1. Podcasts
  2. Tech Brew Ride Home
  3. Nvidia Walks Away From China
Tech Brew Ride Home

18 Nov 2026 · 22m · 4 chapters

Episode artwork

Nvidia Walks Away From China

In short

Tech news roundup focused on (1) Nvidia’s first earnings call without Jensen Huang and its decision to stop selling H200 chips in China amid new US export controls and possible SEC scrutiny; (2) backlash to OpenAI’s GPT 6.2 Astra launch; (3) “AI-driven patch panic” blamed for clustered outages (FAA ground stop, Microsoft, NSA network crash, banking disruptions).

Guest backgrounds

No podcast guests are interviewed in the transcript. Named experts/commentators include Harvey Jones (Nvidia chair), Colette Kress (interim CEO), and analysts/officials quoted (e.g., Brendan Glavin of OpenSecrets, Jacob Feldgoiz at Georgetown CSET, Dan Fisher-Owens export-control attorney, Dan Shipper of Every, Ethan Mollick, Theo Brown, Joe Slowick at Dataminer, Mark Warner).

Key claims

Nvidia is abandoning China’s H200 market; export rules may cut $44–$62B in data-center revenue next year; SEC/Justice investigations may relate to disclosures about Alice Huang/Megaspeed. OpenAI’s Astra is widely criticized as hard to use/steer. Outages stem largely from rushed patching after AI finds vulnerabilities, not hacks.

Notable examples

H200 China deal previously allowed 25% tariff via US shipment; new rules bar advanced chips in foreign data centers serving Chinese customers; Astra examples include building a space elevator in Blender/Unity and testers seeing inconsistent long-task performance; outages cited include FAA nationwide ground stop and major Microsoft/Chase/NSA disruptions.

Written by AI. May contain mistakes.

Chapters

0:45

Nvidia's Earnings Report

Nvidia reports earnings without Jensen Huang, surpassing estimates but with significant changes ahead.

“They start by identifying where one external shift can impact other areas of your business.”
1:19

Nvidia's Strategic Changes

Discussion on Nvidia's strategic abandonment of China and implications for its future.

“That's pwc.com slash US slash 2026 midterms.”
10:07

OpenAI's GPT 6.2 Astra Reception

Analyzing the mixed reactions to OpenAI's new GPT 6.2 Astra release and its performance.

“saying Astra sets records on work and science evals at under a third of GPT-6 Astra's cost.”
14:00

AI and the Recent Outages

Explores the connection between AI-driven cyberattacks and major service outages.

“In the last week alone, we have had major U.S.”
A man walking through an airport, asking his phone a question

Ask about this episode

Get the gist, explore an idea or ask about something said.

Claude

On Tech Brew Ride Home, what are the key points of this episode?

Searching your listening

Transcript

Automatic transcript. May contain errors.

0:04Welcome to the Tech Brew Ride Home for Wednesday, November 18th, 2026. I'm Brian McCulloh. But today, NVIDIA reported its first Jensen Huang-less earnings, beat estimates, but said it was abandoning China and the SEC might be sniffing around. OpenAI launched GPT 6.2 Astra to a brutal reception from power users, and AI-driven patch panic is getting blamed for this week's many outages. Here's what you missed today in the world of tech.

0:35The midterm elections are behind us, and now everyone is wondering, what's next? What changes are coming down the pike and what impact will they have? They may not have a crystal ball, but PwC does have tried and true methods for helping business leaders navigate times of change. They start by identifying where one external shift can impact other areas of your business. Then they help translate a macro issue like tax policy or AI acceleration into specific business implications. PwC experts are brought in, scenario models are examined, and insights are explored to help you weigh your options. And all that information is used to help build your response plan.

1:12Ready to meet your trusted partner? Get started at pwc.com slash US slash 2026 midterms. That's pwc.com slash US slash 2026 midterms. NVIDIA reported earnings yesterday with the first ever Jensen Wongless earnings call. NVIDIA calms some nerves reporting Q3 revenue up 103 % year on year to$115.8 billion. above the$109.1 billion estimate, and they forecast Q4 revenue above estimates. NVIDIA stock rose 3%. On the call, NVIDIA chair Harvey Jones said the board does not intend to reinstate Jensen Huang as CEO, and they had unanimously accepted his resignation. Quoting The Verge, The NVIDIA board unanimously accepted his resignation and does not intend to reinstate Jensen as CEO.

2:03Jones added, we will not be addressing or taking questions on today's call regarding the circumstances of Jensen's resignation and the ongoing investigation. President Trump's closeness with Wong could complicate things for NVIDIA. The company has ramped up its lobbying in Washington, D.C., spending$5 million in 2025 and$4.7 million so far in 2026. But as Democrats prepare to take Congress, they're planning to investigate Trump's dealings with big tech and other industries, alleging potential corruption. The election wasn't very positive for AI, and the news around Huang's resignation is only going to complicate things, especially for the company in particular, but for AI as a whole going forward.

2:45Brendan Glavin, director of Insights at Open Secrets, told The Verge in an interview last week. Liar Procer, managing partner at Deep33, a tech VC firm, said last week that he didn't expect Huang's resignation to pop the AI bubble, although huang has had a significant hand in mobilizing funds for the ai infrastructure build out in the u.s under huang's leadership in august nvidia entered into partnerships with six financial institutions for about 500 billion dollars to fund such endeavors nvidia supplies roughly 75 percent of the ai compute hardware the industry consumes and demand continues to dramatically outpace supply deep 33 as prosser said in an interview with the verge neither its customers nor its competitors can easily replace the capabilities it provides and the economic incentives driving the AI trade remain unchanged.

3:35Even so, a potential market cooling has some worried as Anthropics IPO slated to potentially be the largest of all time is set for Thursday. The chipmaker had agreed to buy$10 billion in Anthropics shares via private placement on top of a$10 billion funding promise NVIDIA had already made last year, end quote. Now, this is interesting. Also on the call, NVIDIA interim CEO Colette Kress said NVIDIA will no longer even sell those lesser H200 chips in China, ending its role in a deal brokered by Jensen Wang and President Trump that let the U.S. collect a 25 % tariff. So they are, at least for now seemingly abandoning the China market entirely, which makes me wonder, what other details around this do they know that we don't yet?

4:25Quoting TechCrunch, NVIDIA's deal to sell H200 chips in China was brokered by Huang and President Donald Trump. The arrangement required the chips to be shipped through the US, allowing the government to collect a 25 % tariff. The pathway angered China hawks who wanted to deny Chinese labs computing power and maintain the United States' lead in AI, but Huang long argued that it was more important for the U.S. to compete in China. In April, he told podcaster Dwarakish Patel that to concede that market for the United States technology industry is a disservice to our country. Huang's ouster, in part over an affair with the CEO of a company under investigation for smuggling chips into China, has coincided with new rules issued Monday barring advanced U.S.

5:08chips from foreign data centers that serve Chinese customers remotely. Investment bank Jefferies forecast that those rules might trim between$44 and$62 billion from NVIDIA's data center revenue next year. Referencing the new export controls, Kress said that the backlog of demand for advanced chips like Vera Rubin means the rules would have no impact on the company's top-line revenue. In the long term, the company may face the loss of potential customers in Singapore, Malaysia, Thailand, and Indonesia. Critics of NVIDIA's dependence on China claim that as much as 20 % of its revenue comes from demand linked to the country.

5:43Annual revenue at its Singapore subsidiary hit$104 billion in 2025 before NVIDIA stopped disclosing those figures, but that's clearly going to change. Even before NVIDIA decided to stop shipping H200s, China had largely refused to accept their delivery. The government would prefer to generate demand for the country's domestic chip industry, even if major AI compute users want the still superior NVIDIA equipment, according to Jacob Feldgoiz, a researcher at Georgetown's Center for Security and Emerging Technology. Meanwhile, Chinese frontier labs are finding ways to access NVIDIA chips elsewhere.

6:20All of China's leading AI models show evidence of having been trained on NVIDIA Grace Blackwell chips that are prohibited from being exported to China. That training may have taken place at data centers outside of China, but sales to those facilities will be barred under the new rules. If large data centers suited for training AI models want to buy NVIDIA chips, they will have to certify that they aren't servicing any Chinese customers, export control attorney Dan Fisher-Owens told TechCrunch. If they can't provide that certification, the new rules may force them toward other chip options and further eat into NVIDIA's market share, end quote.

6:54One more headache for NVIDIA. The Journal says the US SEC is investigating NVIDIA, including whether its disclosures may have misled investors. Quote, NVIDIA had denied earlier this year that Jensen Huang had an unusual or unexpected relationship with Alice Huang, then an executive at the Singapore-based data company Megaspeed, who US intelligence officials believe is working for China's intelligence services. Megaspeed was one of several companies using NVIDIA chips to set up AI data centers in Southeast Asia and suspected of diverting some of the products to China. It had been under investigation by the Commerce Department since late 2024.

7:35The Justice Department is now leading multiple investigations involving Alice Wong, the Wall Street Journal has reported. NVIDIA last year changed how it discloses international chip sales, eliminating a disclosure that showed its sales to Singapore, where Megaspeed was based. Since the Wong relationship was disclosed, NVIDIA's business in Asia has been in the spotlight. Congress has called for investigations, and Treasury Secretary Scott Besant this week openly questioned whether NVIDIA knowingly sold powerful chips to China to lift its earnings. What was the culture at NVIDIA, and were they pressing these seemingly illicit sales to goose their revenues and their earnings, Besant said?

8:10I think we need answers in the next 10 days from the board and from the outside counsel, end quote.

8:22It's that time of year when your car starts feeling like a second home from business meetings, sports practices, road trips, and school pickups. You might be spending a lot of time on the road, too. AT &T Connected Car helps you stay connected wherever life takes you. It turns eligible vehicles into a Wi-Fi hotspot on wheels, helping you and your passengers stream, browse, work, and stay in touch on the go. Because your calendar isn't slowing down anytime soon, and your connectivity shouldn't either. Whether you're catching up on work between stops or keeping passengers entertained, AT &T Connected Car helps keep everyone connected.

8:56That's tech that works where life happens. See if your vehicle is eligible at att.com slash techbrew. That's att.com slash techbrew. Requires eligible vehicle and plan. Service and coverage not available everywhere. Restrictions apply.

9:14Think about the one person on your team who knows everything about a critical process. What happens when they're out of office or leave the company? That's the problem today's sponsor, Scribe, was built to solve. It captures workflows as work happens and automatically turns them into step-by-step guides. So no one has to write documentation from scratch. Your team reviews each guide and sharing takes seconds. Anything outside admin-approved apps is never captured. And the guides do double duty. See, most AI was trained on the internet, not on your company. With Scribe, captured workflows can give your AI agents a foundation in how your team actually works.

9:53Visit scribe.how slash ridehome for your first month of Scribe capture free on select plans. That's S-C-R-I-B-E dot how slash ridehome. OpenAI has rolled out GPT 6.2 Astra and Sol in chat GPT Work Codex and API, saying Astra sets records on work and science evals at under a third of GPT-6 Astra's cost. Artificial analysis says that GPT-6.2 Astra trails only Claude Fable 5.2 in its main index and is flat versus GPT-6 Astra encoding. Epic AI says Astra tied with Fable 5.2 and yet what is odd about this model launch is how quickly people have rushed out to say they don't like it. Online, you see folks saying the new model is frustrating to use, has no concept of knowledge gaps, its outputs are very hard to parse, and it runs in circles on long tasks.

10:54Quoting VentureBeat, investor and former HyperWrite AI CEO Matt Schumer said he won't be releasing a longer written review as he has customarily done for other preceding OpenAI flagships. The new models were essentially unusable, Schumer wrote in a post on X. Developer and YouTuber Theo Brown also posted on X that 6.2 Astra was the worst model OpenAI has ever released. They single out the model's shortcomings in delegated work and creative tasks. Wharton professor Ethan Mollick had a more mixed experience with Astra. It built a playable space elevator in Blender and Unity, but he found the model very uneven and weird and harder to steer than rival models.

11:37Astra now costs$8 per million uncashed input tokens and$40 per million output tokens, while GPT 6.2 Sol keeps its predecessor 6.1 Sol's$2 input and$10 output rates. Both are proprietary hosted models. They're available now in ChatGPT Work and Codex for Plus Pro Enterprise and EDU users and through the OpenAI API as GPT 6.2 Astra and GPT 6.2 Sol. Dan Shipper, CEO of generative AI publication Every, said his team spent several weeks testing both models. He said Astra performed about as well as its predecessor on most of Every's coding benchmarks, but often stopped early on end-to-end builds. He found that its push for brevity could make writing hard to follow or leave out necessary explanation.

12:26He praised its default design instincts but said its outputs were difficult to vary and had a noticeable house style. That distinction matters for teams that need one model to work across several brands or change the depth of an explanation for different readers. In a follow-up post on X, Brown described how he and another tester independently asked Astra to design new levels for different games, Outer Wilds and Majora's Mask. Both got water levels centered on music and bells with the same name, Bellwether. The two runs show how similar Astra's choices could be even when the projects differed.

13:02Brown laid out even more objections in a long X post entitled Why I Hate GPT 6.2 Astra, saying it was the full text of a letter he had sent OpenAI during early testing. On Brown's experimental TypeScript compiler rewrite in Rust, he said Astra spent seven days without landing an improvement and produced a version that ran 56 % slower after focusing on reducing its size. When he redirected it to improve speed, he said it spent another four days without landing a change. Brown did, however, credit Astra with catching issues other models missed in code reviews and praised its 3D game development and computer use capabilities.

13:40He said it helped him devise a reverse proxy that he expects will save him$25 ,000 a month. His criticism focused on how hard it was to keep Astra moving on long projects and understand what it had done afterward. According to Brown, OpenAI addressed a few of his complaints before launch, and he still preferred Claude Opus 5.5 for his development work, end quote.

14:08We kind of didn't talk about this because I have a sort of standing rule where if there's a major outage, if it's likely to be resolved by the time I get to tell you about it, we don't talk about it, trying to avoid stale news. But in this case, that rule failed me. In the last week alone, we have had major U.S. banking disruptions, the NSA's classified internal network has crashed, there was a huge outage to Microsoft services, and of course, that huge nationwide FAA ground stop. Why all these outages all at once? Is AI behind it? Well, the answer is yes, but in an oblique way. Basically, everyone is rushing to fix bugs that AI have exposed, and it is in applying those patches quick and dirty that the problems are arriving, quoting the journal.

15:00Behind the scenes, there have been relentless AI-driven cyberattacks, an avalanche of software patches amid bugmageddon, sleepless nights for IT staffers, and so much cybersecurity work that incident responders are turning away business. While outages are nothing new, AI has created a pressure cooker that makes it difficult to discern cybersecurity threats from the ceaseless updates that amount to their remedy. We have two sets of things. We have systems like the FAA systems, which are known to be ancient and really need an upgrade, said, and Neuberger, a general partner with the venture capital firm Andreessen Horowitz and former White House Deputy National Security Advisor.

15:37In addition to that, we have this whole set of AI attackers that are able to find vulnerabilities in code. The entities affected by these outages, Microsoft, the FAA, Chase, the NSA, Workday, Apple, and United all say that the root problem wasn't a hack. In most cases, it was a system upgrade gone wrong. Patches are essentially changes to code, Newberger said. In some cases, you get unexpected outcomes. Just two months ago, technology executives were warning the public of an AI apocalypse and Dario Amodai, whose company Anthropic is expected to make its initial public offering later this week, warned that rogue AI systems could take down the internet.

16:15It hasn't helped calm nerves that cybersecurity pundits dubbed this past month Black October as investigators responded to an unprecedented number of intrusions, many of which were fueled by a new AI hacking tool called Artex. Cybersecurity experts say that there is no doubt that attackers on our infrastructure pose a national security risk. Earlier this year, U.S. water systems were targeted in a cyberattack linked to Iran. Refrigerators across Europe were shut down in another Iran-linked incident last month. On Friday, the same day the Federal Aviation Administration ordered a nationwide ground stop, the Federal Bureau of Investigation put out a warning about ARTEC saying it was being widely used by hackers to speed up their attacks.

16:56The timing, it turns out, was coincidental. On Tuesday, an AI spokesman said the alert wasn't connected to the current outages. There's been so much reporting of the impending AI apocalypse, and then when ordinary people see systems going offline, they immediately jump to the conclusion that it's cyber, even if it's not, Mulholland said. This year, there have been an unprecedented number of system upgrades in North America, and security experts say that it's fair to blame that phenomenon on AI. In April, Anthropic released Mythos, a product tuned for cybersecurity that has proved to be uncannily good at fighting software bugs.

17:29OpenAI followed with models of its own. As a result, this year has been unlike any other when it comes to the number of bugs found and patches issued. Last month, Microsoft released patches fixing about 1 ,400 security flaws in its products during its monthly software update. It was a record number of patches a year earlier. The tally was about 170. Flaws in another software product called PHP came to light in October and were considered so serious that many organizations shut down their websites rather than risk being hacked. In recent weeks, other serious bugs were uncovered in the Linux operating system and the Java programming language, both of which are ubiquitous among American corporations.

18:08This isn't the first time the world has found itself in a global race to patch. Back in 1999, the Y2K problem was fixed with barely a hiccup. But that was a time when many Americans connected to the internet sporadically via a landline. Since then, the internet has wormed its way into our lives in ways that were unimaginable back in the 1990s. That has made the race to patch all of these systems more chaotic, and it's made the side effects of a botched patch even more significant, end quote. End quoting Bloomberg. Breakdowns over the past week have caused enough havoc to spur American consumers into a panic buying spree.

18:44At Costco, surging purchases of non-perishable foods, water, and battery supplies led analysts to increase its share price target to$1 ,150 from only$1 ,100. Most of the company's high-ticket generators were sold out online as of Sunday. Late-night comedians have also feasted on the tech-related turmoil. Saturday Night Live dedicated its cold open over the weekend to a faux CIA meeting where an agent played by comedian John Mulaney used problems at the Federal Aviation Administration and Microsoft as a punchline. It's unusual and concerning to have so many high-profile incidents clustered so closely together, said Joe Slowick, director of threat research and cyber engineering at the intelligence firm Dataminer.

19:27We have become far too reliant on infrastructure that is shaky being more available than it is, he said. Seeking to quell disquiet, Secretary of State Marco Rubio on a trip to Asia with President Donald Trump told reporters early Wednesday that the U.S. had no indication or evidence that the technology breakdowns were related or the result of hostile acts. Taken together, however, the episode's touch on the broader anxieties surrounding artificial intelligence, including the possibility that AI tools could make it easier for malicious actors to knock out essential services or conduct cybercrime sprees.

19:57Senator Mark Warner, Democrat from Virginia, said in a statement that it appears the incidents are the consequence of how quickly artificial intelligence is changing the cybersecurity landscape. AI is making it possible to identify vulnerabilities across enormous numbers of systems at unprecedented speed, creating enormous pressure on companies and government agencies to patch those vulnerabilities immediately, often all at once, putting tremendous strain on systems that weren't built to handle that pace of change, Warner said, end quote.

20:34Nothing more for you today. Talk to you tomorrow.

20:41Your sausage McMuffin with egg didn't change. Your receipt did. The sausage McMuffin with egg extra value meal includes a hash brown and a small coffee for just five dollars only at mcdonald's for a limited time price and participation may vary promotion pricing may be lower than real pricing

From the publisher

Nvidia reported its first Jensen Huang-less earnings, beat estimates, said it was abandoning China, as the SEC probed. OpenAI launched GPT-6.2 Astra to a brutal reception from power users. And AI-driven patch panic was blamed for this week’s outages.

Nvidia reports Q3 revenue up 103% YoY to $115.8B, above $109.1B est., Data Center revenue up 110% to $107.6B, and forecasts Q4 revenue above est.; NVDA rises 3% (Nvidia)

Nvidia’s board unanimously accepted Jensen Huang’s resignation and does not intend to reinstate him as CEO, and won’t take questions on the circumstances or the ongoing investigation (The Verge)

Nvidia is abandoning China after Huang’s ouster, no longer selling even the lesser H200 chips, as new export rules bar advanced US chips from foreign data centers serving Chinese customers (TechCrunch)

The SEC is investigating Nvidia, including whether its disclosures misled investors about Huang’s relationship with Megaspeed executive Alice Huang and its Singapore chip sales (WSJ)

OpenAI debuts GPT-6.2 Astra and Sol in ChatGPT Work, Codex, and API, saying Astra sets records on work and science evals at under a third of GPT-6 Astra’s cost (OpenAI)

OpenAI’s GPT-6.2 Astra gets a negative reception from power users, with Matt Shumer calling the models “essentially unusable” and Theo Browne calling Astra “the worst model OpenAI has ever released” (VentureBeat)

Sources: Wall Street execs and the NSA told Congress this week’s outages stemmed from systems overwhelmed by a rush to patch vulnerabilities found by AI models (Washington Sun)

AI-driven cyberattacks and an avalanche of patches created a pressure cooker; the outages at Microsoft, the FAA, Chase, the NSA, Workday, Apple and United were mostly upgrades gone wrong, not hacks (WSJ)

Outages roil public trust: panic-buying at Costco, Rubio saying there’s no evidence the breakdowns were related or hostile, and Sen. Warner blaming the speed of AI-driven vulnerability discovery (Bloomberg)

Subscribe to the ad-free feed.
Learn more about your ad choices. Visit megaphone.fm/adchoices

More from Tech Brew Ride Home

All episodes