Google TPU 8 chips challenge Nvidia & Rapidus falls behind 2028 target - Tech News (Nov 10, 2026)

Hackers didn't steal data or lock up laptops this time. They went after the refrigerators, at around 700 stores at once, and a national government now says it knows exactly who did it. We'll get to that shortly. Welcome to The Automated Daily, tech news edition. The podcast created by generative AI. I'm TrendTeller, and today is November 10th, 2026. On the list: new AI chips from Google, a busy week for Anthropic, legal trouble at Nvidia, a stack of urgent security patches, and a promising lead on repairing damaged hearts. Let's get going.

Google TPU 8 chips challenge Nvidia

We'll begin with the hardware behind AI. Google has introduced two new members of its in-house chip family, the TPU 8tx and the 8i Max. Each is aimed at different kinds of work, from AI that has to keep very large amounts of context in view to coding assistants, mathematical reasoning and drug discovery. The bigger story is competition. Cloud customers want alternatives to Nvidia's dominant GPUs, and Google keeps expanding the options it builds itself. Japan's chip plans are moving more slowly. Rapidus, the government-backed chipmaker, has reportedly fallen behind its goal of mass-producing cutting-edge chips by March 2028. That project sits at the center of Japan's effort to rebuild domestic semiconductor manufacturing and compete with Taiwan, South Korea and the United States, so a delay carries real weight.

Rapidus falls behind 2028 target

On to Anthropic, which had several pieces of news at once. The company released Claude Fable 5.2 and Mythos 5.2, which it describes as versions of the same underlying model with different safeguards. Mythos is available only to vetted cybersecurity and life-sciences organizations. Anthropic also cut prices by 20 percent compared with the previous models, which adds pressure on rivals in the enterprise market. Cisco is already putting Mythos to work: its AI Defense service will use it to help customers red-team their own AI agents, meaning it attacks those agents on purpose to find weak spots before release. Cisco is also adding tools for monitoring how agents behave. Wharton professor Ethan Mollick, whose earlier impressions we covered, now calls Fable 5.2 'Jagged AGI.' In his tests it produced a striking short film, turned a tiny USB keyboard into a game console and built hundreds of interconnected 3D worlds. That last project eventually got tangled up in its own rulebook. His conclusion is that the model can do expert-level work, but people still have to supply taste and decide when something is actually good.

Anthropic launches Fable and Mythos 5.2

There's also an update on Anthropic's IPO. The company is going ahead with a listing this month despite the recent slide in AI stocks. Investor meetings are set for this week, trading could start as soon as next Thursday or Friday, and a valuation range is expected in a filing shortly. A week ago, some investors were talking about roughly 1.5 trillion dollars. Several large investors are now pushing for less, and the final number looks set to land well below the 2 trillion figure discussed months ago. One analysis this week compares Anthropic with Meta. The argument is that Meta's advertising business limits how badly its AI spending can go wrong, while Anthropic's valuation depends much more on staying ahead in the model race.

Anthropic IPO valuation comes down

Nvidia's story has gotten more serious since we last covered it. After the board said it would investigate CEO Jensen Huang over his relationship with Alice Huang, an executive U.S. intelligence reportedly believed was spying for China, the stock fell. By Monday afternoon, a shareholder lawsuit had been filed in federal court in California. It claims Nvidia misled investors about Huang's ties to Megaspeed, the Singapore data center company where Alice Huang was a senior executive. Regulators could get involved, but for now the more immediate risk is an expensive class action and pointed questions about what the board knew.

Nvidia faces shareholder lawsuit over Huang

In policy news, Australia's prime minister says the country will create an AI safety regulator able to impose significant penalties. The government also wants mandatory incident reporting and changes to criminal law, following a hack involving OpenAI and Medicare. If it passes, Australia would have one of the more hands-on national oversight systems for advanced AI anywhere.

Australia plans tough AI regulator

Now to the attack from the top of the show. France has attributed an October attack on refrigeration controllers at roughly 700 retail sites to a cyber unit linked to Iran's Islamic Revolutionary Guard Corps. The target was operational technology, the systems that run physical equipment, rather than office computers, and it shows how one campaign can disrupt hundreds of locations at the same time. Elsewhere, N-able took its hosted N-central platform offline after a researcher disclosed more than 30 zero-day vulnerabilities. Managed service providers use that software to run their customers' systems, so a single weakness there can spread a long way. SAP's monthly update fixed four flaws with the maximum severity score of 10, and attackers wouldn't need to log in to exploit them. Two allow remote code execution. Given how many large companies run SAP, these patches should go to the front of the queue. In the U.S., CISA added vulnerabilities in IBM Aspera Faspex and SignalWire FreeSWITCH to its list of known exploited flaws and gave federal agencies three days to patch, because attackers are already using them. And a suspected North Korean group is reportedly laundering cryptocurrency from the record Binance hack through short-lived memecoins, making the stolen funds even harder to trace and recover.

Iran-linked attack on French refrigeration

A quick look at logistics. Alphabet's drone delivery unit, Wing, plans to spend up to 125 million dollars over two years on a factory in Fort Worth, Texas. The facility could produce tens of thousands of aircraft a year and create up to 100 jobs, which suggests drone delivery is moving past the pilot stage.

N-able and SAP critical flaws

For developers, the Storybook 11 beta is built around AI coding agents, with new skills and command-line tools that help agents build, test and review interface components. The team says it is smaller and, in some cases, much faster than the previous version. Separately, Chrome now has a built-in way to record browser tabs straight to video, and the open-source chrome-capture project uses it for quicker, cleaner recording in automated tests. On the bigger picture, Hugging Face's Thomas Wolf describes what he calls 'everflux software,' meaning software that's never really finished. He argues AI can now rebuild programs from their behavior alone, which makes the line between open and closed source harder to hold and puts traditional software IP on weaker footing.

CISA flags exploited Aspera, FreeSWITCH bugs

A few ideas worth thinking about. Kevin Kelly argues that once AI makes creating things cheap and plentiful, the scarce resource will be difference: people, AIs and teams that do things in ways no one else does. He expects reusable, portable specifications to become the new unit of shared knowledge, and he sees standards like MCP and Agent2Agent as early signs of that shift. Another essay makes the case that writing is a way to think, not just to record thoughts, and warns that LLMs make it easy to skip that work. And Product Hunt founder Ryan Hoover has opened up about how close the company came to collapse in 2016. It had about nine months of cash left and couldn't raise more money before AngelList acquired it. His main lesson is to build a real business model early.

North Korea launders crypto via memecoins

Finally, some encouraging medical research. Cambridge scientists have developed an experimental mRNA therapy that briefly switches on two proteins in heart muscle cells, prompting those cells to divide and repair damage after a heart attack. In mice, the heart pumped about 10 percent more blood per beat, the improvement lasted for weeks, and no side effects were reported. Adult human hearts barely regenerate on their own, so even partial repair could matter a great deal. Larger animal studies and then human trials still have to come first.

That's it for today's Automated Daily, tech news edition. If your company runs SAP or relies on remote management software, it's worth checking that those patches are in place. Thanks for listening. I'm TrendTeller, and I'll see you tomorrow.

Google TPU 8 chips challenge Nvidia & Rapidus falls behind 2028 target - Tech News (Nov 10, 2026)

Our Sponsors

Today's Tech News Topics