Entelligence AI’s investigation of 2,444 companies, released on May 26, shows that for every $1 spent on AI engineering, $0.82 is lost: $0.44 is used to fix errors, $0.27 to rewrite AI-generated code, and $0.11 is lost due to review and merge delays—only about $0.18 becomes a usable production system.
Confirmed AI Engineering Loss Data: Entelligence AI and Lightrun
（Source: Entelligence AI）
According to Entelligence AI’s investigation of 2,444 companies, for every $1 of AI engineering spend: $0.44 is used to fix AI code vulnerabilities, $0.27 is used to rewrite AI-generated code, $0.11 is lost due to review and merge delays, for a total loss of $0.82.
Lightrun’s “2026 State of AI-Driven Engineering Report” also notes that 43% of AI-generated code still requires manual debugging in production environments even after passing quality checks; none of the surveyed engineering leaders expressed complete confidence in AI outputs that have already been deployed.
Oracle’s Confirmed Financial Situation
Oracle’s total accumulated liabilities are approximately $108 billion, and it plans to raise another $50 billion in 2026 through debt and equity financing to build AI data centers, with free cash flow at about negative $13 billion.
Oracle’s remaining performance obligations (RPO) backlog orders total $553 billion, of which more than $300 billion are related to OpenAI; OpenAI recorded an approximately $14 billion loss last year. Oracle’s next earnings report is planned for release on June 16, 2026.
OKX’s Confirmed Policy Changes for AI Talent
In a public statement, OKX CEO Xu Shijie said that AI agents can speed up execution while revealing employees who rely on impression management rather than actual results. He said directly: “Fundamentally, it’s not artificial intelligence that changes the phenomenon of layoffs—rather, in the age of artificial intelligence, it fundamentally changes the demand for talent.” OKX has so far tied employee evaluations to AI proficiency.
FAQ
How is Entelligence AI’s 82% loss data calculated?
Entelligence AI’s investigation of 2,444 companies shows that for every $1 of AI engineering spend, losses sum to $0.82: $0.44 for fixing vulnerabilities, $0.27 for rewriting code, and $0.11 for review delays. This means that only about 18% of AI engineering spending directly turns into usable production systems.
What does the meaning of Oracle’s RPO backlog orders and the link to OpenAI?
In Oracle’s $553 billion RPO backlog orders, more than $300 billion are related to OpenAI. OpenAI’s loss of about $14 billion last year is seen by the market as one of the risk factors for Oracle’s AI business due to this concentration. Oracle’s next earnings report is planned for release on June 16.
What are the specific details of OKX tying employee evaluations to AI proficiency?
OKX CEO Xu Shijie confirmed that OKX has directly linked employee evaluations to AI proficiency, but the specific evaluation criteria, implementation timeline, and performance metrics were not disclosed in detail in this announcement.
