OpenAI investor memo touts computing advantage over Anthropic

OpenAI sent a memo to investors this week arguing that its computing infrastructure gives it a structural lead over Anthropic, characterizing its rival as compute-constrained at a moment when both companies are moving toward IPOs.

According to the memo, which was reported by Bloomberg and CNBC, OpenAI put its own 2025 capacity at 1.9 gigawatts — a figure it said was three times what it had the year before — and placed Anthropic’s equivalent figure at 1.4 gigawatts. OpenAI told investors it foresees its own footprint climbing into the “low-double-digit range” of gigawatts within a year and hitting 30 gigawatts by 2030, and that Anthropic would top out somewhere between seven and eight gigawatts before the close of 2027.

“Even at the high end of that range, our ramp is materially ahead and widening,” OpenAI wrote in the memo.

The memo cast heavy infrastructure investment as self-reinforcing: cheaper costs follow from stronger models and hardware, and those savings in turn fund the product improvements that attract more customers and revenue. “Each new generation of infrastructure lets us train more capable models, making every token more intelligent than the one before,” the memo said.

OpenAI also pointed to comments by Anthropic CEO Dario Amodei, who last year described his firm’s approach as deliberately conservative, saying some competitors were “YOLO-ing.” The memo characterized that caution as a miscalculation. “In hindsight, that caution looks less like discipline and more like underestimating how fast demand would arrive,” OpenAI wrote.

To support its case, OpenAI pointed investors to a Stratechery analysis by Ben Thompson, which raised the possibility that capacity shortfalls had pushed Anthropic into rolling out Mythos exclusively through a handful of partner companies rather than the general public.

When both outlets sought comment, Anthropic pointed them to remarks CFO Krishna Rao had made in connection with the recently announced Google $GOOGL and Broadcom $AVGO partnership. “We are making our most significant compute commitment to date to keep pace with this unprecedented growth,” Rao said. The agreement, announced earlier this week, is structured to deliver approximately 3.5 gigawatts of computing capacity to Anthropic starting in 2027.

Bloomberg has reported that OpenAI’s capital plan envisions around $600 billion flowing toward chips and data centers through the end of the decade, a buildout the company has partly financed through a $122 billion fundraise. Profitability remains a distant prospect under that spending trajectory.