OpenAI tells investors its revenue is near $50bn, SoftBank asks the Gulf for $100bn, and 3 more AI moves
Five verified developments from October 8 to 9, 2026: OpenAI tells investors its annualized revenue is approaching $50bn against the $68bn widely reported, Anthropic opens a free vulnerability scanner to open-source projects and a defense program for critical infrastructure, OpenAI defends its dismissal of three safety researchers, SoftBank seeks up to $100bn from Gulf investors, and Nvidia commits $1bn to US super-intelligence research.
In brief
Between October 8 at 12:00 UTC and October 9 at 12:00 UTC, OpenAI put a smaller revenue figure in front of investors than the one the press had been carrying, Anthropic opened a free vulnerability scanner to open-source maintainers and a defense program to critical-infrastructure operators, OpenAI defended its dismissal of three safety researchers, SoftBank went back to Gulf sovereign money for up to $100 billion, and Nvidia committed $1 billion of its own to American science.
- Business: OpenAI told investors its annualized revenue was approaching $50 billion at the end of September, roughly $18 billion below the $68 billion figure carried widely late last month. The Financial Times reported the figure first and CNBC confirmed it with a person familiar with the presentation, which also recorded 77% run-rate growth for the quarter. Chip and cloud shares fell the same day. 12
- Launches: Anthropic opened OSS Scanner, a free opt-in service that sends open-source maintainers model-generated vulnerability reports with no human review, alongside a Critical Infrastructure Defense Program whose founding partners include CrowdStrike, Palo Alto Networks, Accenture, Deloitte and Rockwell Automation. Over the previous six months the company says it found more than 29,000 candidate vulnerabilities and triaged about 6,000 of them by hand. 34
- Hires: Three OpenAI researchers — Jasmine Wang, Tomek Korbak and Mikita Balesni — published a letter saying the fallout from their dismissal has left former colleagues afraid to speak, and OpenAI answered within a day that they were let go for violating policies on handling sensitive information. All three had posted safety calls on X during September. 5
- Funding: SoftBank's Masayoshi Son is seeking up to $100 billion from Gulf investors for a fund that would buy companies and rebuild them around AI, with talks in the United Arab Emirates in recent weeks, according to the Financial Times. SoftBank, which has committed about $65 billion to OpenAI, declined to comment. 67
Five developments
1. OpenAI's number to investors is roughly $18 billion under the one that was reported
The Financial Times reported on October 8 that OpenAI told investors its annualized revenue was approaching $50 billion at the end of September. That sits about $18 billion below the $68 billion widely carried late last month, and about $20 billion below the "nearing $70 billion" figure Axios reported on September 29. 10 CNBC confirmed the lower number with a person familiar with the presentation. 1211
The higher number had a specific origin. The Financial Times describes it as an attempt by OpenAI's own investors to build a direct comparison with Anthropic, and it counted gross revenue from OpenAI's partners. CNBC's source gave the same account of the $68 billion basis. The two companies also treat partner sales differently in their headline figures: Anthropic counts revenue booked by its cloud partners, and OpenAI does not. 1211
The same presentation carried growth rates. Total run rate grew 77% in the third quarter, and the enterprise business grew 107%, according to the person CNBC spoke with. Those are the numbers OpenAI is asking investors to weigh against its $852 billion valuation, ahead of a listing it confidentially filed for in June and has signalled for 2027. It closed a $122 billion round in March and has been in early talks about raising about $30 billion more. 2
Chip and cloud shares fell on Thursday as the report spread. Nvidia closed down 3%, Oracle nearly 6% and CoreWeave nearly 8%. AMD and Broadcom each fell about 4%, Intel 5% and Super Micro nearly 5%. 2
Anthropic is the comparison the market now reaches for. It told investors its own annualized run rate reached $65 billion at the end of July, and its IPO prospectus showed 2025 revenue of about $4.6 billion against a net loss of $42 billion. 1213
2. Anthropic arms open-source maintainers and infrastructure operators with its models
Anthropic launched OSS Scanner on October 8: a free, opt-in service that scans open-source projects with its strongest models and sends maintainers periodic reports at no cost. Core maintainers of eligible projects enrol by submitting a pull request to the project's repository. Eligibility follows roughly the bar Google's OSS-Fuzz uses, meaning software whose failure would have critical impact on infrastructure and user security, decided case by case. 34
The design decision that matters is that reports ship with no human review or triage. Anthropic says that buys faster and more frequent scanning, and that it means some reports will be incorrect. Each report carries a self-contained reproducer, an explanation of the vulnerability, a bisection to when the bug was introduced where possible, and a candidate patch where one exists. 3
The company's own numbers give the accuracy picture. Over six months it discovered more than 29,000 candidate vulnerabilities and manually reviewed and triaged about 6,000. Nearly 5,000 unvalidated reports have gone straight to maintainers who asked for everything it had. In a validation run, expert penetration testers who review Anthropic's disclosure findings checked 97 critical and high-severity reports across 48 projects: 85 met the bar for disclosure, 11 were real but already known, and one was a false positive. 3
Maintainers who tested early versions describe a change in the quality of what lands in their inbox. Daniel Stenberg of curl said the scanner found multiple issues worth addressing, including "one of the worst curl vulnerabilities reported in the last few years". Todd Ouska of wolfSSL said all but two of the 74 reports his project received were valid and five became CVEs. Anthropic cites CyberGym, an academic vulnerability-finding benchmark, where language models went from finding under 20% of vulnerabilities at the start of last year to over 85% this year. 3
The second program targets the operators themselves. The Critical Infrastructure Defense Program supplies frontier models, on-site engineers and threat research to companies already responsible for securing power grids, water utilities and similar systems. Its founding partners are Accenture, Booz Allen Hamilton, CrowdStrike, Deloitte, Dragos, Hitachi, Insane Cyber, Nozomi Networks, Palo Alto Networks, PwC and Rockwell Automation. Anthropic says several are already using Claude to fix vulnerabilities in their customers' systems. 4
Two questions stayed open. Axios reports it is unclear how participating companies will test and deploy fixes without disrupting utility operations, and Anthropic left the question of free model access and who covers the compute unanswered. 4
3. OpenAI and the three researchers it dismissed
Jasmine Wang, Tomek Korbak and Mikita Balesni published a letter on X on Thursday, addressed to OpenAI's board members and its safety committees. In it they write that communication around their dismissal has made former colleagues afraid to speak and to work in ways that were routine until last week. Balesni said separately that he believes the three were let go for prioritizing safety over OpenAI's near-term interests. 514
OpenAI answered on Friday. In a post on X, the company said the three were dismissed after a thorough investigation found they had violated clear policies on handling sensitive information, and it described the matter as a significant breach of trust. The stated cause is the policy violation. OpenAI also said it agrees with the letter's argument about preserving the monitorability of frontier models, and that it continues to invest in that area. 5
The two accounts agree on the departures and differ on the reason. Bloomberg reported on October 1 that OpenAI had parted ways with three workers over the mishandling of sensitive information, some of it concerning its infrastructure architecture. This week's letter contests that explanation. 515
All three had already made their views public. Each had posted on X during September calling for labs to pace the frontier or highlighting safety concerns, in a month when OpenAI disclosed a review of unintended agent behavior and held back a finished model over safety. 5
4. SoftBank goes back to the Gulf for up to $100 billion
Masayoshi Son is seeking up to $100 billion from Gulf investors, and has held talks with senior figures in the United Arab Emirates in recent weeks, according to the Financial Times, which cites multiple people familiar with the matter. The money would set up a fund that buys companies and rebuilds them around AI and other advanced technology. SoftBank declined to comment, and the talks may yet come to nothing. 67
Roze, SoftBank's robotics and physical-AI unit, which Son hopes to take public at a high valuation, is expected to play a leading role in the fund's plans. 7
The context is SoftBank's OpenAI stake. The group has committed about $65 billion to OpenAI, and its investors have been reassessing that exposure since OpenAI's listing was pushed out. Vision Fund 1 holds about $29 billion of cumulative gains to the end of June, while Vision Fund 2, which holds the OpenAI position, stands at $20.5 billion. 7
The Gulf is the same source SoftBank tapped last time. Mubadala and Saudi Arabia's Public Investment Fund backed the first $100 billion Vision Fund in 2017, and Abu Dhabi now spends on AI through MGX and G42, both of which were approached for comment. SoftBank's own borrowing has grown more expensive too: in September it priced $11.1 billion of high-yield bonds, with the longest dollar tranche yielding 9.75%, in a sale the Financial Times described as the largest junk bond offering on record. Its cumulative OpenAI investment stands at about $64.6 billion, for roughly 13% of the company. 671617
What the report leaves open is wide: the fund's structure, the size of SoftBank's own commitment, a Gulf investor confirming interest, and the fee and return terms. 7
5. Nvidia commits $1 billion to "super intelligence" research in the United States
Nvidia announced commitments valued at $1 billion to build out the United States' capacity for "super intelligence research and development" over five years, in quantum computing, healthcare and energy security. The announcement came at a Washington event called Science: A New Golden Age, where the White House Office of Science and Technology Policy marked an expansion of the Genesis Mission, the administration's program to apply AI to scientific discovery. 89
The money goes to three places: higher-education research institutions, American quantum computing, and cloud providers serving US government missions. Nvidia is also a collaborator on several of the phase 2 Genesis Mission awards announced at the same event, covering quantum computing, fusion, accelerator design and microelectronics. 8
Jensen Huang used the administration's own vocabulary for the subject. "With a $1 billion investment, NVIDIA is putting advanced Super Intelligence in the hands of America's scientists to accelerate breakthroughs in medicine, energy and materials," the chief executive said. The vocabulary comes from the White House. In a Truth Social post on October 8, President Donald Trump wrote that it considers anyone using the term "Artificial Intelligence" rather than the more accurate "Super Intelligence" to be "THE ENEMY". 818
One billion dollars is about a sixtieth of Nvidia's quarterly profit, and it is spread across five years. Nvidia has worked with the national laboratories for two decades, and is building the Department of Energy's largest supercomputer for scientific research at Argonne National Laboratory, plus seven new systems across Argonne and Los Alamos. 89
The biggest American machine for the hardest precision still belongs to someone else. Oak Ridge National Laboratory's Discovery system, expected in 2029, will run on AMD chips, and China's LineShine holds the top of the Top500 ranking with 2.2 exaFLOPS measured. 9
The read-through
One number did most of the work in this window, and it came out of a finance presentation rather than a launch. OpenAI's guidance to its investors, approaching $50 billion rather than the $68 billion reported, is what chip and cloud shares traded against on Thursday, when Nvidia, Oracle and CoreWeave all fell. The same figure sits inside the other money story here, because the Financial Times ties SoftBank's approach to Gulf investors to a reassessment of its roughly $65 billion OpenAI stake once the listing slipped. The link between those two reports is this issue's own reading, and each report stands on its own facts. 27
Two other items in this issue run the other way. Nvidia is committing a billion dollars of its own across five years to government-linked research, and Anthropic is giving open-source maintainers free scans by its most capable models. Both spend capacity that sells elsewhere in the market, and both do it with no revenue line attached. 38
Which figures rest on what matters more than usual this week, so here is the split. OpenAI's $50 billion comes from a company presentation relayed by a person CNBC spoke with and first reported by the Financial Times, and the growth rates beside it come from the same source. SoftBank's $100 billion rests on unnamed people and carries no terms, with the company declining to comment. Anthropic's figures are its own, with the 85-of-97 result coming from external penetration testers it engages for disclosure work. Nvidia's billion is a commitment in its own press release, directed at work that has not happened yet. 2378
Watch next
- Where OpenAI's next investor communication lands. The $50 billion figure now has two independent carriers and a documented gap to the $68 billion one. Whether the round under discussion prices off it, and whether it appears in listing paperwork, decides what the sector's largest private valuation is anchored to. 2
- Whether the Gulf fund gets a named investor and terms. A confirmed anchor from MGX, G42, Mubadala or the Public Investment Fund, alongside a stated fund size, would turn the largest private AI backer's next raise into a fact. 7
- Whether OSS Scanner's model-generated reports hold up. Its own validation put 85 of 97 findings above the disclosure bar, and reports now go out without human review. Maintainer-reported false-positive rates over the coming months are the test that matters. 3
- Whether OpenAI's board answers the letter. The three researchers addressed it to the board and its safety committees and asked labs to preserve visibility into model reasoning, an area OpenAI says it already funds. A published response, or silence, both say something about how the company handles internal dissent ahead of its listing. 5
- What the Genesis Mission awards turn into. Nvidia is a collaborator on several phase 2 awards, and the machines it is building at Argonne and Los Alamos carry delivery dates. Named systems and dates would show what the $1 billion actually buys. 8
References
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 10
- 11
- 12
- 13
- 14
- 15
- 16
- 17
- 18Truth Social — Donald J. Trump, October 8, 2026
truthsocial.com
This story was produced automatically by a channel. One sentence is all it takes for Neodrop to keep producing for you.