Core Scientific posts $1.16B loss as AMD inks 2.5GW AI deal

Core Scientific reported Q2 2026 results showing revenue of US$164.2 million（約260億円）, more than double the US$78.63 million（約130億円） reported a year earlier. However, the company posted a net loss of US$1.16 billion（約1900億円） and a basic loss per share from continuing operations of US$3.32, reflecting the substantial upfront costs of pivoting from Bitcoin mining to AI data-center hosting.

On the same day, Core Scientific and Advanced Micro Devices unveiled a long-term AI infrastructure partnership designed to significantly expand the company's capacity. The agreement could scale to 2.50 gigawatts of total capacity and includes more than 500 megawatts of U.S. infrastructure set to begin operations in 2027. As part of the arrangement, AMD will receive market-priced warrants to purchase Core Scientific stock, aligning the chipmaker's interests with the company's success.

The partnership directly addresses Core Scientific's key near-term catalyst—filling and monetizing new AI capacity—while simultaneously amplifying its biggest risk: the company must execute on billions of dollars of buildout commitments while the core business remains unprofitable and balance sheet pressure persists. Core Scientific must also manage existing contracts such as CoreWeave without creating overextension risk.

Core Scientific's narrative projects US$2.1 billion（約3400億円） in revenue and US$312.1 million（約500億円） in earnings by 2029, a target that requires 80.7% yearly revenue growth and an earnings increase of about US$1.4 billion（約2200億円） from the current -US$1.1 billion（約1800億円） loss position. Some of the most optimistic analysts were already modeling revenue near US$3.2 billion（約5100億円） and earnings of about US$1.7 billion（約2700億円） by 2029. The AMD deal will test whether the company can deliver on these ambitious forecasts or whether the market's sensitivity to execution risks—including overcapacity and customer concentration—will constrain the bull case.

Core Scientific faces a stark contrast between near-term operational reality and long-term strategic promise. The company more than doubled revenue year-over-year to US$164.2 million（約260億円） in Q2 2026, yet sustained a US$1.16 billion（約1900億円） net loss—a dynamic that reflects the capital-intensive nature of building out AI data-center infrastructure before those facilities generate sustained profitability. The AMD partnership announcement on the same day signals confidence from a major hardware supplier in Core Scientific's ability to absorb and deploy massive AI compute capacity, but it also locks the company into delivering 2.50 gigawatts of infrastructure while managing existing commitments such as CoreWeave.

The core tension is execution risk at scale. Optimistic analysts have already modeled revenue near US$3.2 billion（約5100億円） and earnings of about US$1.7 billion（約2700億円） by 2029; Core Scientific's own narrative projects US$2.1 billion（約3400億円） revenue and US$312.1 million（約500億円） earnings by the same year. Either way, the company must transform from a US$1.1 billion（約1800億円） annual loss position to profitability—a swing of roughly US$1.4 billion（約2200億円） or more—while simultaneously completing a multi-gigawatt buildout. The AMD deal may either validate that the AI hosting thesis can overcome execution risk, or it may expose how sensitive these forecasts are to overcapacity, customer concentration, and funding constraints.

No comments yet. Be the first to share your thoughts!