Enovum can sell beyond traditional racks and cages into customer-specific AI infrastructure. But it also concentrates the growth story. If Cerebras renews, expands, and attracts adjacent demand, MTL3 becomes a proof. If Cerebras changes architecture, shifts capacity, negotiates lower pricing, or does not expand, MTL3 becomes a single-customer retrofit with limited public evidence of replacement demand.

WhiteFiber's broader business shows concentration risk more starkly. The company disclosed that its initial cloud customer represented 70.7% of revenue in 2025 and 96.6% in 2024, while data center revenue was less concentrated. The lesson is not that Enovum's data center segment has the same customer concentration as the cloud GPU segment. The lesson is that AI infrastructure companies can appear diversified by technology while still relying heavily on a few large counterparties. (Securities and Exchange Commission (SEC))

The public support and informal chatter trail is thin. Enovum's support portal has an outage category and tag that show no public outage articles, and searches yield many more company announcements, facility listings, and transaction coverage than customer complaint threads. That silence is weak evidence. Enterprise colocation customers often do not complain publicly, and problems may be resolved through service credits under private contracts. Nevertheless, it is relevant: there is no visible pattern of public complaints under the Enovum name that would materially change the commercial interpretation. (support.enovumdc.com) (support.enovumdc.com)

Competitors can erode different parts of the bundle

Enovum does not face a single type of competitor. It faces several, each eroding a different part of the bundle.

Cologix erodes the interconnection argument. Its Montreal presence has scale, carrier density, cloud on-ramps, and a long operating history as a network-neutral provider. Customers who value interconnection above customization will naturally compare Enovum to Cologix and ask why they should not deploy into a deeper carrier ecosystem. (Cologix)

Vantage erodes the scale argument. Its Montreal III campus is described as an 8-acre site with 30 MW of critical IT load, powered by nearly 100% renewable electricity from Hydro-Québec, and part of a combined Montreal presence of 89 MW. It also includes a dedicated 45 MW substation with a direct private connection to Hydro-Québec. This is a different scale class from MTL1. (Vantage Data Centers)

eStruxture erodes the Canadian-provider argument. It describes itself as headquartered in Montreal, serving nearly 1,000 customers across Canada, and operating the largest Canadian data center platform with AI-ready, carrier-neutral data centers in key Canadian markets. Enovum cannot rely on "Canadian" or "Montreal" as unique differentiators against that platform. (eStruxture)

QScale erodes the HPC renewable-campus argument. Its Q01 materials describe 142 MW of electrical capacity, 100% renewable energy, ultra-high-density workloads, and heat-reuse positioning. That product is not identical to Enovum's metro colocation model, but it competes for AI and HPC buyers who care more about large blocks of renewable power than about urban proximity. (QScale)

The public cloud erodes the entire colocation decision. AWS, Azure, Google Cloud, and specialized cloud GPU providers can turn infrastructure into a service. A customer that can rent compute, storage, and managed services without owning hardware may never need Enovum. Enovum's defense is the set of workloads where cloud abstraction is too expensive, too inflexible, too slow, too opaque, too jurisdictionally complicated, or unable to meet hardware density requirements.

The failure path is easy to describe. If the big carrier hotels keep winning the network-sensitive workloads, hyperscale campuses win the big standardized megawatt blocks, eStruxture wins Canadian enterprise buying, QScale wins the renewable mega-loads, and cloud platforms absorb flexible compute, Enovum is left with a narrower market: high-density customers who need Montreal, need customization, trust the operator, and accept its network depth. That is still a real market, but not a monopoly.

The North Carolina revelation

WhiteFiber's move into North Carolina is not a distraction from the Enovum story. It clarifies it. If the company were primarily a Montreal colocation operator, North Carolina would look like diversification. If it is an AI infrastructure platform hunting electricity, North Carolina is the natural next step.

WhiteFiber's annual report indicates that Enovum signed a Capacity Agreement with Duke Energy for NC1 to receive 24 MW by September 1, 2025, an additional 40 MW by April 1, 2026, and a final permanent service of 99 MW within four years from May 16, 2025. It also says actual rates will be determined when facilities go into service. (Securities and Exchange Commission (SEC))

WhiteFiber later announced a 10-year colocation agreement with Nscale for the first 40 MW of IT load at NC1 in Madison, North Carolina, with an initial contract value of roughly USD 865 million, excluding electricity and certain pass-through expenses. This is a different order of magnitude from MTL1 and larger than the 5 MW Cerebras deployment at MTL3. (fierce-network.com)

Market commentary has interpreted the move the same way. The publicly visible preview of CoStar's article said that Enovum was expanding to North Carolina because the huge amount of electricity needed for AI was becoming harder to get in Quebec. The article itself is subscriber-only, so the preview should be treated as market commentary rather than primary evidence. But it aligns with the hard evidence: Hydro-Québec is proposing a higher rate and a selection process for large data centers, while WhiteFiber is contracting much larger power blocks in North Carolina. (CoStar)

The commercial conclusion is not that Quebec is unattractive. It is that Quebec is no longer obviously the easiest place to develop large AI loads. Montreal remains valuable for Canadian enterprises, sovereign or location-sensitive workloads. It can be excellent for inference when latency and Canadian presence matter. But if the largest growth prize is quick access to tens or hundreds of megawatts, WhiteFiber will follow power availability rather than provincial loyalty.

Ownership and the sovereignty narrative

Enovum's original story had a local-sovereignty flavor. The 2021 press coverage emphasized Quebec ownership, the absence of public subsidies, and a founder who criticized larger, foreign-capital-backed operators. That positioning mattered because data center buyers often care about jurisdiction, procurement optics, local accountability, and trust. (TVA Nouvelles)

After Bit Digital acquired Enovum, that story changed. Bit Digital is a Nasdaq-listed company headquartered in New York, and WhiteFiber is a public AI infrastructure company controlled by Bit Digital following its IPO structure. WhiteFiber's annual report indicates that Bit Digital owned approximately 71.5% of WhiteFiber after the offering and the exercise of the over-allotment option. (Securities and Exchange Commission (SEC))

This does not mean Enovum is no longer Canadian in operational terms. Its facilities, electricity contracts, employees, and customers may still be in Canada. But ownership control matters to some buyers and regulators. Public-sector customers, sovereignty-sensitive enterprises, and policymakers may distinguish between Canadian location and Canadian control. Private AI customers may not care if the physical jurisdiction, security controls, power density, and price are right.

The most precise commercial position is therefore mixed. Enovum can credibly sell Canadian location, Montreal latency, Quebec hydro power, and local operational competence. It can no longer sell the same purely-local-rival story that appeared in 2021.

The dependencies underlying the margin

A data center operator's margin is exposed to a set of dependencies. Enovum's stack is unusually visible because WhiteFiber's filings detail the risks.

Electricity is first. In Quebec, WhiteFiber states that all hydroelectric power for MTL1, MTL2, and MTL3 is supplied by Hydro-Québec, a government-owned utility with predetermined rates based on sector and demand. This means a fundamental cost input and the gatekeeper of expansion are controlled by a single provincial utility. (Securities and Exchange Commission (SEC))

Network is second. WhiteFiber states that its data centers may require building and operating a sophisticated redundant fiber network, that connecting multiple carrier facilities is complex and subject to factors outside the company's control, and that failure or delay in diverse connectivity could materially impact operating results and cash flow. (Securities and Exchange Commission (SEC))

Suppliers are third. WhiteFiber states that it depends on a limited number of suppliers for certain products and services, such as electricity, electrical equipment, GPU servers, construction materials, and construction services. It specifically notes long-lead-time items such as generators, and the risk of supply chain constraints, budget overruns, labor issues, financing availability, access to electricity, and permitting or utility delays. (Securities and Exchange Commission (SEC)) (Securities and Exchange Commission (SEC))

Insurance is fourth. WhiteFiber discloses that it does not carry business interruption or service interruption insurance, apart from directors' and officers' liability insurance, because it determined such insurance was not practical on commercially reasonable terms. This is a significant risk in a business where downtime, equipment damage, or prolonged service interruption can trigger SLA credits, customer loss, and reputational damage. (Securities and Exchange Commission (SEC))

Customer commitments are fifth. Service level agreements can turn an operational failure into credits, damages, or termination rights. The filings state that virtually all customer contracts include SLA commitments, and that failures can lead to rental credits, legal liability, monetary damages, regulatory penalties, or termination rights for repeated failures. (Securities and Exchange Commission (SEC))

The margin is therefore not simply "sell electricity at a markup." It is the spread after reserving power, building redundancy, funding capex, staffing operations, managing cooling, buying equipment, securing carriers, passing audits, meeting SLAs, and bearing expansion risk.

The business model in a single equation

The simplest way to model Enovum is:

Contracted customer load × price per kW or MW
minus electricity and demand charges
minus lease or ownership cost
minus cooling, labor, maintenance, and connectivity cost
minus depreciation, financing, and overhead
minus SLAs, customer concentration, and expansion risk.

MTL1 shows the traditional version of that equation. A small, full site produces recurring monthly colocation revenue, positive gross contribution, and visible contracted backlog, but overhead, depreciation, and financing reduce profitability. MTL3 shows the AI version. A single 5 MW IT load customer can produce much larger absolute revenue, but the operator must build or retrofit for high density, manage cooling complexity, and accept anchor-customer exposure. NC1 shows the platform version. A 40 MW, 10-year contract can generate a huge backlog, but moves the company into larger financial, construction, utility, and execution risk.

Enovum's economic value therefore does not reside in a single facility. It resides in whether the team can repeat this equation with attractive margins. That is exactly what Bit Digital bought. The acquisition price cannot be justified by MTL1's historical revenue alone. It only makes sense if Enovum is an operating core for higher-density, customer-backed, AI-oriented data center growth.

WhiteFiber's 2025 segment data show the initial outcome. Colocation services generated USD 8.91 million in revenue in 2025, with USD 1.44 million in electricity cost, USD 1.03 million in data center lease expense, USD 407,000 in salary expense, USD 580,000 in other segment items, and USD 5.46 million in segment gross profit. Cloud services were much larger, with USD 68.75 million in revenue. Canada represented USD 8.91 million in service-location revenue and USD 122.78 million in long-lived assets at year-end 2025. (Securities and Exchange Commission (SEC)) (Securities and Exchange Commission (SEC))

This tells the story clearly. Data center revenue was still smaller than cloud, but Canadian long-lived assets had become substantial. The company was building an asset base before the full revenue contribution arrived. That can create operating leverage if facilities fill at good rates. It can destroy capital if power, customers, or build timelines disappoint.

What the public record still cannot answer

The public record is enough to identify the business model. It is not enough to fully underwrite the company.

It does not reveal Enovum's exact effective electricity price per site, including demand charges, transitional treatment, pass-through mechanisms, or contractual rate protections. It does not reveal customer-by-customer pricing, gross margin, or renewal terms. It does not show a detailed history of uptime, SLA credits, incident reports, audit reports, or customer satisfaction. It does not show whether the claimed Tier-3 standard is independently certified at each site or self-assessed against Tier-3 requirements. It does not disclose the full list of carriers physically present in Enovum's meet-me rooms versus those reachable via dark fiber to Cologix or other carrier hotels.

It also does not answer whether MTL3 can materially expand. The public filings say future expansion is subject to utility approval; that is precisely the unknown that determines the upside potential. A site with 5 MW of sold IT load is valuable. A site that can more than triple available power under acceptable rate conditions is far more valuable. A site blocked on utility approval is a constrained asset.

The informal public record is also limited. There is no dense trail in customer forums, no obvious public outage narrative, and no large body of carrier gossip under the Enovum name that materially changes the evidence. The only market commentary that shifts the interpretation is the industry framing around North Carolina: power availability appears to be pushing expansion out of Quebec. That commentary is consistent with the utility and filing evidence, but should not be taken as proof of any specific Hydro-Québec decision about Enovum.

Evidence book

The facts that would change the Montreal arithmetic

The commercial view would change if five facts became public. First, the effective electricity rate per site after demand charges, pass-through clauses, and Hydro-Québec's new large data center process. Second, the approved power expansion path at MTL3 and any recovered path at MTL2. Third, actual churn, renewal pricing, and SLA credit history at MTL1. Fourth, carrier usage and cloud connectivity data, rather than directory-level availability. Fifth, actual capex per megawatt and gross margin at MTL3 after the Cerebras deployment stabilizes.

If those facts show protected low-cost power, low churn, strong SLA performance, real expansion approvals, and durable AI contracts, Enovum is a scarce AI colocation platform in Montreal. If they show repriced power, delayed approvals, customer concentration, weak interconnection pull, and capex inflation, the business is less dramatic: a competent local colocation operator bought at a platform multiple because AI infrastructure capital briefly valued every credible megawatt as strategic.