Tech Brew Ride Home - OpenAI Takes a $150B Haircut

Episode Date: October 23, 2026

OpenAI neared a $30B bridge round at a $1.25T valuation, down from the $1.4T it pitched, ahead of an IPO. Treasury proposed curbs on US tech investment in China, banks blocked rogue AI agents, and OpenAI launched its Decisions API. Sources: OpenAI nears an up to $30B round led by investors from the UAE and Coatue at a $1.25T pre-money valuation, as an “opportunistic” bridge ahead of an IPO (CNBC) The US Treasury proposes a rule to crack down on US high-tech investments in areas like AI and supercomputing in China, Russia, and other “countries of concern” (AP) DeepSeek and MiniMax have won approval to deploy their models overseas, as Beijing says much of Chinese AI will remain free for export despite its draft export-control proposal (Bloomberg) OpenAI makes Decisions API, its Jev competitor, generally available with new voice and visual modalities, at $0.10/1M input tokens and no output-token charges (TechCrunch) Sources: US banks are restricting third-party AI agents from accessing customer accounts, after some agents appeared to act independently of users’ instructions (Axios) Sources: Major US banks, Morgan Stanley among them, have been blocking third-party AI agents from operating inside customer accounts and making transactions, as Jamie Dimon and Schwab’s Richard Wurster say outside agents aren’t getting in (Reuters) Longreads Matt Cortland used an ElevenLabs voice AI agent to make more than 3,000 calls to pubs across Ireland about the price of a Guinness, then used Claude to build the “Guinndex,” a living consumer price index for a pint (Fortune) Subscribe to the ad-free feed.

Welcome to the TechBrew right home for Friday, October 23rd, 2026. I'm Brian McCullough today. Open AI neared a $30 billion bridge round at a $1.25 trillion valuation down from the valuation it pitched. Treasury proposed curbs on U.S. tech investments in China. Banks blocked rogue AI agents. OpenAI launched its Decisions API and in the weekend, Long Reads, the best use of AI I've heard yet. Here's what you miss today in the world of tech. Oh, my pod. That's what you might say when your first international. produced to the pod by 8 sleep. It's a smart mattress cover that goes on top of your existing mattress, actively cooling or heating each side of the bed independently, which is ideal

for falls changing temperatures. The new gen Pod 6 is 50% smaller than the previous generation, packs nine times more sensors, and delivers more powerful active heating and cooling. This means the pod can now feed an even more accurate signal into autopilot. Members see up to 34% more deep sleep. No wearable is required. Use code ridehome at 8Sleep.com for a special offer. The Pod 6 starts at $1,99 and you can have 30 nights to try it at home. Return it if you don't love it.

Sources say OpenAI is finalizing an up to $30 billion round, led by investors from the UAE and CO2 at a $1.25 trillion pre-money valuation, as what they are calling an opportunistic bridge round ahead of an I. IPO. But note, that means that OpenAI has cut its valuation target to $1.25 trillion down from the $1.4 trillion valuation it pitched earlier in October, quoting CNBC. OpenAI kickstarted discussions for a new funding round after investors approached the company with proposals in September, as CNBC previously reported. OpenAI told investors on Thursday that it is aiming to secure signatures by the end of next week, but the final size of the round is still being determined, the people said. alongside the new investment, OpenAI is negotiating a commercial deal that would secure more computing capacity from the UAE according to one person.

Open AI told investors in February that it's targeting roughly $600 billion in total compute spend by 2030. The company confidentially filed its prospectus with the Securities and Exchange Commission in June, but it has not disclosed exactly when it plans to debut. opening I CFO Sarah Friar told employees in August that the AI Lab will be a public company in 27 the following month CEO Sam Altman said that right now would be an ill-advised moment to go public, end quote. And quoting the FT. The chat GPT maker is in the final stages of a deal which is expected to value the company at $1.25 trillion, according to people with knowledge of the matter. That would be a jump of almost 50% from its current $852 billion valuation, but is low. lower than the price pitched to investors earlier this month, according to people familiar with the matter.

The lower valuation target comes as U.S. tech stocks have sold off sharply due to intensifying concerns about the sustainability of the debt-fueled AI infrastructure boom. Morgan Stanley and San Francisco-based tech fund Dragonere had floated Open AI to the bank's private wealth clients at a pre-money valuation of $1.4 trillion, according to documents seen by the F.T. Multiple people with knowledge of the Morgan Stanley-led process said demand from clients was weaker than in the past. There has been zero clamoring, said one of the people. All the prior times opening I have been on, there's been a frenzy. This time it's a bit of a thud. Some investors were also suffering from fatigue or indigestion having invested heavily in prior rounds, said some of the people.

Opening I raised $40 billion in March 2025 and a record-breaking $122 billion in March this year. Arch rival Anthropic has also raised close to $100 billion this year. Other investors told the F.T, they would rather wait and see how Anthropic prices in an initial public offering expected later this year before committing to its rival. Open AI is targeting raising $30 billion as a bridge before its own public offering that was likely to come in the first half of 2027, said a person close to the company, though no date has been fixed. Jimmy Chang, chief investment officer at the Rockefeller Global Family Office, which has exposure to Open AI and Anthropics, said uncertainty around both of the company's IPOs was weighing on the broader U.S. U.S. stock market. The market needs to see a successful IPO, said Chang. These companies need funding, end quote. Two conflicting stories here, first quoting the AP. The U.S. Treasury Department proposed a rule Thursday to crack down on American high-tech investments such as supercomputing and artificial

intelligence in China, Russia, North Korea, and other countries of concern. The U.S. government has been working for years to deny China and other adversaries' investments that could help them boost their military power. That program began in 2023 when then President Joe Biden signed an executive order putting restrictions on outbound U.S. investments in sectors such as semiconductors, quantum computing, and artificial intelligence going into China. Treasury later established the Outbound Investment Security Program. In December 2025, Congress passed a law that codified and expanded the program to target more technologies and more foreign countries, including Cuba, Iran, North Korea, and Russia. The rule announced Thursday would expand the restrictions to cover

investments in hypersonic systems capable of traveling at five times the speed of sound, super computing, biotechnology, and artificial intelligence, though Treasury used Trump's favored term for AI, superintelligence, or SI. Their proposed rule sends an important signal to markets that Washington is tracking where U.S. capital is flowing and what capabilities it is helping build, said Craig Singleton, a senior fellow at the Think Tank Foundation for Defense of Democracies. The screening of outbound investments recognizes that capital can accelerate military relevant capabilities just as surely as technology transfer can, Singleton said. Jason Chipman, a former Justice Department lawyer who specializes in cybersecurity investment rules and export controls as a partner at the law firm Wilmer Hale, said the new rule will require more substantial compliance programs to manage risk. Treasury's rule to implement the 2025 law, the Comprehensive Outbound Investment National Security Act, would only apply to investments that accept.

exceed $50,000 and would exclude many financial investments, including stock purchases. Treasury will hear public comments on the proposal before issuing a final rule, end quote. This is sort of the inverse of that, then, quoting Bloomberg. DeepSeek and Minimax Group have won approval to deploy models overseas, part of a wave of homegrown AI technologies that Beijing will allow to operate globally, a Chinese state media affiliate said. Much of the Chinese AI will remain free for export, according to a post-published by a social media account affiliated with state broadcaster China Central Television. It addressed Beijing's draft proposal last week for export control measures that would restrict

certain artificial intelligence tools and services, including large language models and agents. Their proposed restrictions could limit overseas access to China's most advanced models, potentially slowing an international surge in adoption driven by cost-effective Chinese open-source AI. Chinese startups have made significant strides closing the gap on U.S. market leaders this year, Beyond more affordable pricing, they've also attracted users by making their models easier to customize than proprietary systems like the ones on offer from OpenAI and Anthropic PBC. Beijing's interest in controlling the export of what's seen as strategically important technology has shown up most prominently in the way it forced the unwinding of meta-platforms acquisition of Manus and AI agent startup with Chinese roots. There are significant carve-outs to the proposed regulations. providing API access, web applications, or software-as-a-service features will not be subject to export

restrictions, even if powered by restricted underlying models, according to the Post. The Post contrasted Beijing's approach with that of Washington, accusing the U.S. of using targeted export bans and technology blockades to protect its competitive dominance. China, however, is not following this path. The Post added, end quote. OpenAI has a Jev competitor. It's called Decisions API, and it is generally available with new voice and visual modalities at 10 cents per 1 million input tokens and no output token charges, quoting TechCrunch. TypeSafe's Jev model created a category for deep learning models that act as superclassifiers, given a set of options they can quickly make choices and offer a confidence. rating in their decision. Software developers quickly found that these decisions or system one models could replace LLMs in a variety of workflows operating more quickly and at a dramatically lower cost.

Nukunj Honda, the product lead for OpenAIs APIs, told TechCrunch that his colleagues saw Jev and said, wow, we have like a million use cases for this. Honda's team sprinted to offer their own version rather than building a specifically calibrated decision model. The team took ChatchipT, Luna and configured it to speed up the inference and lower the cost. The big unlock with Decisions API has been the speed, shrinking things from 1 to 3 seconds to 120 to 180 milliseconds, Honda said. In the future, we may sort of like have specific training to make our models like even better in these ways, but right now it's just plain old Luna that gets the job done. Announced at the company's Developer Day on September 29th, the Decisions API

became available to all customers on Thursday. Opening I added new voice and visual modalities to the public version and worked to ensure lower latency in regions around the world by targeting its inference resources geographically. In initial trials with customers like Box, Datadog, Ramp, and Speak, engineers used the API to replace OpenAI's LLMs with higher throughput. Andrew Sue, the CTO at Speak, which builds AI language tutors, said the Decisions API allowed the company to cut the time between actions from 1.4 seconds to just 0.13 seconds. also use the product to improve their dots voice assistance responsiveness. Still, it's not clear

exactly what makes a good decision model other than it works for you. Optimization across intelligence, costs, calibration, and speed seems to differ by workload. TypeSafe CEO Diego Almeida distrust benchmarks and urges people to trial the models on their own work, but even TypeSafe released its own rankings last week alongside the release of its latest Jev model, which includes new features including support for images and smart if statements. Public benchmarks like the AIM System 1 Index and artificial analysis decision index tend to favor D1 Ultra, a new decision model that Liquid released last week and BaseLabs, Elephant 1. The latest Jev model remains competitive on both indices in 7th and 3rd, with OpenAI's decision

model following 8th and 14th, respectively. However, TypeSafe's offerings are for now the most widely queried on OpenRouter. The most fun Jev benchmark, however, is Jev Man, a benchmark where decision models play Pac-Man. The latest Jev model tops that ranking. Felix Wonderlich, the chief growth officer at Upper AI, which built the benchmark, said that the classic game is an interesting test because of its nonstop stream of quick high-stakes decisions. At every junction, the model must decide where to go in a few hundred milliseconds. Where Open AI's arguably more capable model may have an advantage is in multimodality,

For image classification, Luna is extremely powerful and has been showing better calibrations, Honda told TechCrunch, adding that Open AI customers want the decision API to be more flexible about using images as outputs as well. Honda also saw further increases in speed enabling new use cases. Can we get this to something like, you know, 60 milliseconds? And then you could imagine, like, an API-like call for every single frame in like a video. As these offerings get faster and cheaper, we can expect them to live. up to Jeb's namesake, an economist who explained how using a resource more efficiently makes it more ubiquitous.

Or, as Honda sees it, there will be an expanding crazy set of use cases that were hiding in plain sight until the decision models revealed them, end quote. Cyber security was built for a different time, one without AI. Rubrik built agentic cyber resilience for what's next. Their platform is built to help you keep pace with AI and keep your organization secure. The agentic cyber resilience platform covers your data, your identities, and your AI systems, so no part of your organization is left exposed. Move fast and stay secure at rubric.com slash mb. That's rubr-i-k.com slash mb. If you're struggling to scale your business, it's probably not because you need more headcount or a bigger budget. Without documentation of critical knowledge, your team can't work efficiently. That's what Scrib.

is built to fix. When specialists and experts leave your team, their knowledge leaves too. With Scribe, you can avoid the scramble for information. Here's how it works. Scribe captures workflows in real time and automatically generates the documentation for how your company works with no manual writing. It automatically redacts sensitive information like names and account numbers from screenshots, and nothing is published without human review. With Scribe, your team can launch real-time on-screen guidance that shows exactly where to click inside the actual tool so they can move forward with confidence. Captured workflows also become the foundation for AI agents so they can operate on how work actually happens, not guesses. Learn more at scribe.com. How slash right home

and mention TechBrew Right Home for your first month of Scribe Capture free on select plans. That's S-C-R-I-B-E dot how slash ride home. Sources say U.S. banks are restricting third-party AI agents from accessing customer accounts after some agents appeared to act independently of users' instructions. Quoting Reuters, major U.S. banks have been blocking third-party AI agents from operating inside customer accounts and making transactions to people familiar with the situation.

Said, following a report that AI agents were accessing bank accounts and taking actions independent of user instructions. The banking industry has become increasingly concerned that sophisticated AI agents could harm the banking system and the institutions trusted to hold people's wealth. An Axios report on Thursday said some AI agents attempted to take actions that appeared to be independent of user intentions, citing people familiar with the matter. In some cases, bank customers have shared their login, ID, and password with agents with banks typically flagging and blocking those transactions, a person in the banking industry said. Still, such transactions cannot be deemed unauthorized if credentials were shared by the customer.

Morgan Stanley has blocked third-party agents from accessing its platform, but had not seen any instance. of unauthorized or rogue AI agents, according to a person familiar with the matter. Top banking executives earlier this month discussed the risks of AI agents acting on customers' behalf with some warning the technology is adding significant risks for both customers and banks. J.P. Morgan Chase CEO Jamie Diamond earlier this month said the largest U.S. lender was developing tools to navigate the risk of AI agents carrying out tasks on their own. We're not going to allow just any outside agent to come inside the company because that's just too risky for clients and for J.P. Morgan, he said at the time. Charles Schwab CEO Richard Rooster earlier this month also said the brokerage was blocking unauthorized agents.

I get a report every day tracking how many times we've had to block one of the different agents, and it's not many because I think people aren't using it in that way in their financial life. We are blocking it because we want to have control over authentication and the experience Worcester said, end quote. Finally today, only one long read this week, and it is hands down the best use of AI I've heard of yet. Quoting Fortune, have you ever overpaid for a beer? Matt Cortland has and hit set him on a path to never repeat the mistake. That is, for Cortland's drink of choice, a pint of Guinness. after paying 7 euro 80 cents about $8.93 for an Irish dry stout at a pub in Dublin earlier this month. The 37-year-old grew curious about the average cost of a pint across Ireland. To his astonishment, the country's central statistics office had dropped price tracking of the nation's most popular beer in 2011.

That led Cortland to the wild idea of tracking the price himself. Cortland, founder of an AI startup, turned to AI to lend him a hand. a voice. He devised Rachel with AI voice generation platform 11 Labs, made as an homage to Rachel Duffy, the winner of the UK version of the reality TV show The Traders, and equipped with a northern Irish accent. The voice-enabled AI agent made more than 3,000 calls across the island inquiring about the price of a pine of Guinness. I was like, well, can I just call every pub in Ireland and conversationally ask them with AI? Cortland told Fortune. I pulled the thread, and I just kept pulling the thread, and here we are.

Using the data accrued from the thousands of phone calls, he then turned to Anthropics Claude to devise the Gindex, which he calls a living, breathing, consumer price index for a pint of Guinness across Ireland. It also allows bartenders and beer drinkers to contribute and modify prices. Now Cortland can see how his 7 euro 80 cent pint weeks earlier matches up with the rest of Ireland. On Monday, the average price was about $6.08, $0.08. and the most common price was five and a half euros, or $6.30. Guinness Parent Company Dia Geo didn't respond to Ford's request for comment.

Beer prices are independently set by pub owners across Ireland. Rachel, Cortland's AI agent is one of a growing number of voice AIs that are appearing on the other end of your phone line. Data from Voice AI firm Regal showed that customers are finding the AI as credible as humans. Based on data from millions of calls with Voice AI agents, people are taking 14% more time to chat with AI than they would with a human representative. They're also giving 22% longer responses, sharing details they'd normally skip. Cortland said he saw similar results.

The conversations his AI had across Ireland showed that most didn't realize that they were communicating with AI. The transcripts of some of those conversations reviewed by Fortune make that clear. The cost of a pint of Guinness, 25 pounds. But if you're coming in for a wee drink, I'll give it to you for a four. Fiver, a bartender at Dugies in Eniskillen, Northern Ireland, told Rachel, listen, they're normally 620, but if you can't afford one, we'll buy you one. We'll look after you, a bartender at Malzard's pub in Kilkenny, Ireland, told the AI. While the Gindex hasn't yet led to a dramatic price shift,

Cortland said he has already seen it yielding results. In one instance, he said a pub owner reportedly lowered the cost of his Guinness by 40 cents and then updated the entry on the Gindex himself. Cortland's hoping to replicate the success of the Gindex for other products, perhaps for prescription drugs in the U.S., where he is originally from, or even for a slice of pizza in New York City, end quote. I am going to have some weekend bonus episodes for you next weekend. I swear, famous last words. Talk to you on Monday.