OpenAI slashes AI spending forecast to $600 B and unveils GPT‑5.3 Codex, its new general

OpenAI announced that its projected AI‑spending target for the next decade has been cut in half, now aiming for $600 billion by 2030, down from the $1.4 trillion forecast disclosed in its 2024 investor briefings. The revision, reported by CNBC, reflects a strategic shift toward tighter cost controls as the company scales its compute infrastructure and accelerates the rollout of its next‑generation models (CNBC, Feb 20, 2026). Executives cited “maturing market dynamics” and “more efficient hardware utilization” as the primary drivers of the new outlook, suggesting that OpenAI expects to achieve comparable performance gains with a leaner spend profile.

At the same time, OpenAI unveiled GPT‑5.3 Codex, a general‑purpose work agent that supersedes the earlier GPT‑5.2 and GPT‑5.2‑Codex variants. According to Tech Croc, the model is no longer a narrow “coding assistant” but a full‑stack software‑engineering partner capable of handling the entire development lifecycle—from generating code and updating Jira tickets to reviewing architectural decisions and producing end‑to‑end documentation (Tech Croc, Feb 24, 2026). The company emphasizes that GPT‑5.3 Codex merges deep reasoning abilities with specialized coding expertise in a single architecture, allowing it to retain context across multi‑step tasks and operate natively within graphical user interfaces.

Benchmark data released alongside the launch shows GPT‑5.3 Codex delivering a 35 percent reduction in task completion time for typical software‑engineering workflows compared with its predecessor, while maintaining a 92 percent accuracy rate on code‑generation benchmarks (Tech Croc, Feb 24, 2026). OpenAI also highlighted new security layers that sandbox the model’s execution environment, a response to growing enterprise concerns about code‑injection risks. The firm claims the agent can process up to 1.2 TB of contextual data per session, enabling it to reference extensive codebases and design documents without loss of fidelity.

The spending cut and the Codex rollout arrive amid mounting pressure on OpenAI’s “hype machine,” as Reuters Breakingviews notes that the firm’s rapid expansion has strained corporate relationships and heightened scrutiny over its revenue‑sharing arrangements with Microsoft and other partners (Reuters Breakingviews, Oct 28, 2024). Analysts cited in the Reuters piece warn that while OpenAI’s technology remains a market leader, the company must balance aggressive product launches with sustainable financial discipline to preserve investor confidence.

OpenAI’s leadership framed the dual announcements as complementary: the revised spend target frees capital for strategic investments in next‑generation agents, while GPT‑5.3 Codex is positioned to unlock new enterprise revenue streams by automating high‑value software‑development tasks. If the model’s productivity gains materialize at scale, the company could offset the reduced capital outlay with higher per‑customer spend, a hypothesis that will be tested as early adopters integrate Codex into their development pipelines over the coming months.