TeraWulf Q3 Report: HPC Accounts for 80% of Revenue—What’s the Outlook for the AI Computing Business?

The buzz surrounding AI infrastructure ultimately comes down to whether data centers can deliver, when customers will pay their rent, and whether the investment can be recouped. TeraWulf announced on November 6 thatThird-Quarter 2026 Results...provides a snapshot: Total revenue for the quarter was $76.2 million, of which $63 million came from high-performance computing (HPC) leasing revenue, accounting for approximately 83 percent.

This revenue structure indicates that HPC has become the core business for the quarter. However, the net income of approximately $709.9 million reported for the same period must be understood in the context of accounting items. The company disclosed that this figure was driven by non-cash changes in the fair value of warrants, while non-GAAP adjusted EBITDA for the quarter was negative $12.8 million. Relying solely on net income can easily lead to a misjudgment of the company’s operating performance.

First, figure out what the company sells.

When discussing AI computing power, models, chips, cloud services, and data centers are often lumped together. They operate at different stages of the value chain and generate revenue in different ways. When analyzing the financial results of infrastructure companies, you should first determine whether the revenue comes from facility leasing, computing services, or other business lines before deciding which metrics are appropriate for comparison.

In the data center leasing business, customers are not just paying for physical space; they also expect the facility to meet agreed-upon power supply, handover, and operational conditions. Signing long-term leases can improve demand visibility, but construction expenses are typically incurred upfront, while revenue is not realized until the facility meets the handover conditions. This time lag underscores the importance of financing arrangements and construction execution.

Therefore, the increase in HPC’s revenue share has clear significance: it indicates that the transformation is already reflected in revenue. However, the share itself does not answer the question of whether the return on investment is sufficient, nor does it indicate that every batch of new capacity will be subject to the same conditions. The analysis must continue to examine how much investment is required to generate the new revenue, whether costs are manageable, and when rental income will begin to contribute to cash flow.

Net income and operating cash flow are not the same figure.

Changes in fair value, simply put, refer to the remeasurement of the value of relevant financial instruments in accordance with accounting rules. The resulting book gains may boost current-period profits, but they do not mean that customers paid an equivalent amount more in rent during that quarter. Treating such net income directly as a measure of the business’s profitability confuses items of different natures.

On the other hand, adjusted EBITDA is not a one-size-fits-all solution that can replace all other metrics. It is subject to adjustments defined by the company and is not equivalent to operating cash flow, much less the cash that remains available for discretionary use after deducting capital expenditures. A reasonable approach is to examine revenue, profit adjustments, cash flow, and capital expenditures simultaneously, while ensuring consistency in the definitions used for each.

This distinction is particularly important for infrastructure companies that are expanding. A company may have a long order backlog but still need to continuously raise funds for construction; or it may post strong quarterly profits, but this does not necessarily mean that its cash flow pressures have disappeared.

A $19 billion contract must be placed within a 20-year timeline

The Anthropic lease disclosed by TeraWulf represents approximately $19 billion in contract revenue over the initial 20-year term. This figure represents the size of the long-term contract; it is not the amount of this funding round, nor is it cash received or third-quarter revenue. Framing it as “securing $19 billion in funding” would alter the nature of the event.

According to the companyForm 8-K filed on November 4... The relevant capacity is expected to be delivered in phases from the end of 2027 through the second quarter of 2028, and Anthropic’s lease obligations will commence upon the delivery of the corresponding data center. Google provides credit support for certain lease obligations; however, each form of support is subject to agreed-upon conditions precedent and cannot be construed as an unconditional assumption of all risks associated with the entire project.

Credit support helps facilitate discussions on financing terms, but it does not replace on-site construction, nor does it convert future rent into recognized revenue in advance. For long-term projects, it is more important to monitor whether delivery milestones have changed, whether financing has been secured, and whether actual rent is being recognized as planned. The total contract value provides a long-term outlook, while phased execution determines whether that outlook will be realized.

To Assess AI Infrastructure, We Need a Timeline

When monitoring TeraWulf going forward, focus on three key issues: whether the facilities already delivered will continue to generate revenue; whether new projects are progressing on budget and on schedule; and how new financing and construction investments are affecting cash balances. Comparing these metrics over several consecutive quarters is more useful than looking at a single profit figure in isolation.

The customer base is also worth examining. Large tenants can generate substantial orders, but the more a project relies on a small number of customers, the more important it is to understand the delivery, payment, and change conditions outlined in the contracts. Long-term contracts can reduce some of the uncertainty surrounding demand, but they also require facility operators to consistently deliver the agreed-upon service levels over the long term.

The opportunities and constraints outlined in its third-quarter report are quite specific. HPC leasing is currently driving revenue, but future growth depends on the successful completion of long-term projects. The key to understanding companies like this is to cross-check customer commitments, facility deliveries, and actual cash collections item by item—rather than substituting an assessment of operational details with the assumption of strong AI demand.