OpenAI Tells Investors Revenue Is $20 Billion Lower Than Reported Figure

OpenAI's Revenue Comes In $20 Billion Below Hype

OpenAI has told investors its annualized revenue is approaching $50 billion, well short of the $70 billion figure that had been widely reported just over a week earlier. The gap rattled tech stocks, with the Nasdaq and chip-linked names like Nvidia, Oracle and Micron all falling on the news. Sources: Financial Times - AI, TechCrunch - AI

The discrepancy traces back to an attempt by OpenAI's own investors to make the company's numbers comparable to Anthropic, which reports revenue that includes sales through cloud partners such as AWS and Google Cloud. OpenAI's own accounting excludes that cloud-partner revenue, so the $70 billion estimate was never truly apples to apples. Sources: The Guardian - AI, Financial Times - AI

The revised figure lands as questions are raised over OpenAI’s growth and demand. Sam Altman said last month the company will not float on the stock market this year, citing safety concerns. Fifty billion dollars in annualized revenue is still a massive number, just not the one some investors had been led to expect. Sources: The Guardian - AI, TechCrunch - AI

Anthropic's Cheap New Haiku Undercuts Rivals on Price

Anthropic has released Claude Haiku 5.5, its first update to the small-model line in about a year, calling it the cheapest, fastest and most capable small model the company has ever shipped. Anthropic says it costs roughly 75 percent less to run on average than Haiku 4.5, with prices dropping by up to 90 percent for prompts under 100,000 tokens, a bracket the company says covers about 90 percent of past Haiku usage. Sources: The Decoder, Latent Space

Independent benchmarking from Artificial Analysis backs up the performance claims: Haiku 5.5 scores 43 on the Intelligence Index, edging out GLM-5.3 Flash and GPT-6 Luna, and its score on Humanity's Last Exam jumped from 10.2 percent to 45.9 percent without tools. Source: The Decoder

The catch is token efficiency. At its highest effort setting Haiku 5.5 burns through about 162,000 output tokens per task, roughly three times what GPT-6 Luna needs for comparable work, meaning the advertised sticker price understates the real cost of heavier jobs. Source: The Decoder

Manus Raises Over $500 Million After Meta Deal Collapsed

Chinese AI startup Manus has raised more than $500 million, its first funding round since Meta's $2 billion acquisition of the company fell apart. Boyu Capital and IDG Capital led the round, with existing backers Tencent and HSG, formerly Sequoia China, also participating. Source: TechCrunch - AI

Chinese authorities ordered the Meta deal unwound back in April over concerns about losing AI talent to the West, and Manus resumed independent operations in August. The company has not disclosed a valuation, though it was reportedly in talks last month for the same amount at a $4 billion valuation. Source: TechCrunch - AI

The funding round was led by Boyu Capital and IDG Capital, with existing shareholders including Tencent and HSG also participating. Source: TechCrunch - AI

Google Gives Gemini Its Own Inbox and a Job to Do

Google Cloud has introduced a unified agentic version of Gemini aimed first at businesses, letting the AI plan work rather than simply follow instructions. CEO Thomas Kurian said the new agent can be given “objectives, not just instructions,” and it can use tools and connect to internal systems like Slack, Jira and BigQuery. Source: TechCrunch - AI

The agent gets its own Workspace account complete with an email address, tracks colleagues' time zones and calendars, and logs an audit trail under its own name instead of a human's. Google says Gemini already has more than a billion monthly active users and that nearly 90 percent of Fortune 100 companies use Gemini Enterprise, giving the rollout a large built-in audience. Source: TechCrunch - AI

Early testers include sportswear brand On, Shopify and PayPal, and users will be able to swap in third-party models starting with Anthropic’s Claude. Source: TechCrunch - AI

Anthropic Bans Cruelty Toward Its Own Chatbot

Anthropic has updated its usage policy for the first time in over a year, adding a ban on what it calls sustained and needless abusive or cruel behavior toward Claude. The company says the rule is meant only for extreme cases with no discernible purpose, and that ending the conversation remains the primary way it will be enforced. Sources: The Verge - AI, Anthropic news

The update also sharpens language around weapons, now explicitly covering software and components that make weapons function, including arming drones and other autonomous vehicles. It additionally bars using Claude to decide or recommend who should be investigated, arrested or charged. Sources: The Verge - AI, Anthropic news

Anthropic says the changes respond to misuse patterns it has actually observed, including attempts to build weapons-control software and surveillance tools used to track political dissidents. Sources: The Verge - AI, Anthropic news

FTC Refunds Millions Misled by Cleo AI Cash Advances

The FTC is returning more than $15.8 million to consumers misled by cash advance promises from the app company Cleo AI. The agency's March 2025 complaint alleged Cleo advertised access to hundreds of dollars in cash advances when almost no one received close to that amount, charged hidden fees for advances that often arrived a day late, and made subscriptions difficult to cancel. Source: FTC press releases

The FTC plans to begin sending payments on October 27, 2026, to 2,124,796 Cleo AI customers who paid for eligible instant cash advances. Source: FTC press releases