A：The financial results for the first fiscal quarter ending September 30, 2025, include a fifth consecutive quarterly increase in revenue reaching $240 million and adjusted EBITDA of $92 million.

A：The $9.7 billion AI cloud contract with Microsoft is a defining milestone for the business, validating the company's position as a trusted provider of AI cloud services, opening access to a new customer segment among global hyperscalers, and expected to generate approximately $1.94 billion in annual recurring revenue.

A：The company plans to scale its GPU fleet from 23,000 GPUs to 142,000 GPUs by the end of 2026, which is expected to support approximately $3.4 billion in annualized run-rate revenue.

A：The benefits of the company's vertical integration model include control over the entire stack from substation to GPU, which eliminates dependence on third-party colocation providers and removes counterparty risk. This allows for faster GPU deployments with control over execution, uptime, and customer service, resulting in scalability, cost efficiency, and superior performance.

A：The company plans to manage its growth in a capital-efficient manner by leveraging existing data centers for the deployment of additional GPUs. The incremental 40,000 GPU build-out will be executed efficiently, with the ASIC to GPU swap-out program progressing well across its campuses.

A：The British Columbia Data Center supports the expansion by fast-tracking growth in high-performance AI workloads and scaling the company's GPU fleet into one of the largest in North America, which is part of the ASIC to GPU migration process.

A：The next phase of data center construction is focused on accelerating the development of Horizons 1 to 4 to accommodate the phase delivery of Nvidia GPUs for Microsoft. The facilities are being engineered to tier 3 equivalent standards for concurrent maintainability, with a key feature being the establishment of a network core architecture capable of supporting a 100 MW supercluster for high-performance AI training with future-proofing for next generation GPUs. This phase also incorporates Flex rack densities to accommodate future chip generations without requiring major structural overhauls, providing long-term value protection and reducing recontracting risks.

A：Sweetwater, the flagship data center in West Harvey, Texas, has a full build-out potential to support up to 2 GW (2000 MW) of gross capacity, all of which has been secured from the grid. While recent headlines have been more about AI cloud expansion at other sites, Sweetwater remains an exciting platform asset poised to service the wave of AI compute demand. Sweetwater One energization is on schedule with more than 100 people mobilized on site to support the construction of one of the largest high voltage data center substations in the United States.

A：For Q1 FY 26, the company reported a total revenue of $240 million, marking a 28% quarter-over-quarter and a 355% year-over-year increase. Operating expenses increased due to higher depreciation and share-based payment expenses resulting from a materially higher share price. Net income and EBITDA were $385 million and $663 million, respectively, both significantly up from the previous period, largely attributable to unrealized gains on prepaid forward and capco transactions related to convertible note financings. Adjusted EBITDA was $92 million, reflecting continued margin strength.

A：The AI Cloud partnership with Microsoft is a significant milestone for the company, expected to deliver approximately $1.9 billion in annual revenue once the four phases come online, with an estimated 85% project EBITDA margin. This strong margin incorporates all direct operating expenses across cloud and data center operations, including power, salaries, wages, maintenance, insurance, and other direct costs. The investment in cloud returns is supported by a $5.8 billion CapEx for GPUs and ancillaries, which, after internal colocation charges, is expected to yield robust returns on the cloud investment.

A：The company is planning to finance the transaction with Microsoft in a capital-efficient way by aligning payments for CapEx with the phase delivery of GPUs across 2026, receiving customer prepayments of 20% of total contract revenue upfront, and leveraging the strong credit profile of Microsoft to potentially raise significant additional funding secured against the GPUs and contracted cash flows. The company is targeting approximately $2.5 billion through these initiatives, with options for funding the remaining balance through existing cash, operating cash flows, equity, convertible notes, and corporate instruments. The company also issued 1 million in 0-coupon convertible notes and secured GPU financing to support AI cloud expansion.

A：The strategic value of servicing a large technology company with a proprietary data center design is twofold: it demonstrates to capital markets and investors that the company is on the right track, and it provides validation in the broader customer ecosystem.

A：The expected unlevered IRR for the cloud returns transaction is low double digits. When a $2.5 billion additional leverage is taken into account, the levered IRR is expected to be in the order of 25 to 30%. With a potential $3 billion leverage package, the levered return could increase by an additional 10%.

A：The deal with Microsoft includes a colocation charge structure that projects revenue for the data centers at an internal rate of $130 per kilowatt per month, which is considered a market rate of return. The first five years of this arrangement is underwritten by a hyperscale credit, contributing to the revenue projections.

A：The company receives a third of the CapEx for the GPU fleet upfront, which amounts to $3.4 billion by the end of 2026. This allows the company to retain a larger percentage of equity while still accessing significant capital for growth.

A：The anticipated number of GPUs to be deployed for the Microsoft deal is not explicitly stated in the transcript. However, the data center design is future-proofed by accommodating up to 200kW of rack power, which is more than the 135kW utilized by the GB 300 S GPUs. This indicates the potential for future generations of chips to be supported in the data centers. Other enhancements include full tier 3 equivalent concurrent maintainability, designed to handle multiple generations of GPUs.

A：The company continues to see very good levels of demand across all offerings, particularly for air-cooled servers that are ideal for 500 to 4000 GPU clusters with rapid scalability capabilities. The transition of existing data centers to AI workloads is a quick process, which allows the company to meet the growth requirements of customers in that category effectively.

A：The company offers high-end liquid-cooled capacity (Horizon One build-out) from Nvidia, which is in high demand. This allows them to serve a diverse range of customers from hyperscalers and foundational AI labs to large enterprises and top-tier AI startups, as well as smaller scale inference and enterprise users.

A：The company is seeing strong interest in cloud services, including GPU as a service and colocation, across all potential service offerings, including at the Sweetwater site. Initial energization is expected in April 2026, and while it is early days from a construction perspective, the potential return profile and future cash flow from engaging in the cloud business at Sweetwater is very attractive.

A：The $5.8 billion order from Dell is mostly allocated to GPUs, with a substantial amount also set aside for top-tier networking equipment for the GPU clusters, specifically the Infiniband offering. The future viability of the specific equipment for the clusters, however, remains to be seen, as it may not be reusable for future generations of GPUs.

A：The transition of data centers to AI workloads is relatively cost-efficient with minimal CapEx requirements. This involves mainly removing existing AC systems and replacing them with standard data center racks and power distribution units (PDUs) compatible with AI servers. In some cases, adding redundancy might be required, which could cost around $2 million per megawatt. However, the overall process is capital and CapEx efficient, especially in the context of large-scale liquid cooled capacity deployments like those at Horizon.

A：The company has never missed a construction or commissioning date as a listed company, indicating a strong track record of on-time delivery.

A：The CapEx includes infrastructure for a 100 MW supercluster, which necessitates significant additional infrastructure. Some of the costs mentioned are related to this capability and may not be a requirement for all customers, offering a level of uniqueness in the potential impact from tariffs or networking cabling.

A：Out of the initial 23,000 GPUs, 11,000 have been contracted, and an additional thousand GPUs have been contracted subsequently. The remaining GPUs are mainly for delivery later in the timeline, and while there is increased appetite from customers to pre-contract, discussions are still ongoing for a significant portion of the yet-uncontracted capacity.

A：The speaker views risk differently in cloud deals compared to colocation deals. Colocation deals with hyperscalers have a lower equity return potential over the contract term and rely on recontracting after the term for any return. In contrast, the cloud deal has a better risk profile. For short-term colocation contracts without hyperscale credit, there's a risk of not being able to refresh GPUs against companies with insufficient balance sheets.

A：Upon the expiry of the Microsoft contract, the company has optionality to sign a colocation deal, relaunch a new cloud offering with the latest generation GPUs, or potentially operate the GPUs for additional years as with prior generations like the A100s.

A：The company is uncomfortable with the trade-off of transferring an asset, which is tied to technology and demand growth, for bond exposure to the counterparty, especially when the counterparty may not have a strong balance sheet. The colocation deals involve exchanging equity exposure for bond positions, which is not seen as a favorable decision for shareholders.

A：The projected IRR of 35% equity IRR is calculated by considering the equity portion of the IRR and assuming it is funded with 100% ordinary equity. This is deemed conservative given the company's track record of raising convertibles and lack of debt at the corporate level, and it aligns with a trillion-dollar credit and efficient funding.

A：The company's current strategy focuses on providing a bare metal offering that caters to highly advanced AI and software companies requiring raw compute and performance benefits. While there has not been a significant demand for anything other than the bare metal service, the company continues to monitor customer needs and is prepared to layer in additional software if required by customers in the future.

A：The company does not develop its own software and operating layer because the largest and most sophisticated technology companies prefer to access GPUs to run their proprietary software. These companies have a competitive advantage in their respective spaces and are looking for bare metal infrastructure. As the market develops, there may be a role for the company to service smaller customers without the internal capability or budget, but for the company's current strategy of pursuing scale, focusing on the commoditized software market is not seen as a pathway to achieving scale.

A：The direct fiber loop between Sweetwater 1 and Sweetwater 2 adds an additional layer of optionality for the customers. It enables the potential for scaling deployments where the two sites can operate as an individual campus despite being physically separated. This capability is valuable for customers seeking scale and is the reason the company pursued the direct fiber connection.