OpenAI’s Enterprise Pivot: Codex Surge Powers $1B API Boom Amid Anthropic Rivalry

OpenAI is ramping up its push into corporate accounts with new tools to quantify AI’s return on investment and a revamped sales approach, as it seeks to claw back ground from rival Anthropic in the battle for business dollars. CEO Sam Altman revealed that the company’s API operations added more than $1 billion in annual recurring revenue in a single month, highlighting the shift toward enterprise as the core growth driver. “People think of us mostly as ChatGPT, but the API team is doing amazing work!” Altman posted on X.

This surge comes as OpenAI prepares updates to its Codex coding model, which will for the first time reach the company’s “High” cybersecurity risk level, enabling advanced capabilities like automating attacks on protected targets or detecting vulnerabilities. Altman announced new features shipping the following week, with initial product restrictions to curb cybercrime use, evolving toward bolstering defenses. Enterprises will need stringent safeguards for deployment, per OpenAI’s guidelines reported by The Decoder.

The API boom underscores OpenAI’s diversification beyond consumer products. Startups like Clay, which automates sales prospecting using OpenAI’s models, hit $100 million in ARR in December, processing trillions of tokens and becoming one of OpenAI’s largest API customers. Founder Kareem Amin credits the partnership, including Slack channels and weekly meetings for API tweaks, as detailed by Inc. Other adopters include Perplexity for AI search and Harvey for legal tech, fueling the revenue spike noted by Business Insider.

API Explosion Fuels Revenue Records

OpenAI’s overall annualized revenue crossed $20 billion in 2025, up from $6 billion the prior year, with API demand—particularly Codex—driving a twentyfold increase in code generation via the platform. Developers are embedding models into IDEs and CI/CD pipelines, as shared by AI specialist Amelia Edwards on X. Codex usage grew 20x in six months after shifting to local editor integration, per OpenAI’s Codex product lead in a podcast with Lenny Rachitsky.

At Davos, OpenAI CFO Sarah Friar emphasized enterprise now comprises 40% of business, aiming for 50% by year-end, positioning large firms as the funding source for massive compute needs—from 0.2 GW to 1.9 GW in 2025. This aligns with partnerships like Salesforce embedding GPT-5 and Codex into Agentforce for regulated sectors, reported by Reuters.

Anthropic’s Enterprise Edge Challenges OpenAI

Anthropic has surged ahead in enterprise LLM spend, capturing 40% market share in 2025 versus OpenAI’s 27%, down from 50% in 2023, driven by Claude’s dominance in coding at 54% share. Menlo Ventures’ report highlights code as AI’s killer app, with 50% of developers using tools daily. Anthropic derives 80% of revenue from business, projecting $10 billion in 2025 and break-even by 2028, per Menlo Ventures and CNBC.

OpenAI counters with Codex advancements, including GPT-5.2-Codex for agentic engineering, cloud sandboxes, and multi-day autonomous sessions via “compaction.” Benchmarks show tight races: Claude Opus 4.5 at 80.9% on SWE-Bench Verified versus OpenAI’s 80.0%. Developers debate preferences—Claude for refactoring, Codex for speed—on forums like Reddit, while Cisco deploys Codex for multi-repo migrations in compliance settings.

In response, OpenAI is developing enterprise offerings to overhaul systems, including financial ROI calculators and sales force tweaks to lure Anthropic clients, as exclusively reported by The Information. This includes potential revenue shares from AI-aided discoveries, signaling a deeper embed in operations.

Codex Evolution Redefines Developer Workflows

Codex, reborn from its 2021 origins, now powers feats like building Sora’s Android app—#1 on app stores—in weeks with few engineers. It generates plans, implements cross-platform features, and handles gnarly bugs. OpenAI’s internal guide details uses beyond code: note organization, prototyping, task queues. Best practices include structured prompts like GitHub issues and context files for business logic, from OpenAI’s PDF.

The shift alters roles: designers code prototypes, engineers review AI output. Bottlenecks move to human typing speed and validation, with models running 60+ hours autonomously. Embirico notes writing code may become how all AI tasks occur, embedding coding in every agent.

Competitive heat intensifies with Anthropic’s Claude Code CLI for local workflows and agentic search across codebases. Paul Kedrosky’s analysis flags Claude’s edge prompting OpenAI’s “code red,” though recent Reddit threads claim Codex 5.2 outperforms on debugging, per Paul Kedrosky.

Strategic Bets on Enterprise Scale

OpenAI’s enterprise focus addresses $1.4 trillion infrastructure pledges amid $17 billion annual burn projections. Friar ties revenue to compute scaling, while Anthropic faces margin squeezes from inference costs, slashing 2025 forecasts. Salesforce’s dual partnerships give Claude access in finance and healthcare via secure clouds, per Constellation Research.

At Davos, Anthropic CEO Dario Amodei stressed Fortune 500 scaling, while OpenAI’s Jason Kwon focused on deepening ties. Both eye healthcare, with Anthropic’s HIPAA-ready Claude rattling software stocks, as RBC analysts noted in Business Insider.

OpenAI’s playbook—API scale, Codex potency, ROI tools—positions it to reclaim enterprise primacy, but Anthropic’s safety-aligned, coding-strong models demand vigilance in this high-stakes rivalry.