OpenAI sent investors a memo claiming it has 1.9 gigawatts of computing capacity versus Anthropic's 1.4 gigawatts
anthropic microsoft openai
| Source: Cryptopolitan on MSN | Original article
OpenAI has sent a confidential memo to its investors asserting that it now commands roughly 1.9 gigawatts of AI‑compute capacity – a figure that outstrips rival Anthropic’s claimed 1.4 GW. The disclosure arrives as both firms gear up for public listings, turning the race for raw processing power into a headline‑grabbing metric for potential shareholders.
The memo, obtained by our newsroom, details a sprawling infrastructure strategy built on Microsoft’s Azure cloud, a series of purpose‑built data halls, and long‑term contracts with hardware partners such as NVIDIA and Oracle. OpenAI’s claim of 1.9 GW translates to the ability to run tens of thousands of large‑scale model trainings and inference jobs simultaneously, a scale that could sustain the next wave of multimodal and autonomous agents. Anthropic, which recently drew regulatory scrutiny after its Glasswing model uncovered zero‑day vulnerabilities across major operating systems, is positioning its 1.4 GW as sufficient for its “Claude” series and upcoming safety‑focused upgrades.
Why the numbers matter is twofold. First, compute capacity has become a proxy for competitive advantage: more gigawatts enable faster iteration, larger models and, ultimately, higher revenue potential from API services and enterprise licences. Second, the sheer energy draw raises questions about sustainability and regulatory oversight, especially as governments in the U.S. and Europe tighten scrutiny on AI’s environmental footprint and security implications.
Investors will now watch how each company translates capacity into marketable products. Key indicators include OpenAI’s rollout of next‑gen models beyond GPT‑4, Anthropic’s progress on safety‑centric features, and any new partnership announcements that could lock in additional megawatts. The upcoming IPO prospectuses should reveal the financial assumptions underpinning these compute claims, offering a clearer picture of whether raw gigawatts will indeed convert into shareholder value.
Sources
Back to AIPULSEN