Applied Digital Posts $126.6M Quarter but $100.9M Loss Raises Questions About Cloud Write-Down
Applied Digital Corporation (NASDAQ: APLD), which began as a blockchain infrastructure company before pivoting to AI-focused data centers, reported fiscal third-quarter 2026 results on April 8, with revenue surging 139% year-over-year but a large non-cash accounting charge pushing its net loss well past analyst expectations.

The Nasdaq-listed data center operator posted $126.6 million in total revenue for the three months ended February 28, 2026, beating the analyst consensus of $76.56 million by roughly $50 million. Despite that top-line performance, the company recorded a GAAP net loss of $100.9 million, or $0.36 per share. Analysts had forecast a loss of $0.16 per share. The shortfall traces almost entirely to a $59.7 million non-cash charge tied to the reclassification of its Cloud Services business ahead of a reverse merger that will combine the unit with EKSO Bionics Holdings to form ChronoScale Corporation, with Applied Digital retaining roughly 97% ownership.
From Applied Blockchain to AI Infrastructure
Applied Digital was founded as Applied Blockchain, Inc. before rebranding in November 2022 following the Ethereum Merge, which sharply curtailed proof-of-work mining economics and accelerated the company's shift toward high-performance compute hosting. That history makes it a bellwether for the broader mining-to-AI infrastructure transition reshaping the blockchain sector, and it explains why a NASDAQ data center company remains directly relevant to crypto markets.
The Write-Down Explained
Applied Digital has agreed to combine its Cloud Services unit with EKSO Bionics Holdings to form a new entity called ChronoScale Corporation. When the company signed that definitive agreement, accounting rules required it to remeasure the cloud unit at fair value rather than carrying it as an asset held for sale. The resulting $59.7 million loss is non-cash but still hits the GAAP income statement. Applied Digital expects to retain roughly 97% of ChronoScale after the transaction closes, which is targeted for the fourth fiscal quarter of 2026.
The full GAAP-to-adjusted swing is approximately $134 million. The $59.7 million charge accounts for a portion of that gap, with additional adjustments including stock-based compensation, depreciation, and amortization making up the remainder. After those adjustments, the company reported adjusted net income of $33.2 million, or $0.09 per diluted share, alongside adjusted EBITDA of $44.1 million.
The stock's after-hours reaction varied across sessions. Some reports cited a roughly 3% decline following the earnings release, while others tracked gains in a range of approximately 2.37% to 8%.
Shares traded between $27.35 and $29.39 during the regular session on April 8. The stock sat about 40% below its 52-week high of $40.20 heading into the print, even after a roughly 237% gain in calendar year 2025. Analysts carried an average price target of $45.27 on the shares heading into the report, compared with a pre-print price of approximately $25.22, according to Motley Fool.
AI Data Centers Now Lead Revenue
The company's HPC Hosting segment, which houses AI-focused data centers, generated $71.0 million in the quarter. That figure breaks down into $44.1 million in base rent, $18.9 million in tenant fit-out reimbursements, and $8.1 million in power pass-throughs. The legacy Data Center Hosting segment (Bitcoin mining colocation) contributed $37.5 million, up 7% year-over-year, with 286 megawatts of operating capacity running at full occupancy. Cloud Services contributed $18.1 million in revenue for the quarter; the segment was subsequently reclassified on the balance sheet in connection with the ChronoScale transaction.
CEO Wes Cummins pointed to demand conditions as a tailwind. "We are seeing a clear acceleration in demand for high-performance AI data center capacity," he said in the earnings release. Cummins also noted that annual capital expenditure from the largest U.S. hyperscalers has climbed to nearly $700 billion, describing those customers as "as aggressive as we have ever seen them."
Applied Digital's flagship Polaris Forge 1 campus in Ellendale, North Dakota runs at 400 megawatts and is fully leased to CoreWeave, the Nvidia-backed GPU cloud operator, under a contract valued at roughly $11 billion over 15 years. The company says the first 100 MW building is fully operational and uses direct-to-chip liquid cooling to manage the heat loads from large-scale AI model training. Cummins called it "one of the only 100 MW direct-to-chip liquid-cooled data centers online today."
The company closed the quarter with approximately $2.1 billion in cash and equivalents against $2.7 billion in total debt, following a $2.15 billion senior secured note offering. A second campus, Polaris Forge 2 in Harwood, North Dakota, is under construction at 200 MW and leased to an undisclosed investment-grade U.S. hyperscaler. That tenant holds a right of first refusal on up to 800 MW of expansion, representing up to $5 billion in contracted revenue. Combined contracted future revenue across Polaris Forge 1 and Polaris Forge 2 totals approximately $16 billion, according to Applied Digital's investor relations disclosures.
Applied Digital broke ground on its Delta Forge 1 campus in the southern United States during the quarter, with 430 MW planned and operations targeted for mid-2027. The company also maintains a 4 GW active development pipeline across multiple U.S. states, according to TIKR. Management has set a five-year target of $1 billion in net operating income.
Why This Matters Outside the United States
Applied Digital's trajectory is a working model of the crypto-to-AI infrastructure pivot, and it has direct implications for markets in South Asia and Africa. Bitcoin miners that have secured AI contracts are projected to derive less than 20% of total revenue from mining by the end of 2026, down from 85% in early 2025, according to CoinDesk reporting from March 2026. More than $70 billion in AI and HPC contracts have been signed across the sector, with competitors including Core Scientific and Hut 8 executing similar transitions. Hut 8 recently secured a $7 billion Google-backed deal as part of that shift.
In India, where data center capacity reached 1.93 GW in 2025 and is projected to reach 4 GW by 2028, the technical demands are converging with what Applied Digital has built. AI workloads require rack densities of 40 to 50 kilowatts per rack, roughly ten times standard configurations, creating the same need for liquid cooling that Applied Digital has deployed at scale in North Dakota. Microsoft has committed $17.5 billion to Indian AI infrastructure, Google has committed $15 billion, and Adani has pledged $100 billion over a decade. Analysts note that domestic operators are closely evaluating whether to build or to lease capacity from global players as those commitments take shape.
In Africa, sub-Saharan Bitcoin mining operators face a version of the same strategic crossroads. South Africa's data center market is projected to attract $1.5 billion in new investment in 2026, and up to 56 new facilities are expected to launch across the continent by 2027. Operators in Ethiopia and elsewhere that currently host mining rigs hold a broadly similar asset profile to the one Applied Digital leveraged for its AI pivot, including cheap power, large land parcels, and existing cooling capacity, according to analysis from CryptoMeter and ETF Trends. That analogy has limits: unreliable grid infrastructure across much of the continent remains a real constraint. The buildout also carries direct commodity implications for the region. AI data centers are projected to account for 2% of global copper demand by 2030, a figure with concrete stakes for the DRC, Zambia, and South Africa as dominant copper producers.
For developers in emerging markets building compute-intensive applications such as zero-knowledge proof systems or on-chain AI agents, analysts observe that the consolidation of hyperscale AI infrastructure in North American facilities deepens latency and cost dependencies that will not resolve without comparable regional buildout.