Hut 8 Corp.: From Bitcoin Pioneer to AI Power Broker

I. Introduction & Episode Roadmap (0:00 - 10:00)

On the morning of July 20, 2026, Hut 8 Corp. put out a press release that would have been unintelligible to anyone who had followed the company at its founding. It announced that a single anonymous tenant — described only as a "high investment-grade" counterparty — had signed a second fifteen-year lease for 352 megawatts of IT capacity at a place called Beacon Point in Nueces County, Texas. That tenant now held 704 megawatts at the campus, and the campus itself was fully contracted at a base-term value of $19.6 billion, rising to $50.2 billion if every renewal option were exercised.²

Nowhere in that release did the word "bitcoin" appear.

Rewind two and a half years and Hut 8 was something else entirely: a Canadian bitcoin miner with a large hoard of coins, a mediocre machine fleet, a bruising merger behind it, a short-seller report circulating, and a stock trading at $6.77.⁴ The distance between those two documents is the story of this episode.

The post-halving identity crisis

The proximate cause of the transformation was arithmetic. In April 2024 the Bitcoin network executed its fourth "halving," cutting the block subsidy paid to miners from 6.25 BTC to 3.125 BTC. Overnight, the revenue side of every miner's income statement was cut roughly in half while the cost side — power, machines, people — stayed exactly where it was. What survived was not the companies with the most hash rate. It was the companies with the cheapest, most flexible, most abundant electricity.

That timing collided with a second event: the AI training boom turned grid-connected megawatts into the scarcest input in American technology. Hyperscalers discovered that the binding constraint on building a frontier model was not GPUs or capital, but an interconnection queue that could run five to seven years. Bitcoin miners — who had spent a decade quietly acquiring exactly those interconnects in exactly those overlooked places — suddenly found that the least glamorous asset on their balance sheets was the one everybody wanted.

Every miner noticed. Only some managed to convert. Hut 8 is the most aggressive conversion story in the group, and also one of the most contested.

The subject of today's story

Hut 8 Corp. (HUT, listed on NASDAQ) is today a Delaware-incorporated energy and digital infrastructure developer. As of its first-quarter 2026 report, it managed roughly 710 MW of operating capacity, carried a development pipeline of 8,375 MW across various stages of diligence, exclusivity, development and construction, and held about $1.3 billion of combined cash and bitcoin on a consolidated basis.¹ Its self-mining business no longer sits inside the company in the way it once did: the ASIC fleet was carved out into a separately listed subsidiary, American Bitcoin Corp. (ABTC), in which Hut 8 held roughly 55% of the economics as of March 31, 2026.¹

Its financial statements, meanwhile, look deeply strange. Revenue in the first quarter of 2026 grew 226% year over year to $71.0 million and gross margin expanded from 14% to 64% — and the company still reported a net loss of $253.1 million, because bitcoin's price fell and mark-to-market accounting ran the damage straight through the income statement.¹ Understanding why both of those things are true at once is most of the analytical work in this story.

The narrative arc

The Canadian roots. A fast-scale build in Alberta using proprietary European chips, and one of the first liquid public proxies for bitcoin anywhere.
The supplier trap. How an exclusive hardware relationship — the thing that made the company possible — nearly killed it.
The great restructuring. The 2023 merger with US Bitcoin Corp, the swift removal of legacy management, and the arrival of a chief executive in his late twenties.
The AI frontier. Coatue's $150 million, a thousand GPUs in Chicago, and a managed-services contract that vanished in a fortnight.
The power-first rebuild. Carving out the miners, selling the gas plants, and converting raw interconnects into investment-grade contracted cash flow.

Episode themes

Three ideas run through everything that follows. First, power as a cornered resource — and the question of whether an interconnect is a durable moat or merely an option with an expiry date. Second, the discipline of grid balancing: how a business learns to make money by not consuming electricity. Third, the migration from commodity economics, where you are a price-taker against a global network, to contracted infrastructure economics, where a triple-net lease with an AA– counterparty pays you whether or not the tenant's business works.

To judge whether Hut 8 has really made that migration, we first need to understand the machine it is trying to escape.

↑ Back to Top

II. The Physics of Bitcoin Mining: Energy, ASICs, and Grid Economics (10:00 - 25:00)

Picture a global lottery that runs every ten minutes, forever, with the prize fixed by software rather than demand. Anyone may buy tickets. Tickets are bought with electricity. And here is the cruel part: if you buy more tickets, the lottery automatically makes tickets harder to win, so that the prize still goes out every ten minutes on average. That is bitcoin mining, and it explains almost everything about how these companies behave.

The arms race nobody can win

The technical name for the ticket-difficulty adjustment is the difficulty retarget, recalculated roughly every two weeks. The consequence is that mining is a zero-sum share game. If Hut 8 doubles its computing power while the rest of the world triples theirs, Hut 8 earns less bitcoin than before despite spending more. There is no product differentiation, no brand, no customer relationship. The only variables that matter are how efficiently you convert electricity into computation and how little you pay for the electricity.

The three pillars of mining profitability

Fleet efficiency. Mining chips are rated in joules per terahash (J/TH) — energy consumed per unit of computing work. Lower is better. The workhorse machines of the 2021–2023 cycle, Bitmain's Antminer S19 family, ran near 30 J/TH. The generation that followed, the S21/T21 class, pushed below 18 J/TH. When the block subsidy halved in April 2024, that gap became existential: a 30 J/TH machine burns almost twice the power per unit of output as an 18 J/TH machine, and after a halving there is only half as much revenue to cover it. Miners either found capital to re-equip or watched their fleets go cash-flow negative.

Energy sourcing. Power cost, measured in dollars per megawatt-hour, is the single largest determinant of survival. The industry's cost curve is brutally steep: a miner at $30/MWh and a miner at $60/MWh are not competitors in any meaningful sense; the second one simply dies first in every downturn.

Capital allocation. Because miners produce a liquid, volatile asset, every management team faces a recurring decision: sell the bitcoin to fund operations, or hold it and raise equity instead. Holding is a leveraged bet on price that transfers risk onto shareholders through dilution. Selling funds the business but forfeits the upside that many shareholders bought the stock for in the first place. There is no neutral choice here — only a disclosed one.

The grids

Geography is destiny. Canada's provincial grids are regulated, cool and relatively stable — good for uptime, less good for opportunism, and increasingly expensive as industrial demand rose. Texas is the opposite. The Electric Reliability Council of Texas (ERCOT) is a merchant grid with enormous wind and solar penetration, chronic transmission congestion, and price behaviour that ranges from negative — generators paying to offload power — to several thousand dollars per megawatt-hour during a summer scarcity event.

Here is the intuition that makes the rest of this story work. A grid operator's job is to match supply and demand every second of every day. Historically that meant paying generators to turn on. But paying a large customer to turn off accomplishes the same thing and is usually cheaper. The problem is that almost no industrial customer can do it: an aluminium smelter that stops mid-pour destroys its own equipment, a hospital cannot, a semiconductor fab certainly cannot. Interruptible load at hundreds of megawatts is genuinely rare, and grids will pay for it.

For a normal industrial user, that volatility is a nightmare. For a miner, it is the business model. A mining load is uniquely interruptible: nobody is harmed if hashing stops for four hours. So a miner can contract for power at a fixed price, then sell that power back into the grid when spot prices spike, pocketing the spread and shutting the machines off. It can also enroll in demand-response programs that pay simply for the promise to curtail. Done well, this turns a power bill into a trading book — and, crucially, it teaches an organisation how to negotiate with utilities and grid operators, a skill that turns out to be far more valuable than mining itself.

The competitive landscape

By 2024 the sector had split into recognisable camps. Marathon Digital (MARA) and Riot Platforms (RIOT) pursued scale, hoarding both hash rate and coins. Core Scientific (CORZ) — which passed through Chapter 11 during the 2022 crypto winter, one of several peers that restructured while Hut 8's future leadership sat on creditors' committees⁵ — re-emerged as an early AI-hosting convert. IREN (IREN) built owned, purpose-built sites and moved into GPU cloud. TeraWulf (WULF) and Cipher Mining (CIFR) leaned on low-cost, contracted power and later signed their own hyperscaler leases.

The strategic insight, arrived at independently by most of these firms between 2023 and 2025, was the same: the enduring asset was never the machines. It was the substation, the interconnection agreement, the land, and the relationship with the utility. Bitcoin mining had accidentally financed a decade of speculative grid development that no data-center developer would have underwritten.

Hut 8's version of that insight came later than some and more violently than most — because for its first three years, the company did not control its own hardware at all.

↑ Back to Top

III. Act I: Canadian Origins & The Bitfury Golden Cage (2017–2020) (25:00 - 40:00)

In the winter of 2017–18, the town of Medicine Hat, Alberta — a place that markets itself as "The Gas City" because of the shallow natural gas fields beneath it — agreed to lease land and supply 42 megawatts of electrical power to a company almost nobody had heard of. The deal came with a promised $100 million of local construction spending, roughly 100 construction jobs and 42 permanent ones. Announced in March 2018, it was, for a prairie city of 60,000, a genuinely large industrial win.²⁰

The company was Hut 8 Mining Corp., and it was less a startup than a distribution vehicle.

Incubated, not founded

Hut 8 was created in 2017 in partnership with the Bitfury Group, the Amsterdam- and Georgia-rooted mining conglomerate founded by Valery Vavilov, which designed its own bitcoin ASICs and packaged them into shipping-container-sized units called BlockBoxes. Bitfury needed a way to monetise its chips at scale without carrying the energy and capital burden itself. Hut 8 was that way: it raised public money, bought BlockBoxes, plugged them into cheap Alberta power, and mined.

The physical build was extraordinarily fast by industrial standards. Mining began in December 2017. By late 2018 the company operated 73 BlockBox units — 56 at Medicine Hat, representing about 67 MW of operating power, and 17 at a smaller site in Drumheller drawing 18.7 MW — and had already mined more than 5,100 bitcoin.²⁰ Speed was the whole proposition: a containerised unit could be trucked in and energised in weeks, where a conventional data center took years.

The capital advantage

Hut 8 listed on the TSX Venture Exchange in early 2018 through a reverse takeover, backed by Bitfury and funded by well over C$100 million of private placements.²⁰ The timing was exquisite and terrible: it arrived just as bitcoin was rolling over from its December 2017 peak, and the shares fell on debut. But the strategic point held. For Canadian institutions and retail investors who wanted bitcoin exposure in a brokerage account years before spot ETFs existed, Hut 8 was one of the very few liquid proxies available. That scarcity value — being a way to own the trade — became a defining feature of the equity, and a source of investor expectations that would later constrain management badly.

It is worth pausing on how unusual that structure was. Bitfury was one of the few vertically integrated players in the industry — it designed silicon, built immersion and air-cooled enclosures, ran its own mining, and sold blockchain analytics software. Vertical integration gave it a cost advantage over anyone buying merchant chips, and in 2017 that advantage looked permanent. Hut 8's investors were, in effect, buying a share of Bitfury's manufacturing edge without the risk of financing a fab. When the edge held, the arrangement was brilliant. When it stopped holding, there was no exit.

The golden cage

Hamilton Helmer's 7 Powers treats switching costs as an advantage accruing to the seller. Hut 8's early history is the same mechanic running in reverse, with Hut 8 as the captive.

The exclusive relationship that let Hut 8 scale in months also forbade it from buying anyone else's machines. And by 2019, Bitfury's silicon had fallen behind. Bitmain's Antminers and MicroBT's Whatsminers were pulling ahead on both raw hash rate and, more damningly, joules per terahash. Global hash rate was compounding relentlessly, meaning Hut 8's share of each block shrank every month unless it added machines — but the only machines it was permitted to add were the uncompetitive ones.

This is the purest illustration of supplier power in the sector's history. Hut 8 had no product decision to make. Its cost structure was set in a boardroom it did not sit in. The company was, in the most literal sense, someone else's channel.

The birth of the HODL doctrine

Facing a fleet it could not upgrade, management leaned into the one thing its shareholder base did prize: coins. Hut 8 became the sector's most conspicuous practitioner of retaining 100% of self-mined bitcoin on the balance sheet, funding cash operating losses through equity issuance rather than by selling production.

It is worth being precise about what this policy is. It is not a "treasury strategy" in the sense a corporate treasurer would recognise. It is a decision to fund a loss-making industrial operation by selling equity into the market at whatever price prevails, in order to avoid selling an asset whose price the seller does not control. It works spectacularly when bitcoin rises faster than the share count. It compounds destruction when it does not. Hut 8 built a large hoard and a large share count simultaneously, and both facts would matter later — the hoard as the collateral that eventually financed an AI pivot, the share count as the reason management would become almost religious about non-dilutive capital.

By 2020, with an obsolete fleet, a restrictive supplier contract, and a stock that lived and died on the bitcoin price, Hut 8 needed someone who had run a real data center business. It hired one.

↑ Back to Top

IV. Act II: Breaking the Monopoly & The First HPC Pivot (2020–2023) (40:00 - 55:00)

Jaime Leverton arrived in December 2020 from an entirely different world. Her background was enterprise infrastructure and telecom — the business of selling racks, connectivity and managed services to corporate IT departments, where contracts run for years and customers care about uptime service levels, not block rewards. She was, by design, not a crypto native. The board wanted someone who could renegotiate a bad contract and build a business that did not depend solely on a price chart.

Breaking the shackle

The first task was the Bitfury agreement. Through renegotiation, Hut 8 won the right to buy state-of-the-art ASICs from Bitmain and MicroBT. The effect on the fleet's efficiency was immediate and mechanical — a company that had been forced to run yesterday's silicon could now buy today's.

It was also, strategically, a demotion disguised as a liberation. Once Hut 8 could buy the same machines as everyone else, it lost its only differentiator and became a pure price-taker in a commodity industry. Freedom from the supplier meant exposure to the market. From that moment, the only sustainable question was: where does your power come from, and what does it cost?

The TeraGo acquisition: a hedge that was not a hedge

In January 2022, Hut 8 announced it would acquire the data-center business of TeraGo Inc. for C$30 million in cash, adding five conventional colocation facilities across Canada in Toronto, Mississauga, Vancouver and Kelowna.¹⁹

The logic was sound on paper. Bitcoin revenue is violent and uncorrelated with anything a CFO can forecast. Colocation and cloud revenue is contracted, recurring, and boring in the best sense. Bolting one onto the other should smooth the ride and give the equity a valuation anchor that did not move with the coin. At roughly ten times EBITDA, the price was defensible.

What the deal did not do — and this is the part that matters for everything that followed — was buy anything useful for the coming AI build-out. These were traditional enterprise data centers built for racks drawing a handful of kilowatts each: raised floors, air cooling, modest power density, tens of megawatts at most across the entire portfolio. An AI training hall in 2025 would demand liquid cooling and rack densities more than an order of magnitude higher, delivered in hundreds of megawatts at a single site. Hut 8 bought a diversification hedge, not a call option on the future. When the AI wave arrived, this portfolio contributed brand and a Canadian cloud business — real, but small — rather than a platform.

The honest read is that management correctly identified the disease (revenue volatility) and prescribed a treatment that addressed the symptom rather than the cause. The cause was that Hut 8 did not control enough cheap, scalable power.

The bottleneck

Reality closed in from the energy side. Canadian electricity costs climbed, eroding the original locational advantage. And at North Bay, Ontario, a protracted and bitter dispute with the site's energy supplier forced Hut 8 to idle thousands of machines — an especially painful outcome because idle ASICs still depreciate, still carry financing costs, and still count against the hash rate investors were tracking.

The episode was a preview of the central lesson of the next three years: in this business, a contract with a power provider is not an administrative detail. It is the asset. A miner that cannot dispatch its own load is not an infrastructure company; it is a tenant with expensive furniture.

It is worth being fair to the executive who ran that period. Leverton inherited a company with a broken supplier contract, an obsolete fleet and a shareholder base that wanted bitcoin exposure rather than a data-center business, and she fixed the first of those problems — the one that was structurally fatal. What she could not do, inside a Canadian cost base and a collapsing coin price, was manufacture the thing the company actually lacked: hundreds of megawatts of cheap, dispatchable American power. The strategy she executed was the right strategy for a 2019 industry. The industry stopped being that in about eighteen months.

By the second half of 2022, with bitcoin collapsing, energy prices spiking after Russia's invasion of Ukraine, and fixed debt service falling due across the sector, being a mid-sized Canadian miner with an average cost position was not a viable place to stand. Hut 8 needed either a buyer or a partner. It found something in between.

↑ Back to Top

V. Act III: The USBTC Merger & The Boardroom Coup (55:00 - 1:15:00)

The 2022 crypto winter was not a drawdown; it was a culling. Bitcoin fell roughly 75% from its peak, power prices spiked, and the sector's balance sheets — built for a world of perpetual appreciation — buckled. Core Scientific, then the largest miner in North America, went into Chapter 11. Celsius Network, the crypto lender, collapsed into one of the most complex bankruptcies of the cycle.

Asher Genoot spent that winter on the other side of the table. As he later described it, he sat on unsecured creditors' committees of companies going through restructuring, and his firm served as plan sponsor for carving Celsius's mining assets out of bankruptcy into a new entity.⁵ "I learned firsthand meeting with creditors on what could go wrong when you're building in a hypergrowth environment and the music stops," he said years later, on a call where he returned to that phrase — what if the music stops — repeatedly.⁵

Enter US Bitcoin Corp

USBTC, co-founded by Genoot and Michael Ho, was the philosophical opposite of Hut 8. Where Hut 8 owned its sites and hoarded its coins, USBTC ran an asset-light managed-services model: it operated mining infrastructure for third parties for fees, built energy-arbitrage software, and treated power markets as the core competency. It had US sites — critically, in ERCOT — and a reputation for operating other people's assets cheaply.

The two companies fit together like a lock and a key, if you squinted. Hut 8 brought a large bitcoin treasury, public-market history and a listed vehicle. USBTC brought American energy assets, operating software and a services business that generated fees without capital.

There is a temperamental difference buried in that description that turns out to matter more than the balance sheets. Hut 8's model was ownership: buy the site, buy the machines, hold the coins, and let the asset base appreciate. USBTC's model was operating leverage on other people's assets: get paid for competence. The first model is a bet on price. The second is a bet on execution — and it produces a management team that thinks in contracts, counterparties and downside scenarios rather than in price targets. Almost everything Hut 8 did after 2024 is legible as the second culture displacing the first.

The combination

The all-stock "merger of equals" closed on November 30, 2023, creating Hut 8 Corp. — newly incorporated in Delaware, headquartered in Miami, and listed on Nasdaq.¹⁸ On paper it was a balance-sheet-meets-operating-capability trade.

In practice it was a governance event waiting to happen. Merging two management teams with opposite philosophies rarely produces a synthesis; it produces a winner. And the market did not wait long to test the combination. On January 18, 2024, the short-seller J Capital Research published a report attacking the deal; the stock fell 23% that day.²² Investors later alleged in a securities class action that Hut 8 had failed to adequately disclose energy and internet-related problems at King Mountain — a Texas joint venture in which USBTC held a 50% stake before the merger — and had misrepresented USBTC's financial condition in the merger materials.²²

The sudden coup

On February 7, 2024, roughly ten weeks after closing, the board replaced Jaime Leverton with Asher Genoot, then 28.¹⁷ There was no transition period and no elaborate explanation. The company that had hired a traditional data-center executive to civilise a crypto miner had reversed itself and handed the business to the mining operator.

Analysts pushed back on the abruptness — a CEO change that fast after a "merger of equals" invites the reading that it was never a merger of equals at all, and the disclosure record does not fully dispel it. Genoot's own later account is candid about how weak the starting position was: institutional ownership below 10%, a share price of $6.77, and, by his description, a company that had to be rebuilt "from a first principles approach."⁴

For investors, the episode cuts both ways. It was decisive capital-allocation governance — a board acting on a leadership mismatch rather than waiting two years for it to show up in results. It was also an object lesson in what a nominal merger of equals is worth when one side controls the operating capability and the narrative. The relevant question was never whether the coup was polite. It was whether the person who won had a plan.

He did, and its first moves were unsentimental.

↑ Back to Top

VI. Act IV: Asher Genoot's Restructuring Playbook (1:15:00 - 1:30:00)

The first thing a new chief executive reveals is what he is willing to kill.

Rationalising the asset base

Genoot's early decisions were subtractive. The 42 MW Drumheller facility in Alberta — one of the original 2018 sites, and by then a high-cost operation — was shut down, its functioning machines redeployed to cheaper US locations. The symbolism was deliberate: the founding infrastructure of the company was not sacred if its power price no longer worked. (In a small coda that says something about how the business changed, Drumheller was re-energised in 2026 as ASIC colocation capacity — not for Hut 8's own account, but leased to its now-separate mining affiliate.¹)

The ERCOT build and the art of not mining

Alongside the closures came a Texas build-out: Salt Creek at 63 MW, and the far larger Vega site at 205 MW. But the more interesting development was software, not steel.

Rather than mine continuously, Hut 8 ran automated systems monitoring real-time ERCOT prices, curtailing machines within seconds when power became more valuable sold than consumed. In effect the company operated a virtual power plant whose "generation" was the electricity it chose not to use.

The operating results in the third quarter of 2024 showed what that discipline was worth: an average energy cost of $28.83/MWh and a cost to mine one bitcoin of $31,482, against a bitcoin price well above $60,000 at the time.[^19] Read carefully, those two numbers say something specific. They say the company had achieved a genuine cost position — not a marketing claim — in the one variable that determines survival in a commodity business. They do not say the business was good. A miner earning a spread on a commodity it cannot price is still a price-taker; a 50% cash margin at $60,000 bitcoin becomes a negative margin at $25,000 bitcoin, with no management action available in between.

Genoot appears to have understood this better than his own investor base did. The curtailment capability's real value was not the mining margin. It was the proof to utilities and grid operators that this team could manage large interruptible loads — the exact credential required to win interconnection agreements for something else entirely.

The operator's temperament

Genoot is, by the evidence of his own public appearances, an unusual chief executive for a company of this size — young, technically fluent, and given to describing his own paranoia as a design principle. His CFO's characterisation, relayed on an earnings call, is that "for your age, you're really more of a credit guy than anything else."⁵ That is not the vocabulary of a bitcoin executive; it is the vocabulary of someone who watched a leveraged industry break in 2022 and concluded that the job is to survive the downside.

The management style that follows is hands-on to a degree that can be read as an asset or a risk. In the third quarter of 2025, after two senior operators were poached by xAI, Genoot took direct control of procurement, rejected the practice of pricing equipment from vendor quotes, and rebuilt the function around estimating manufacturing cost from raw materials and labour up — putting a former consultant with a chemical-engineering background, and no procurement experience, in charge of some of the largest supply agreements in the sector.⁴ He describes the result as better pricing and lead times. An investor should note both readings: the first-principles instinct that produced Vega's cost per megawatt and the Beacon Point redesign is the same instinct that concentrates critical decisions in one person's judgment, at a company now committing billions of dollars of construction capital.

Incentive alignment, and its limits

Genoot's pay structure was built to make the equity the point: a base salary of $550,000 with a target annual bonus of 100% of salary, and 100% of his 2025 annual long-term equity granted as performance stock units.²¹ Hut 8's own compensation philosophy language is explicit that base salary is "deliberately moderated" in favour of at-risk pay, and the company maintains stock ownership guidelines, a clawback policy tied to accounting restatements, and a board on which six of eight directors are independent.²¹

That is a defensible design — and it produced a number that deserves scrutiny rather than applause. For 2025, Genoot's annual bonus paid out at 200% of target on adjusted EBITDA of $80 million against a $46 million target, and his total reported compensation for the year was approximately $239.9 million, inflated by one-time "Transformation Awards" granted in November 2025 with four-year performance periods tied to market-capitalisation growth and to the value of Hut 8's stake in American Bitcoin.²¹

Two honest observations follow. First, the awards are genuinely performance-conditioned, with two-year post-vesting holding requirements — this is not a repricing or a retention grant dressed up. Second, a quarter-billion-dollar reported figure for a company whose GAAP results were a net loss is exactly the kind of item an activist investor circles in red ink, and shareholders were asked to bless it in an advisory say-on-pay vote at the June 11, 2026 annual meeting.²¹ The defensible view and the uncomfortable view are both available on the same page of the proxy.

Cost discipline and incentive design, though, only produce a better version of a commodity business. The leap required someone to pay Hut 8 for something other than bitcoin.

↑ Back to Top

VII. Act V: The AI Frontier & The Coatue Endorsement (1:30:00 - 1:45:00)

In the first half of 2024, an odd inversion took hold in American infrastructure. The scarcest input in artificial intelligence stopped being GPUs and became the ability to plug them in. Utilities were quoting five- to seven-year waits for large interconnections. Transformers and high-voltage breakers had multi-year lead times. And a group of unfashionable companies discovered that the interconnection agreements they had secured to mine an internet currency were now among the most valuable permits in the country.

Why the wait? Because connecting a large load to a high-voltage grid is not a matter of running a cable. The utility must study whether the addition destabilises the network, potentially upgrade transmission lines and substations far from the site, and procure equipment — large power transformers, high-voltage breakers, switchgear — that global manufacturers were quoting in years rather than months. Then a regulator must approve who pays for it. Each of those steps is sequential, and each is administered by an institution with no commercial incentive to hurry.

The catch — and it is the whole game — is that an interconnect is not a data center. Between "we have 300 megawatts approved at this substation" and "an AI lab will sign a fifteen-year lease here" sits engineering, construction, cooling design, credit, and a counterparty willing to bet its training schedule on your delivery date.

The gap between the two buildings is larger than most investors appreciate. A mining shed is a metal barn full of fans: if a row of machines goes down for six hours, the operator loses six hours of bitcoin and nothing else. An AI training hall is closer to a hospital. It needs redundant power paths so a single utility fault does not halt a training run costing millions of dollars a day; liquid cooling plumbed to the chip because modern accelerators reject more heat than air can carry away; structural floors rated for racks that weigh as much as small cars; and network fabric that keeps thousands of GPUs in lockstep. The cost per megawatt runs an order of magnitude above a mining site — Hut 8 guides to $9–11 million per MW on its AI campuses⁴ against the roughly $455,000 per MW it claims for Vega.⁵ Owning an interconnect gets a developer to the starting line. It does not qualify it to run the race.

Coatue writes the first cheque

On June 21, 2024, Hut 8 agreed to sell a $150 million convertible note to a Coatue Management vehicle, closing on July 1, 2024. The note carried 8.00% interest compounding quarterly, payable in cash or in kind, matured in June 2029, and converted at $16.395 per share, with the company able to redeem beginning at the two-year anniversary.¹⁵

Two things about this deal matter more than the coupon. First, the signal: Philippe Laffont's Coatue is a technology crossover investor, not a crypto fund, and its willingness to underwrite Hut 8 as an AI-infrastructure story arrived long before that framing was consensus. Genoot later noted that Coatue "backed us before the theme was a consensus."⁵ Second, the structure: convertible debt at a conversion price roughly 2.4 times the share price at which Genoot had taken over meant that if the strategy worked, the dilution would come at a price shareholders could live with — and if it did not, Hut 8 owed real money to a real creditor.

By mid-2026 the note was, in the CFO's words, "deeply in the money," the sole remaining piece of parent-level recourse debt, with the company holding the ability to force conversion as early as late June 2026 by electing redemption.⁴ Coatue's bet paid; the dilution arrived on Hut 8's terms rather than the market's.

Highrise AI: the small experiment

In September 2024, Hut 8 launched a GPU-as-a-Service business through a wholly owned subsidiary, Highrise AI, commissioning an inaugural cluster of 1,000 NVIDIA H100 GPUs in HPE Cray systems at a Tier 3 facility in Chicago, under a five-year agreement with an AI cloud developer combining fixed infrastructure payments with revenue sharing.[^16]

Measured against what came later, this was tiny — it added $7.4 million of year-over-year revenue in 2025.⁵ Measured as an information-gathering exercise, it was shrewd. Owning GPUs teaches a landlord what its tenants actually need: how clusters fail, what networking and cooling constraints bite, what "cost per token" means once depreciation is included. Genoot has since described Highrise as a "neo cloud" with its own software stack, bare-metal and multi-tenant offerings, and a CTO who spent a decade and a half running AI in the Israel Defense Forces, and has said the unit gives Hut 8 an understanding of the full stack "from electron to compute, from megawatt to token."⁵

The skeptical framing is equally available: a 1,100-GPU cloud competing against hyperscalers and specialist neoclouds is subscale by two orders of magnitude, and owning depreciating H100s while Blackwell-class silicon shipped is precisely the kind of capital commitment that destroys value when lease rates compress. Hut 8's decision to keep the GPU business small, private, and separate from the landlord business is the more telling signal — it suggests management shares that concern.

Because at the same moment Hut 8 was learning to be an AI landlord, the business it had told investors was its most attractive segment fell apart in a fortnight.

↑ Back to Top

VIII. Act VI: The Ionic Digital Shock & The Managed Services Reality Check (1:45:00 - 2:00:00)

The capital-light dream

Through early 2024, the most flattering slide in Hut 8's investor materials was Managed Services. The pitch was elegant: operate other people's mining fleets for fixed fees plus performance incentives, earn high margins, consume no capital, and let someone else own the depreciating machines. Its anchor client was Ionic Digital, the entity created out of the Celsius bankruptcy — the very restructuring Genoot's team had helped sponsor. Hut 8 managed roughly 12 EH/s, about 302 MW, on Ionic's behalf.

In June 2024 the two parties amended and restated the agreement, reducing the fixed cash fee to $15 million annually and — the detail that mattered — adding the right to terminate at will on 30 days' notice.[^18] At the time this read as a modest concession in exchange for continuity.

By the third quarter of 2024, Managed Services was Hut 8's largest single revenue line, generating $20.8 million of $43.7 million total quarterly revenue, or 47.6%.[^19] An investor reading that disclosure carefully would have seen a company whose biggest revenue segment rested on one customer holding a 30-day exit.

Thirty days

On November 26, 2024, Ionic served notice of termination effective December 10, 2024, announcing a transition to self-operations — and taking the position that no part of the $22.5 million termination fee was payable.¹⁶ Hut 8's response was blunt: the claims supporting the termination were "meritless," and the company would vigorously enforce its rights.¹⁶

Nearly half of quarterly revenue disappeared in fourteen days. The financial hole was visible for a full year afterwards: Power segment revenue fell to $23.2 million in 2025 from $56.6 million in 2024, a decline the CFO attributed directly to the Ionic termination, partially offset by tighter power markets elsewhere.⁵ As of the most recent disclosures reviewed for this piece, no public resolution of the fee dispute has been announced — meaning the $22.5 million remains a claim, not a receivable, and investors should treat it accordingly.

Separately, the securities litigation stemming from the USBTC merger resolved in June 2026: after a judge dismissed the Exchange Act claims entirely in September 2025 and allowed only Securities Act claims regarding King Mountain disclosure omissions to proceed, Hut 8 agreed to a $2.35 million settlement — roughly 19.6% of estimated recoverable damages — without admitting wrongdoing.²² Small in dollars; useful as a marker of how the 2023–24 period was ultimately judged.

The lesson, and what management did with it

The obvious lesson is customer concentration. The deeper one is about the phrase "capital-light." A services business with no contractual lock-in and no owned asset is not capital-light — it is asset-less, which means it has nothing to fall back on when the customer leaves. Hut 8 had been paid for its operating expertise; it had not been paid for owning anything the customer needed.

What is genuinely interesting is that management drew that conclusion rather than the comfortable one. Instead of chasing replacement service contracts, Hut 8 spent 2025 rebuilding the company around assets it owned and contracts that could not be cancelled for convenience.

Carving out the miners. On March 31, 2025, Hut 8 contributed substantially all of its ASIC fleet — carried at roughly $120 million of book value — into a venture with American Data Centers, a company backed by Eric Trump and Donald Trump Jr., taking 80% of the newly named American Bitcoin Corp. and becoming its exclusive infrastructure and operations partner under long-term commercial agreements.¹⁰ In May 2025 American Bitcoin announced a go-public transaction with Gryphon Digital Mining, and on September 3, 2025 the combined company began trading on Nasdaq as ABTC.¹¹ Genoot's own summary is characteristically direct: "I think we spun out $100 million of machines that probably would have gone liquidated for sub-$50 million."⁴ The structural point is that mining's capital cycle and volatility moved into a separate vehicle that funds itself, while Hut 8 kept the power and the buildings and collected rent — five-year triple-net leases at a 20–25% yield on cost, against a tenant targeting two-year paybacks on its chips.⁴

Building Vega. On June 30, 2025, Hut 8 energised Vega in the Texas Panhandle: 205 MW, 162,000 square feet, a proprietary closed-loop glycol-water direct-to-chip cooling design supporting up to 180 kW per rack, and capacity for up to 15 EH/s of next-generation ASIC servers.¹² Genoot has emphasised a different number: Vega was built for roughly $455,000 per megawatt, from a grass field, with no legacy constraints — and its 180 kW/rack density was developed when NVIDIA's reference designs were at 120 kW.⁵ Whether or not one accepts the cost claim at face value, the intent is clear: prove to prospective AI tenants that this team could engineer at densities the incumbents were only approaching.

Selling the gas plants. In February 2024, Hut 8 and Macquarie Equipment Finance had bought Far North Power Corp. — four Ontario natural gas plants totalling 310 MW, at Kingston, Iroquois Falls, Kapuskasing and North Bay — out of bankruptcy for $42.7 million, with Hut 8 holding about 80% of the venture.¹⁴ Hut 8 secured five-year capacity contracts across the portfolio, then agreed on November 17, 2025 to sell it to TransAlta for $95 million, closing in February 2026.¹³ Buy distressed, fix the commercial structure, contract the cash flows, sell to a strategic buyer — it is a small transaction that demonstrates a real capability, and the CFO used it pointedly: "A lot of data center companies now call themselves power-first. For us, this is not a marketing clip."⁴

Selling megawatts to the AI complex. On December 17, 2025, Hut 8 announced a fifteen-year triple-net lease for 245 MW of IT capacity at its River Bend campus in Louisiana with Fluidstack, with total base-term contract value of $7.0 billion and a 3% annual escalator — and, decisively, a backstop from an Alphabet subsidiary rated AA+/Aa2 covering all lease payments and pass-through obligations for the full base term.⁷ The lease sat inside a broader partnership under which Hut 8 would develop at least 245 MW and up to 2,295 MW for Anthropic, with a right of first offer on up to 1,000 MW of additional capacity at River Bend and joint diligence on up to 1,050 MW beyond it.⁸ Power came from Entergy Louisiana: 330 MW initially, with a path to roughly 1,000 MW more.⁷

Doing it a second time. On May 6, 2026 — the same morning it reported first-quarter results — Hut 8 commercialised the first phase of Beacon Point, a 1 GW campus in Nueces County, Texas, with a fifteen-year triple-net lease for 352 MW of IT capacity (about 500 MW of utility capacity) at a base-term contract value of $9.8 billion, a 3% annual escalator, three five-year renewal options taking potential value to roughly $25.1 billion, and expected average annual NOI of $655 million at stabilisation.⁶ The campus was designed around NVIDIA's DSX reference architecture with Jacobs as engineering and construction manager, Vertiv supplying infrastructure systems, and AEP Texas providing utility capacity; initial energisation was targeted for the first quarter of 2027 and first data hall delivery for the third.⁶ That transaction took Hut 8's total contracted IT capacity to 597 MW, aggregate base-term contract value to roughly $16.8 billion, and average annual NOI across both campuses to approximately $1.1 billion.⁶ Ten weeks later, on July 20, 2026, the same tenant took the rest of the campus.²

The reversal is worth stating plainly. In November 2024 Hut 8 lost half its revenue on 30 days' notice from a single customer. Thirteen months later it signed a fifteen-year, take-or-pay obligation from one of the strongest credits on earth, with no termination for convenience. That is not a pivot in messaging. It is a change in the legal character of the company's cash flows — and it is the single most important fact in this story.

↑ Back to Top

IX. Playbook: Business & Investing Lessons (2:00 - 2:15)

Strip away the narrative and Hut 8 in 2026 is a wager that a specific set of advantages compound. Some of them are real and demonstrable. Others are assertions that have not yet been tested by a full cycle. It is worth sorting them.

What a triple-net lease actually is

Because the entire investment case now rests on lease structure, it is worth translating the jargon, since these three features do most of the analytical work.

Triple net means the tenant pays operating expenses, property taxes and insurance directly; the landlord's only obligations are the building's structure and, in Hut 8's description, the landscaping. That is why management can say revenue converts to net operating income at above 99.9% margins — an almost absurd-sounding figure that is simply what happens when a landlord has no costs left to bear.⁴

Take-or-pay with no termination for convenience means the tenant owes rent whether or not it uses the capacity, and cannot walk away because its strategy changed. This is precisely the clause Hut 8 did not have with Ionic Digital — and its presence is the difference between a services contract and an infrastructure asset.

Credit backstop is the subtlest and most important. Fluidstack is a GPU cloud intermediary, not a household-name balance sheet; a fifteen-year, multi-billion-dollar obligation from such a counterparty would be nearly unfinanceable on its own. By having an Alphabet subsidiary rated AA+/Aa2 guarantee the lease payments and pass-through obligations for the full base term,⁷ the cash flow is effectively repriced to Google's credit. That single structural feature is what allowed a construction-stage project to reach investment grade — and it is also why investors should read "$7.0 billion contract" as a claim on a guarantor, not on the AI market.

Applying Helmer's 7 Powers

Cornered resource — the strongest claim, with an expiry date. Hut 8's pipeline stood at 8,375 MW across diligence, exclusivity, development and construction as of the first quarter of 2026.¹ More important than pipeline size is provenance: the Corpus Christi interconnect underlying Beacon Point was secured in 2023, before ERCOT tightened its interconnection study procedures — a permit that, by management's account, would be materially harder to obtain today.⁵ That is a genuine cornered resource, because the scarcity is created by regulation and queue mechanics rather than by capital.

The falsification test is straightforward: cornered resources are supposed to be hard to replicate. But every large miner is racing to convert the same kind of asset, utilities are actively expanding, and interconnection reform could in principle release supply. What Hut 8 owns is better described as a portfolio of time-limited options on power scarcity. The right question is not "does it have a moat?" but "how many gigawatts can it commercialise before the queue normalises?"

Process power — plausible, partly proven. The claim here is a repeatable development system: greenfield sites underwritten power-first, long-lead equipment ordered before announcement, contractors and engineering partners locked, financing arranged in parallel. On the Q1 2026 call Genoot stressed that in both River Bend and Beacon Point, 100% of long-lead equipment was ordered and all major contracts signed at announcement — that Hut 8 "didn't just announce a deal," it announced a program.⁴ The evidence for repeatability is two greenfield campuses commercialised in five months, neither converted from an existing mining site.⁴ The evidence against is that neither has delivered a single data hall yet. River Bend's first hall targets the second quarter of 2027, with additional halls roughly every 60 days thereafter.⁵ Process power in construction is only proven on handover.

Scale economies — early, and cutting both ways. Scale shows up most visibly in financing. Hut 8's River Bend project bond priced at $3.25 billion of 6.192% senior secured notes due 2042, fully amortising over 16.5 years, rated BBB– by both S&P (positive outlook) and Fitch, non-recourse to the parent, and described as the first investment-grade construction bond issued for a single-sponsor data center project.⁹ The financing was struck at approximately 95% loan-to-cost and returned $184 million of previously deployed equity at closing.⁴ The progression is instructive: initial discussions contemplated a construction loan at 85% LTC and SOFR+225; that improved to 90% at SOFR+240; then the investment-grade market opened and the team took a different path entirely.⁴ Each notch of cost-of-capital improvement is, in a spread business, pure equity value.

Counter-positioning — mostly rhetorical. Management likes to note that a traditional developer would not have underwritten the Corpus Christi interconnect because a traditional developer has no fallback use for the power. Hut 8's fallback was bitcoin mining, which let it hold the site without a signed AI tenant.⁴ That is a genuine structural advantage — but it is one every miner-turned-developer shares, which makes it a category advantage rather than a company advantage.

Porter's five forces

Supplier power remains extremely high, but it has changed shape. In mining, the dependency was on Bitmain and MicroBT for chips. In the landlord model, it is on the electrical supply chain — transformers, switchgear, breakers — and on partners like Vertiv and Jacobs for delivery. Hut 8's response has been to attack it structurally rather than complain about it: the company has begun designing and contract-manufacturing pieces of the stack itself, with Genoot describing a first-principles procurement approach that starts from raw material and labour costs rather than vendor quotes.⁴ Whether that produces durable savings is unproven, but the attempt is the right instinct for a business whose margin is set by cost per megawatt.

Rivalry is now the sharpest force. In mining, competition is a mathematical certainty against global difficulty. In AI data centers, Hut 8 competes against Core Scientific, IREN, TeraWulf, Cipher and Crusoe among converted miners — and against private, deeply capitalised developers who, in Genoot's estimate, do the large majority of transactions in the sector.⁵ Hut 8's answer is that as complexity rises, the number of developers who can deliver "is limited and shrinking."⁴ That is an assertion about the shape of a market whose supply response has not yet arrived.

Buyer power is the most underrated risk. Two campuses, two tenants. Hut 8 has narrowed the counterparty credit risk to almost nothing by insisting on AA– or better — Genoot disclosed the internal debate about that definition on the Q1 call⁴ — but concentration of demand is a different exposure from concentration of credit. Google's backstop guarantees Hut 8 gets paid at River Bend. It does not guarantee there is a third tenant for site three.

The treasury: from war chest to legacy position

The most consequential change in the investment case is one management describes almost casually. In 2024 the bitcoin hoard was existential — it was the balance sheet that let a small company tell hyperscalers it could finance a build. By early 2026 the CFO framed it as a source of liquidity to be optimised: a $200 million Coinbase facility was refinanced with FalconX at a 7% coupon, down from 9% and 450 basis points better than June 2024, unencumbering roughly 3,300 bitcoin worth about $260 million and bringing total unencumbered coins at Hut 8 to roughly 5,600.⁴ Genoot went further on the prior call: "the reality is we're going to remove Bitcoin exposure on our balance sheet as we move forward," with residual exposure held through the equity stake in American Bitcoin.⁵

For long-term shareholders this is the most important governance signal in the story, and it deserves to be named for what it is: a reversal. The company that pioneered the 100% HODL doctrine has concluded that holding a volatile asset on the balance sheet of a contracted infrastructure business raises the cost of capital rather than lowering it. That is analytically correct. It also means anyone who owns HUT as a bitcoin proxy is holding the wrong instrument, and should say so out loud before the next drawdown does it for them.

↑ Back to Top

X. Analysis & Skeptical Investor Stress Test (2:15 - 2:30)

Myth versus reality

Myth: Hut 8 is a bitcoin miner whose stock tracks the coin. The 2026 tape says otherwise, and violently. Hut 8's shares have risen more than 1,000% from the $6.77 at which Genoot took over in February 2024 — Genoot cited that figure himself in May 2026⁴ — while its own mining affiliate, American Bitcoin, fell more than 95% from a September 2025 peak of $139.65, executed a 1-for-15 reverse split to maintain Nasdaq compliance, and by mid-July 2026 carried a market capitalisation of about $402 million against 8,000 bitcoin worth roughly $509 million.²³ The parent re-rated on contracted leases; the affiliate de-rated below the value of its own coins. Rarely does a corporate structure separate two theses so cleanly, or so publicly.

Myth: the mark-to-market losses mean the business is deteriorating. The 2025 net loss of $248.0 million against 2024 net income of $331.4 million is almost entirely a bitcoin price artefact: a $220 million primarily unrealised loss in 2025 versus a $509.3 million gain the prior year.⁵ Underneath, revenue grew 45% to $235.1 million and gross margin expanded from 47% to 54%.⁵ In Q1 2026 the same pattern repeated more starkly — 226% revenue growth, gross margin from 14% to 64%, and a $253.1 million net loss driven by digital-asset marks.¹ The accounting is not lying; it is measuring something that will soon no longer be there.

Myth: the contracts are already cash flow. They are not. Not one dollar of the $16.8 billion contracted base has been earned. The Digital Infrastructure segment produced $1.3 million of revenue in Q1 2026 and the CFO expects it to become the primary growth driver only from the second quarter of 2027.¹ Everything between now and then is construction risk.

The three KPIs that matter

Forget hash rate. Forget bitcoin held. Three metrics determine whether this thesis works, and each is disclosed in a form investors can track.

Contracted IT capacity and its counterparty credit. Total contracted megawatts, the base-term contract value attached, and the credit rating standing behind it. This moved from zero to 597 MW and roughly $16.8 billion in five months, then to a fully contracted Beacon Point campus in July 2026.¹² It is the cleanest single measure of whether the pipeline converts.
Delivery against announced dates. The first River Bend data hall is targeted for the second quarter of 2027, with subsequent halls at roughly 60-day intervals.⁵ Genoot's own framing is that the contract is "a liability until we deliver and start generating cash flow."⁵ A slipped hall is worth more information than any quarterly revenue print.
Capital cost per megawatt against realised financing terms. Management guides to $9–11 million per MW of capex on these builds, against triple-net leases where revenue converts to net operating income at above 99.9% margins.⁴ Watch whether the next project financing prices near the River Bend bond's terms. The entire equity return is the wedge between the yield on cost and the cost of capital.

War-gaming the peer set

Judged against the other converted miners, Hut 8's position is distinctive on two axes and unremarkable on a third.

On asset provenance, Hut 8 is the outlier: both of its commercialised campuses are greenfield developments rather than conversions of existing mining sites, and management has been pointed about the fact that it has "yet to convert a single facility," leaving its legacy capacity available as optionality.⁴ Rivals who converted got to revenue faster; Hut 8 got a purpose-built asset and an untouched inventory of existing sites. Which was the better trade depends entirely on whether converted mining shells age well against rising rack densities.

On financing, Hut 8 has, at least once, done something no peer had: an investment-grade, fully amortising, non-recourse construction bond that returned equity at closing.⁹ Cost of capital is the whole margin in a leasing business, and this is the most concrete evidence in the story that the company is being repriced by professionals rather than by momentum.

On demand access, Hut 8 is in the same position as everybody else: dependent on a handful of hyperscalers and frontier labs whose capital plans are set in rooms it does not sit in. Its Anthropic relationship — joint diligence on up to roughly a gigawatt beyond River Bend⁸ — is a real head start, not a lock. When a competitor announces a similar backstopped lease, as several have, it is a reminder that the scarce asset is the tenant, not the developer.

The activist's brief

A skeptical investor with a sharp pen would file the following.

Governance and complexity. Hut 8 consolidates a separately listed subsidiary in which it holds roughly 55% of the economics but about 80% of voting power through Class B shares.¹ That means the income statement is polluted by mark-to-market swings on bitcoin that Hut 8 does not economically own in full, and that minority shareholders of ABTC and shareholders of HUT have differing interests in every related-party lease between them. Those leases — five-year triple-nets at a 20–25% yield on cost⁴ — are set between entities under common control. The terms may be fair. The structure guarantees the question gets asked forever.

Political adjacency. American Bitcoin's founding investor group included Eric Trump and Donald Trump Jr., and the company's identity is explicitly American-nationalist.¹⁰ For a business whose core activity is negotiating with state regulators, utilities and grid operators, proximity to a sitting president's family is a governance factor to be disclosed and assessed, not celebrated or ignored. It cuts both ways, and both ways are risks.

Compensation. Roughly $239.9 million of reported 2025 CEO compensation, and a 200% bonus payout, at a company reporting a GAAP loss.²¹ Performance-conditioned or not, the optics invite a say-on-pay fight in any year the stock stops rising.

Disclosure posture. Management has explicitly adopted a policy of not naming tenants. Genoot's stated reasoning on the Q1 2026 call — that most competitors are private, that deal terms attract third-party noise, and that customers prefer confidentiality — is commercially coherent.⁴ It is also a reduction in disclosure at precisely the moment investors are asked to capitalise fifteen-year cash flows from unnamed counterparties. "AA– or higher" is a credit statement, not an identity, and it requires trusting management's characterisation.

The claim that hasn't been tested. Hut 8 has never delivered a hyperscale AI data hall. Its construction track record is roughly a gigawatt of mining infrastructure, self-managed without general contractors⁴ — real, relevant, and not the same discipline as delivering liquid-cooled halls to a hyperscaler's schedule.

Risk radar

Concentration and correlation. Two campuses, two tenants, one of them credit-supported by Google. If AI capex expectations reset, the leases survive — but the pipeline's option value, which is most of the equity story, does not.

Construction and supply chain. Transformers, switchgear and cooling systems remain the binding physical constraint. Hut 8 says all long-lead items for both campuses are ordered.⁴ Delay damages, as Genoot notes, are not the real cost; credibility with a tenant whose GPU deliveries are scheduled around your handover is.

Financing. The River Bend bond removed refinancing risk at that project by fully amortising over 16.5 years.⁹ Beacon Point is not yet financed. Management has committed to a non-recourse structure there too,⁴ but a $9–11 million per MW build across 704 MW implies a capital requirement measured in billions, and credit markets that welcomed the first data-center construction bond may not welcome the tenth on the same terms.

Regulatory and social licence. ERCOT has revised interconnection study procedures; communities in several states have begun resisting data-center development. Genoot named public sentiment toward AI infrastructure as the macro issue he spends the most time on.⁴ This is now a real permitting risk, not a talking point.

Technology. GPU obsolescence is a genuine threat to Highrise's owned H100s. It is a lesser threat to the landlord business, where the tenant owns the chips — though rising rack densities do create the risk that a hall designed today is thermally obsolete before its lease ends. Hut 8's response, redesigning Beacon Point with NVIDIA as technology partner to lift IT capacity from 224 MW to 352 MW within the same land and utility footprint — and lifting base-term contract value from $6.2 billion to $9.8 billion in the process — is the strongest single piece of evidence that this team engineers rather than merely assembles.⁴

The bull case

Hut 8 owns scarce, pre-approved power in a market where power is the binding constraint; it has converted two greenfield sites into fifteen-year, take-or-pay, triple-net leases with counterparties rated AA– or better; it has financed one of them at investment grade, non-recourse, with equity returned at closing; it has removed nearly all parent-level recourse debt; and it retains an 8.4-gigawatt pipeline plus a demand relationship with Anthropic covering up to roughly a gigawatt of further joint diligence.¹⁴⁸⁹ If even a fraction of that pipeline converts on similar terms, the business becomes a contracted infrastructure company with approximately $1.1 billion of average annual NOI from its first two campuses alone,² and the equity is being valued on a different set of multiples than the miner it used to be.

The bear case

Nothing has been delivered. The entire re-rating rests on contracts whose first revenue arrives in mid-2027, executed by a team that has never built this asset class, funded by capital markets currently in an unusually generous mood toward anything with "AI data center" in the prospectus. Customer concentration — the precise failure mode that cost Hut 8 half its revenue in 2024 — has been reduced in credit quality but not in count. Related-party dealings with a majority-owned, publicly traded affiliate that trades below the value of its own bitcoin add complexity that a skeptic will discount. The bitcoin treasury that funded the pivot is being deliberately wound down, removing the optionality some shareholders own the stock for. And the cornered resource at the heart of the thesis is an option on regulatory scarcity, which regulators can unwind.

Both cases are legitimate. They resolve on the same date: the day the first data hall at River Bend energises.

↑ Back to Top

XI. Epilogue & Outro (2:30 - 2:40)

There is a version of this story that is simply about luck — a Canadian miner that happened to own interconnects when the world went mad for electricity. That version is wrong, but it is wrong in an instructive way.

Luck explains the asset. It does not explain the sequence. Between February 2024 and July 2026, Hut 8 shut its founding facility, exited legacy managed services after having them cancelled out from under it, bought four bankrupt gas plants for $42.7 million and sold them for $95 million,¹³¹⁴ carved its entire mining fleet into a separately listed vehicle, built a 205 MW data center at a cost per megawatt it claims is a fraction of industry norms, signed $16.8 billion of contracted base-term revenue across two greenfield campuses, and financed the first with the first investment-grade construction bond ever issued for a single-sponsor data center project.¹⁹ Institutional ownership went from under 10% to over 70% across that span.⁴ Whatever else one concludes, that is not a company being carried by circumstance.

What it is, still, is a company whose entire thesis lives in the future tense. Genoot's own stress test, offered unprompted on the Q1 2026 call, is the right one: what happens if I'm wrong? His answer is structural — fifteen-year contracts, investment-grade tenants, no termination for convenience, non-recourse debt, minimal risk capital deployed before a lease is signed (only 2.5% of River Bend's total investment went in before the December signing).⁴ It is a good answer. It is also, by construction, an answer about surviving a bad outcome rather than achieving a good one.

What to watch. Three things, in order. Does River Bend's first data hall energise in the second quarter of 2027, and do the following halls arrive at the promised 60-day cadence? Does Beacon Point get financed on terms comparable to River Bend's bond, without recourse to the parent and without equity issuance? And does a third campus get commercialised with a third tenant — the test of whether Hut 8 has built a repeatable development program or has simply been very good twice?

The company that began life as a distribution channel for someone else's chips now sells the one thing that cannot be shipped in a container: a place to plug in, at scale, on a date certain. Whether that is a generational infrastructure business or the most successful narrative trade of the AI cycle will not be settled by a press release. It will be settled by a switch, in a field in West Feliciana Parish, Louisiana, sometime in the spring of 2027.

Hut 8

Table of Contents

Hut 8 Corp.: From Bitcoin Pioneer to AI Power Broker

I. Introduction & Episode Roadmap (0:00 - 10:00)

The post-halving identity crisis

The subject of today's story

The narrative arc

Episode themes

II. The Physics of Bitcoin Mining: Energy, ASICs, and Grid Economics (10:00 - 25:00)

The arms race nobody can win

The three pillars of mining profitability

The grids

The competitive landscape

III. Act I: Canadian Origins & The Bitfury Golden Cage (2017–2020) (25:00 - 40:00)

Incubated, not founded

The capital advantage

The golden cage

The birth of the HODL doctrine

IV. Act II: Breaking the Monopoly & The First HPC Pivot (2020–2023) (40:00 - 55:00)

Breaking the shackle

The TeraGo acquisition: a hedge that was not a hedge

The bottleneck

V. Act III: The USBTC Merger & The Boardroom Coup (55:00 - 1:15:00)

Enter US Bitcoin Corp

The combination

The sudden coup

VI. Act IV: Asher Genoot's Restructuring Playbook (1:15:00 - 1:30:00)

Rationalising the asset base

The ERCOT build and the art of not mining

The operator's temperament

Incentive alignment, and its limits

VII. Act V: The AI Frontier & The Coatue Endorsement (1:30:00 - 1:45:00)

Coatue writes the first cheque

Highrise AI: the small experiment

VIII. Act VI: The Ionic Digital Shock & The Managed Services Reality Check (1:45:00 - 2:00:00)

The capital-light dream

Thirty days

The lesson, and what management did with it

IX. Playbook: Business & Investing Lessons (2:00 - 2:15)

What a triple-net lease actually is

Applying Helmer's 7 Powers

Porter's five forces

The treasury: from war chest to legacy position

X. Analysis & Skeptical Investor Stress Test (2:15 - 2:30)

Myth versus reality

The three KPIs that matter

War-gaming the peer set

The activist's brief

Risk radar

The bull case

The bear case

XI. Epilogue & Outro (2:30 - 2:40)

References