$DGXX: The Bitcoin Miner That Sold Itself to Cerebras

On May 5, 2026, a company that eighteen months earlier was still called Digihost Technology and was primarily running Bitcoin miners in upstate New York announced a 10-year, $1.1 billion Master Services Agreement with Cerebras Systems for a purpose built 40 megawatt AI colocation campus in Columbiana, Alabama. The stock ran to a market cap of roughly $723 million that day. It has since settled back to approximately $380 million. The contract has not changed.

The setup is unusual enough to be worth spelling out. The company’s Q2 2026 revenue was $6.63 million. Management is guiding to $250, 300 million for full year 2027. The sole analyst with a formal price target, HC Wainwright, has it at $9.00. The stock is at $3.72. Something is either deeply mispriced or deeply wrong, and we think it is worth figuring out which.

Our thesis is straightforward: Digi Power X is not a crypto company wearing an AI hat. It is an energy infrastructure business with permitted, grid connected power capacity that hyperscalers and AI compute customers cannot replicate on their timeline, and it has already signed the contract that proves the market will pay for that. The gap between management’s 2027 guidance and the street’s $201 million consensus estimate is 37%, and we think the market is wrong on the timing rather than wrong on the direction.

The 60 second version

• Digi Power X ($DGXX) is a vertically integrated AI infrastructure company building modular Tier III data centers and renting bare metal NVIDIA GPU compute, converted from a Bitcoin mining operation formerly known as Digihost Technology.
• The shift: AI compute demand has outrun available power secured colocation capacity, and the binding constraint is now grid interconnection, not chips or capital.
• The number that matters: a 10-year, $1.1 billion MSA with Cerebras Systems, with total potential value of $2.5 billion including renewal and expansion options, signed May 5, 2026.
• Catalyst with a date: Phase 1 of the Alabama Tier III campus, representing 15 MW of IT load, targeted ready for service in December 2026, with Phase 2 (25 MW) following in March 2027, the point at which material Cerebras colocation revenue begins flowing.
• Main risk: single customer concentration on Cerebras, Phase 2 financing not yet secured, and a 91% increase in shares outstanding over the past year that will continue if additional equity is needed.
• Market cap: approximately $380 million as of September 10, 2026, against $1.1 billion of contracted revenue spread over ten years.

Here is the fact that made us look twice. In Q2 2026, Digi Power X generated $1.1 million in GPU bare metal rental revenue from its initial fleet of B200 and B300 GPUs deployed at Columbiana over roughly five weeks of operation. That is a small number by any measure. What it represents is the first AI revenue this company has ever recorded, and the run rate it implies, annualized from five weeks of partial deployment, is the leading edge of what management says will be a $140 million annualized contracted revenue run rate during 2027.

The thesis in one sentence: Digi Power X controls permitted, grid interconnected power capacity in a market where time to power is the primary bottleneck for AI infrastructure deployment, has proven that capacity with a $1.1 billion signed contract, and is currently valued as though neither of those things is fully true.

We want to be clear about what we do not know. Phase 2 of the Alabama campus requires additional financing that has not been announced in final form. The 2027 guidance assumes that financing closes. The company has diluted shareholders substantially and retains the ability to do so again. These are not minor footnotes. They are the conditions under which the thesis either works or does not.

The shift

For most of the last decade, the constraint on AI compute infrastructure was semiconductors. You could not build fast enough because you could not get the chips. That changed as NVIDIA ramped production of its Hopper and then Blackwell architectures, and as hyperscalers placed orders large enough to guarantee supply. The constraint shifted.

What you cannot manufacture quickly is permitted, grid connected power capacity at scale. A new grid interconnection request in most US markets takes three to seven years to process. The queue in PJM alone stretched past 1,200 projects at last count. The companies that already hold interconnection agreements and permitted substations are sitting on something that cannot be replicated on any AI company’s deployment timeline, and the market has been slow to price that.

As Alliance Global noted in its coverage of DGXX, the biggest bottleneck for hyperscalers right now is time to power. That framing matters. It means that a company with 78 megawatts of operating power across New York sites under grandfathered status, a dedicated substation and power delivery agreement already in place in Alabama, approximately 40 acres adjacent to one of North Carolina’s largest electrical switchyards, and a letter of intent covering 1.3 gigawatts of potential power access in West Virginia is not primarily an infrastructure company. It is a queue skipping service for AI compute customers who have the chips but not the power.

The sector is at the early steep section of its adoption curve. Hyperscaler capital expenditure on AI infrastructure is running at a pace that the existing colocation market cannot absorb, and purpose built AI campuses are filling before they are finished. Digi Power X signed a 10-year contract for a campus that is not yet fully built. That sequence tells you where we are on the curve.

The company

Digi Power X started life as Digihost Technology, a Bitcoin mining operation with facilities in upstate New York. The name changed in March 2025. The business model changed more gradually, and the full pivot is still underway: Q2 2026 revenue still includes cryptocurrency mining as a segment, alongside energy sales and the new colocation line.

The core product today is the AI Ready Modular Solution platform, which the company calls ARMS. It is a modular Tier III data center design that can be deployed at sites where power infrastructure already exists. The customer facing compute product is branded NeoCloudz, a portal through which enterprise customers rent bare metal NVIDIA GPU capacity. The company deployed B200 and B300 GPUs at Columbiana and began billing in Q2 2026.

Michel Amar serves as CEO. The management team has been building the AI infrastructure pivot while still operating the legacy mining business, which is not an easy thing to do operationally, and the Q2 results show the friction: overall revenue was down 18% year over year as mining contribution fell while AI revenue was still in its first weeks. Adjusted EBITDA turned positive at $3.3 million in Q2, which is the clearest signal that the cost structure of the new business is better than the one being replaced.

One detail from the Q2 filing that does not fit the promotional narrative cleanly: the company committed $35 million toward NVIDIA Vera Rubin systems in June 2026, before Phase 1 of the Alabama campus is even live. That is an aggressive capital commitment on a facility that has not yet generated a single dollar of colocation revenue under the Cerebras contract. It reflects confidence in the pipeline. It also reflects the pressure of a 10-year contract with a major customer that expects equipment to be there when the building is ready.

The numbers and what the street expects

Q2 2026 revenue came in at $6.63 million, down 18% from Q2 2025. Net loss widened 38% year over year. By any conventional screen, these are bad numbers. The context is that they are bad on purpose: legacy Bitcoin mining revenue is being wound down while the AI build out burns cash without yet producing revenue at scale. Management described Q2 as a substantial turning point precisely because of the first AI revenue recognition, not because of the total.

The balance sheet, as of June 30, 2026, is the strongest element of the story. Cash and equivalents stood at $142.4 million, rising to approximately $150 million by August 14. Total assets were $279.6 million. Shareholders’ equity was $265.0 million. The company carries no long term debt. Approximately $110 million of capital expenditure had already been deployed toward GPU equipment and data center build out for the Cerebras contract through the end of Q2.

Gross margin in FY2025 was negative 9.4%. That number comes from the legacy mining business operating in an unfavorable cost environment, and it is not the right baseline for the AI colocation business. The adjusted EBITDA figure of $3.3 million in Q2, up $3.2 million year over year, is a better leading indicator of where margins go when the GPU fleet is billing. We would want to see two more quarters of positive adjusted EBITDA before calling the margin turn confirmed.

Management is guiding for Q3 2026 revenue to increase by over 100% compared to Q2, implying roughly $13 million or above. There is no formal sell side Q3 consensus available, which means that number is purely management’s forecast against no external check. We are flagging that explicitly. The guidance is plausible given that GPU utilization was ramping through the quarter, but there is no independent estimate to triangulate against.

For full year 2027, management guided to $250, 300 million in revenue. The analyst consensus, drawn from an extremely thin coverage base, sits at approximately $201 million. The midpoint of management guidance is $275 million, which is 37% above the consensus figure. That gap is either a sign of management overconfidence or of a consensus that has not yet been updated for the Cerebras contract’s contribution schedule. Given that Phase 2 of the Alabama campus, carrying 25 MW of IT load, is targeted for March 2027, and that is when material colocation revenue begins, we think the consensus has simply not modelled the Phase 2 ramp.

The company also anticipates reaching approximately $140 million of annualized contracted revenue run rate during 2027, based on existing contracts and scheduled deployments. That figure is not the same as the 2027 full year revenue guidance, which is higher, implying the company expects additional contracts to be signed before the year closes. HC Wainwright maintains a Buy with a $9.00 target, raised from $2.50 in November 2025. Alliance Global lowered its target to $9 from $12 after the Q2 print on August 14, citing execution risk on Phase 2 financing. Two analysts, two targets, both at $9.00, both well above the current price. The coverage base is thin enough that a single new initiating analyst would meaningfully change the visibility of this name.

Dilution is the number the bulls underweight. Shares outstanding have risen 91% in the past year. The company has an active ATM program allowing up to $200 million in additional share sales, and had already distributed roughly 20 million shares for gross proceeds of approximately $76.5 million under a prior supplement. Phase 2 requires additional financing. If that financing comes via equity rather than debt, existing shareholders absorb another round of dilution before the revenue materializes.

Why it wins

The moat is not the GPU fleet. NVIDIA sells B200s and B300s to many buyers. The moat is the power. Specifically, it is power that is already permitted, already interconnected, and already under agreement with the local utility. You cannot manufacture that moat in the timeframe that matters for a 10-year AI infrastructure contract signed in 2026.

Digi Power X controls 78 megawatts across its New York sites, operating under grandfathered status that competitors cannot simply replicate. The Alabama campus has a dedicated on site substation and a live power delivery agreement with Alabama Power. The West Virginia LOI covers a potential 1.3 gigawatts, which, if converted to real agreements, would make this one of the larger power secured land banks in independent AI infrastructure. The North Carolina acreage sits adjacent to one of the state’s largest electrical switchyards.

The conversion from Bitcoin mining to AI colocation is the specific mechanism by which this company captures the shift. Mining facilities already have high power electrical infrastructure, cooling systems, and often grid agreements. Repurposing that infrastructure for GPU compute colocation is faster and cheaper than building from scratch, which is exactly what Cerebras is paying for. A competitor starting from raw land in 2026 cannot get interconnected and permitted in time to compete for the contracts being signed right now.

What could go wrong

The bear case has one central argument and several supporting ones. The central argument is customer concentration. The entire Alabama thesis rests on Cerebras. If Cerebras has difficulty with its own financing or competitive position, the anchor contract is at risk. A 10-year MSA is a serious legal commitment, but it is worth asking what happens to DGXX’s valuation if that contract is renegotiated, delayed, or disputed. The answer is that most of the current premium disappears.

Phase 2 of the Alabama campus, representing 25 MW of the 40 MW total and therefore the majority of the contracted capacity, is contingent on additional financing that has not yet been secured in final form. The discussions described as “advanced” with a major bank are unconfirmed in their terms. If Phase 2 financing does not close on schedule, the March 2027 revenue ramp does not happen, and the 2027 guidance becomes impossible to achieve. That is not a tail risk. It is the central execution dependency.

Dilution may continue. The ATM program is active, Phase 2 needs capital, and the company has already demonstrated willingness to issue equity at prices well below where analysts have their targets. A shareholder who buys today and is right about the 2027 revenue picture may still underperform if the share count grows another 40% between now and when the revenue arrives.

The stock’s weekly move averages 16%, which puts it in the top 10% of US equities by volatility. That cuts both ways, but a negative surprise on Phase 1 timing, a Cerebras related headline, or a broad risk off period in small cap AI names could easily return the stock to the sub-$2 range it visited in September 2025. The short interest has risen 737% since that period and now sits at 5.8% of float, which is not extreme but is growing, and growing short interest in a low float name means someone is making a structured bet that the execution misses.

The investment thesis

The thesis: Digi Power X holds a power secured AI infrastructure position that the market is valuing at roughly $380 million against $1.1 billion of signed contracted revenue, and the gap between those numbers closes as Phase 1 delivers in December 2026 and Phase 2 begins billing in March 2027.

What has to be true for this to work: Phase 1 of the Alabama campus comes online on schedule in December 2026. Phase 2 financing closes, either through the discussed debt arrangement or through additional equity. Cerebras remains a viable counterparty through at least the first two years of the contract. GPU utilization on the NeoCloudz platform continues ramping through Q3 and Q4 2026, providing evidence that the business outside the Cerebras contract is also working.

What the market appears to believe instead: the 2027 guidance is too aggressive, Phase 2 is uncertain enough to discount heavily, and the dilution risk makes the equity level return unattractive even if the infrastructure thesis is correct. The consensus revenue estimate of $201 million for 2027 implies the street is modelling roughly the annualized contracted run rate of $140 million and adding a small increment for other activity, rather than crediting management’s guidance of $250, 300 million.

The specific gap we are watching is the Phase 2 financing announcement. That single event converts a contingent asset on the balance sheet into a confirmed revenue line. It is not in estimates in any form we can find. If it closes before December 2026 and Phase 1 delivers on time, the 2027 consensus gets revised upward significantly, and this is a name that re rates on two events in two months rather than one.

This is a name to watch closely through the December 2026 Phase 1 delivery, not to buy and file away. The thesis plays out over a six to twelve month window, and there are enough binary dependencies that position sizing relative to conviction on Phase 2 financing matters more than the headline contract number. The falsification is simple: if Phase 2 financing is not announced by Q1 2027, the 2027 revenue guidance is unachievable, the thesis is broken, and the correct response is to reassess rather than to wait.

What we are watching

• October 1, 2026: Management’s targeted date for the “55 MW AI rollout” milestone. Specific details on what this includes beyond Phase 1 have not been fully disclosed; we want to see a formal announcement confirming scope and timing.
• Mid November 2026 (estimated): Q3 2026 earnings release. The test is whether revenue exceeded $13 million as guided, whether GPU utilization data supports the Q4 ramp, and whether any update on Phase 2 financing is provided.
• December 2026: Phase 1 of the Alabama Tier III campus targeted ready for service. A formal commissioning announcement from the company or a filing confirming service commencement is the single most important observable between now and year end.
• Q1 2027 (target March): Phase 2 (25 MW) scheduled to come online. Any financing announcement that enables Phase 2 before this date would be a major positive revision to the 2027 estimate picture.
• Ongoing: ATM program drawdown activity via any S-3 prospectus supplements filed with the SEC, which would signal the pace and scale of additional dilution before Phase 2 closes.

The bottom line

The power bottleneck in AI infrastructure is real, and Digi Power X has more of the relevant asset, permitted grid connected megawatts, than its market cap currently reflects. The Cerebras contract is not a letter of intent or a memorandum of understanding. It is a signed 10-year MSA with a take or pay structure at $1.1 billion, negotiated by a company that needed a partner badly enough to commit to a dedicated campus. That contract changes what DGXX is, even if the stock price has not fully caught up.

The dilution risk is real and the Phase 2 financing dependency is real, and we are not dismissing either. A thesis that requires two things to go right in sequence over six months, in a stock that moves 16% in a typical week, is a thesis that demands active monitoring rather than passive holding. The 2027 revenue picture is compelling if Phase 2 closes. It is substantially less compelling if it does not.

The thing that would change our mind is not a bad quarter. A bad quarter in Q3 or Q4 2026, if it reflects timing rather than utilization failure, is noise given the December Phase 1 catalyst ahead. What would change our mind is a Phase 2 financing announcement that does not come, or a Cerebras related headline that puts the contract’s counterparty at risk. Either of those events breaks the primary revenue event entirely, and the stock price would reflect that faster than any re rating on the upside would.

Not financial advice.

References

1. Digi Power X Inc. Form 8-K Q2 2026 Earnings Release. US Securities and Exchange Commission. Published August 14, 2026. Accessed September 10, 2026. https://www.sec.gov/Archives/edgar/data/0001854368/000121390026089450/ea030173001ex99-1.htm
2. Digi Power X Inc. Form 8-K Q2 Earnings Date Announcement. US Securities and Exchange Commission. Published August 4, 2026. Accessed September 10, 2026. https://www.sec.gov/Archives/edgar/data/0001854368/000121390026084921/ea030047601ex99-1.htm
3. Yahoo Finance. DGXX Quote and Summary. Accessed September 7, 2026. https://finance.yahoo.com/quote/DGXX/
4. SimplyWallSt. Digi Power X Stock Analysis. Published September 8, 2026. Accessed September 10, 2026. https://simplywall.st/stocks/us/software/nasdaq-dgxx/digi-power-x
5. StockTitan. DGXX Overview and Short Interest. Accessed September 2026. https://www.stocktitan.net/overview/DGXX/
6. StockTitan. Digi Power X Reports Second Quarter 2026 Financial Results and First AI Revenue. Published August 14, 2026. Accessed September 10, 2026. https://www.stocktitan.net/news/DGXX/digi-power-x-reports-second-quarter-2026-financial-results-and-bks40f9i2tkx.html
7. Digi Power X Inc. (DGXX) Q2 2026 Earnings Call Transcript. The Globe and Mail via Motley Fool. Published August 14, 2026. Accessed September 10, 2026. https://www.theglobeandmail.com/investing/markets/markets-news/motley/3863063/digi-power-x-dgxx-q2-2026-earnings-call-transcript/
8. Digi Power X Inc. (DGXX) Q2 2026 Earnings Call Transcript. Motley Fool. Published August 14, 2026. Accessed September 10, 2026. https://www.fool.com/earnings/call-transcripts/2026/08/14/digi-power-x-dgxx-q2-2026-earnings-call-transcript/
9. Digi Power X Inc. Q2 2026 Earnings Call Summary. Yahoo Finance via Moby. Accessed August 2026. https://finance.yahoo.com/markets/stocks/articles/digi-power-x-inc-q2-205051694.html
10. Digi Power X Q2 2026 Earnings Call Highlights. Yahoo Finance. Accessed August 2026. https://finance.yahoo.com/markets/stocks/articles/digi-power-x-q2-earnings-140427175.html
11. Digi Power X Reports Second Quarter 2026 Financial Results. Yahoo Finance. Published August 14, 2026. Accessed September 10, 2026. https://finance.yahoo.com/technology/ai/articles/digi-power-x-reports-second-113000956.html
12. Digi Power X Targets Phase 1 Ready for Service in December 2026 While Anticipating Q3 Revenue Over 100% Growth. Seeking Alpha. Published August 14, 2026. Accessed September 10, 2026. https://seekingalpha.com/news/4633411-digi-power-x-targets-phase-1-ready-for-service-in-december-2026-while-anticipating-q3-to
13. Proactive Investors. Digi Power X Secures $1.1B Data Center Contract with Cerebras Systems. Published May 5, 2026. Accessed September 10, 2026. https://www.proactiveinvestors.com/companies/news/1091732/digi-power-x-secures-1-1b-data-center-contract-with-ai-chip-maker-cerebras-systems-1091732.html
14. Digi Power X Secures $1.1B Data Center Contract with Cerebras. Yahoo Finance. Published May 5, 2026. Accessed September 10, 2026. https://finance.yahoo.com/sectors/technology/articles/digi-power-x-secures-1-132900027.html
15. Data Center Dynamics. Cerebras Signs on for 40MW of Capacity at Digi Power X Data Center in Alabama. Published May 8, 2026. Accessed September 10, 2026. https://www.datacenterdynamics.com/en/news/cerebras-signs-on-for-40mw-of-capacity-at-digi-power-x-data-center-in-alabama/
16. Baxtel. Cerebras Secures 40MW at Digi Power X Alabama Facility. Published May 8, 2026. Accessed September 10, 2026. https://baxtel.com/news/cerebras-secures-40mw-at-digi-power-x-alabama-facility
17. StockTitan. Digi Power X Signs $1.1B 40MW AI Campus Deal with Cerebras. Published May 5, 2026. Accessed September 10, 2026. https://www.stocktitan.net/news/DGXX/digi-power-x-signs-ai-colocation-agreement-with-leading-ai-compute-6hqq2m3hl0lg.html
18. WebDisclosure. Digi Power X Inc. Reports Second Quarter 2026 Financial Results and Provides Operational Update and 2027 Outlook. Published August 14, 2026. Accessed September 10, 2026. https://www.webdisclosure.com/press-release/digi-power-x-inc-nasdaq-dgxx-digi-power-x-reports-second-quarter-2026-financial-results-and-provides-operational-update-and-2027-outlook-a3K92htXBI1
19. Digi Power X Secures NVIDIA Vera Rubin Systems Advancing Its AI Infrastructure. Nasdaq Press Release. Published June 3, 2026. Accessed September 10, 2026. https://www.nasdaq.com/press-release/digi-power-x-secures-nvidia-vera-rubin-systems-advancing-its-ai-infrastructure-and
20. Digi Power X Commits $35M to NVIDIA Vera Rubin Systems. Yahoo Finance. Published June 3, 2026. Accessed September 10, 2026. https://finance.yahoo.com/sectors/technology/articles/digi-power-x-commits-35m-124700752.html
21. StockTitan. Digi Power X Q1 2026 Earnings: $4.7M Loss, Positive Adjusted EBITDA. Published May 15, 2026. Accessed September 10, 2026. https://www.stocktitan.net/sec-filings/DGXX/8-k-digi-power-x-inc-reports-material-event-aa4cca118fb1.html
22. GuruFocus. Digi Power X (DGXX) Reports Q1 Loss and Revenue Decline, Sets Ambitious 2027 Goals. Published May 15, 2026. Accessed September 10, 2026. https://www.gurufocus.com/news/8863070/digi-power-x-dgxx-reports-q1-loss-and-revenue-decline-sets-ambitious-2027-goals
23. Seeking Alpha. DGXX Earnings Estimates. Published March 17, 2026. Accessed September 10, 2026. https://seekingalpha.com/symbol/DGXX/earnings/estimates
24. Intellectia AI. Digi Power X (DGXX) Q2 2026 Earnings Call Transcript. Accessed August 2026. https://intellectia.ai/news/stock/digi-power-x-dgxx-q2-2026-earnings-call-transcript
25. StockAnalysis. DGXX Statistics and Valuation. Accessed September 2026. https://stockanalysis.com/stocks/dgxx/statistics/
26. StockAnalysis. DGXX Market Cap History. Accessed September 2026. https://stockanalysis.com/stocks/dgxx/market-cap/
27. Seeking Alpha. DGXX Stock: Business Improved While Shareholders Diluted. Published July 23, 2026. Accessed September 10, 2026. https://seekingalpha.com/article/4925072-digi-power-x-stock-business-improved-while-shareholders-diluted
28. SimplyWallSt. Digi Power X Future Forecasts. Accessed September 2026. https://simplywall.st/stocks/us/software/nasdaq-dgxx/digi-power-x/future
29. StockTitan. Digi Power X Files Amended and Restated Prospectus Supplement for $200M ATM Program. Published November 18, 2025. Accessed September 10, 2026. https://www.stocktitan.net/news/DGXX/digi-power-x-files-amended-and-restated-prospectus-supplement-for-nvjwysrdc0m1.html
30. Nasdaq. Digi Power X Announces Nasdaq Ticker Symbol Change to DGXX. Published March 17, 2025. Accessed September 10, 2026. https://www.nasdaq.com/press-release/digi-power-x-announces-nasdaq-ticker-symbol-change-dgxx-2025-03-17
31. The AI Consulting Network. Digi Power X and Cerebras $1.1B Alabama AI Data Center Deal Analysis. Published May 9, 2026. Accessed September 10, 2026. https://www.theaiconsultingnetwork.com/blog/digi-power-x-cerebras-1-1b-alabama-ai-data-center-cre-investors-2026