Top 10 AI News — October 9, 2026

AI stocks fell on Thursday after OpenAI told investors its revenue run rate was $50 billion, well short of the $68 billion figure that had been circulating. Products kept shipping regardless: Google launched a unified Gemini agent for businesses, OpenAI put GPT-6 in front of free users, and Anthropic cut prices on its smallest model.

1. OpenAI's $50 billion run rate sends AI stocks lower

CNBC confirmed that OpenAI told investors it reached roughly $50 billion in annualized revenue at the end of September. That is well below the $68 billion figure widely reported late last month. According to a source quoted by CNBC, the higher number counted gross revenue from OpenAI's partners, which made it easier to compare with Anthropic.

The market reacted the same day. The Nasdaq Composite fell 1.25% and the S&P 500 dropped 0.47%. Nvidia, Oracle, CoreWeave and Micron all fell, because each of them sells capacity into OpenAI's spending plans.

On Friday Bloomberg reported that OpenAI expects to reach or exceed $70 billion in annualized revenue by year end. A run rate projects the current sales pace over a full year; it is not booked revenue. So the open question is which accounting method the market will use from here, not whether OpenAI is growing.

2. Google launches one Gemini agent for businesses

At a Google Cloud event on Thursday, Google showed a unified Gemini agent that takes objectives instead of step-by-step instructions. It plans the work, loads skills, delegates to subagents and connects to internal systems. Connectors cover Workspace, Microsoft 365, Slack, Jira, Confluence, Git, BigQuery, Databricks, Postgres, Snowflake and any MCP server.

Each agent gets its own Workspace account, email address and audit trail, so its actions are logged to the agent and not to the employee who asked. By default it picks a model for each task. Users can override that choice, and the third-party options start with Anthropic's Claude. Google says nearly 90% of the Fortune 100 already use Gemini Enterprise. On, Shopify and PayPal tested the agent early.

Sundar Pichai said businesses get it before consumers so Google can work through "harder problems around security, scale, and performance" first. Google has not announced pricing or a general availability date. It did list spending caps and smart routing as cost controls, which suggests what enterprise buyers asked for first.

3. GPT-6 and Intelligent UI reach every ChatGPT tier

OpenAI began rolling out GPT-6 with Intelligent UI to Plus, Pro, Business and Enterprise users on October 7, and to Free and Go users on October 8. Paid tiers run GPT-6 Sol and the free tiers run GPT-6 Luna. Answers can now include interactive charts, forms, buttons and calculators alongside text, and plain text remains the default when it is clearer.

The change covers ChatGPT's chat experience only, not the models behind Work or Codex. With it, OpenAI's newest model family is in front of its entire consumer base, and ChatGPT has moved from returning text to returning small working interfaces.

4. Anthropic launches Claude Haiku 5.5 at $0.10 per million input tokens

Anthropic released Claude Haiku 5.5 on October 7. For prompts up to 100,000 tokens it costs $0.10 per million input tokens and $0.50 per million output tokens, which works out to about 90% below Haiku 4.5. Above 100,000 tokens the price rises to $0.50 and $2.50. It is the first Haiku with an adjustable effort setting.

Anthropic puts the average saving against Haiku 4.5 at about 75%, below the sticker math. Part of the gap comes from a new tokenizer that uses slightly more tokens per task. The model is available on Anthropic's platform, AWS, Google Cloud and Azure. Classification, summarization and support teams running high volumes now pay a fraction of what they did for this tier.

5. Eleven companies pledge $2.4 billion to the Genesis Mission

At a White House summit on October 8, eleven companies pledged $2.4 billion in resources to the federal Genesis Mission Consortium. Nvidia led with $1 billion, AMD pledged $500 million, OpenAI $200 million, and Anthropic and Google $150 million each. AWS, Crusoe, Micron and Armada pledged $50 million apiece.

Almost none of it is cash. The pledges come as compute capacity, cloud and token credits, discounted model access and engineering support, spread over several years. Anthropic's three-year commitment provides Claude, Claude Code and API credits to several hundred research projects. Federal agencies had already committed more than $5 billion to the program by July.

6. Claude ports the TypeScript 7 compiler to Rust in two weeks

On October 7 Theo Browne released tsc-rs, an MIT-licensed Rust port of Microsoft's Go-based TypeScript 7 compiler, type checker and language server. Claude Opus 5.5 wrote the port. The README says all 181,711 ported Go tests pass. Earlier agent attempts with OpenAI models reportedly spent more than $400,000 in API-priced tokens and stalled around 84% compatibility. The Opus run cost roughly $24,000.

Read the claim carefully. Browne said he did not read the code. The port inherits Microsoft's design and test suite instead of creating new ones, and reports of a speedup conflict: Browne himself says it is faster only when Effect type checks are in use. What the project does show is an agent completing a large, test-gated translation from start to finish, and that is the task where coding agents are most useful today.

7. AP-NORC: 64% of Americans say AI is moving too fast

An AP-NORC poll released October 8 found that 64% of US adults think AI is developing too fast. Another 27% said the pace is about right and 8% said too slow. About eight in ten said it is extremely or very important for the government to keep AI under human control. A similar share want the government to prioritize protecting workers.

The concern crosses party lines. Only 31% approve of President Trump's handling of AI, and 39% say neither party can be trusted on it. NORC surveyed 2,140 adults from September 24 to 28, with a margin of error of ±2.9 points. Lawmakers in both parties now face voters who put protecting workers ahead of speed.

8. Manus raises more than $500 million after the Meta deal is unwound

Butterfly Effect, the parent company of the agent startup Manus, raised more than $500 million in a round co-led by Boyu Capital and IDG Capital. Tencent, HSG and ZhenFund also participated. Meta had agreed to acquire Manus for more than $2 billion, but in April Chinese regulators ordered the deal unwound amid scrutiny of US investment in Chinese AI startups.

The Information reported that Manus's annualized revenue reached about $500 million by June, up from about $100 million when Meta agreed to buy it. Reuters reports that a Hong Kong IPO is not expected before 2027. Blocking the sale did not hurt Manus's fundraising. It did keep a fast-growing agent company out of a US platform.

9. Accenture and Dell form a business group for private AI

Accenture and Dell announced the Accenture Dell Business Group on October 8, extending a partnership of more than 20 years. The group will build private AI stacks, which it calls AI factories, for agentic workloads and inference. The target customers are regulated industries, including financial services, life sciences, manufacturing and high tech. Accenture plans to train more than 3,000 practitioners on Dell's portfolio.

The announcement fits a broader pattern: enterprises that cannot send sensitive data to a public cloud are building on-premises AI capacity instead, and integrators are organizing around that work. Neither company disclosed an investment figure.

10. AI classroom monitoring planned for a Gaza school network

The Associated Press reported on October 9 that Gaza Children Village plans to record classroom conversations at its Academies of Hope schools. An AI system would flag speech it labels "antisemitic," "inciting" or "hateful" and alert the charity's founder, Duke neurosurgeon David Hasan, who would then contact the teacher. Recordings would be kept for 72 hours, and the system would run in dozens of classrooms serving thousands of children.

The pushback came quickly, including from inside the charity. Board member Bishara Bahbah said cameras "to silence students and teachers from being able to have a free exchange in a classroom are not acceptable." An aid partner said it might reconsider the relationship. The plan does not say when monitoring starts or which language will be flagged, and those are the details that would decide how much harm it does.

What to watch

The OpenAI revenue gap will hang over every AI infrastructure financing until the company settles on one reporting method. Watch for pricing on Google's Gemini agent, the date Mistral publishes Large 4's open weights (covered in our October 7 brief), and whether price cuts like Haiku 5.5's force competitors to cut prices on their small models.