CoreWeave and Nebius emerge as tests of AI spending resilience
Heavy infrastructure spending and financing commitments leave both companies exposed if AI compute demand or utilisation weakens.
CoreWeave and Nebius are emerging as sharp tests of whether demand for AI infrastructure can withstand concerns about a slowdown in frontier development.
Michael Burry has challenged warnings about slowing AI progress, arguing that tighter development could benefit incumbent companies by limiting competition. AI-linked shares nevertheless sold off on 14 September, although the available material does not establish that the debate was the sole cause.
CoreWeave ended the second quarter with about US$104 billion in reported revenue backlog, after quarterly revenue more than doubled to US$2.58 billion, according to the source material. It also recorded US$9.4 billion in quarterly capital expenditure, a US$626 million net loss and US$35.6 billion in future debt principal payments.
Nebius reported second-quarter revenue of US$582.3 million and adjusted EBITDA of US$236.2 million. Customer commitments exceeded US$40 billion, while capital expenditure reached roughly US$5.7 billion. The company also closed about US$5.75 billion in convertible notes in August.
The commitments could provide visibility if AI spending shifts from frontier training towards inference and deployment. But it remains uncertain whether those commitments will translate into sustained revenue or high infrastructure utilisation, leaving both companies sensitive to any shortfall.
Insider Monkey data cited in the source showed hedge-fund ownership rising for both companies in the second quarter of 2026. About 45 million Nebius shares were reportedly sold short as at 31 August, equal to 17.89 per cent of the float, though the gross figure may include structural hedging related to the company’s convertible notes.
