OpenAI's Expanding Compute Bill: A Shifting Landscape
OpenAI (find out more) informed investors in February that its anticipated compute bill would stand at approximately $600 billion by 2030. However, a July presentation revealed a significantly higher figure of $856 billion.
The burning (cash burn) figure, a key metric, has seen a reversal, decreasing from roughly $305 billion in May to $278 billion in July. Despite this apparent positive trend, the free cash flow remains negative, projected at -$278 billion between 2026 and 2030.
This contrast is particularly notable given that OpenAI's revenue is estimated to reach $350 billion by 2030, up from approximately $36 billion in the current year.
The Rising Compute Costs
The February announcement set a public target of $600 billion for compute by 2030, which was considered a conservative estimate at the time. However, a private presentation made in July reveals a much higher figure of $856 billion. This represents an increase of approximately 43% over the publicly stated target.
It's important to note that the categories of "compute" and "computing power and infrastructure" may not be entirely identical, leaving room for definitional differences contributing to the gap.
Financing the Build
The unusual combination of rising spending and decreasing burn is attributed to external financing. OpenAI, lacking an investment-grade credit rating, relies on partners and vendors to fund its expansive buildout:
- Nvidia has been in discussions to guarantee $250 billion of data center debt, allowing lenders to base pricing on Nvidia's creditworthiness.
- Oracle, similarly, is investing heavily in data centers, much of it against OpenAI commitments, with the capital expenditure recorded on Oracle's balance sheet.
- Leases also play a significant role, as seen in SB Energy's $5.5 billion investment in OpenAI warrants in exchange for a 20-year lease, transforming a capital commitment into an operating expense.
These financing structures enable OpenAI to scale without shouldering the full financial burden itself. However, they also highlight the complex web of obligations created in the buildout process.
Free cash flow, while a crucial metric, does not reflect these external financing arrangements and obligations, leading to apparent contradictions between rising spending and decreasing burn.