Amazon’s $7B Decart Talks, Nvidia Circles Reflection AI, and Japan’s Cyberspace Emergency

A photorealistic wide view of a modern data center hall with glowing server racks and blue AI chip circuitry, illustrating the AI infrastructure race between Amazon and Nvidia.

Two mega acquisition talks, a funding frenzy that won’t quit, and a cyber emergency showing AI’s darker side — all in today’s Daily Tech and AI.

Amazon in Talks to Buy Decart for Around $7 Billion

The biggest deal headline of the weekend: Amazon is in talks to acquire an AI startup called Decart in a deal valued at around $7 billion, according to the Wall Street Journal.

Decart is a San Francisco startup founded three years ago by three Israeli engineers, and its software solves a problem the whole industry is starting to feel. It makes it easier for developers to move their AI workloads between different AI chips — a real pain point, because most AI models are tuned for one specific chip, usually Nvidia’s, and moving them takes a lot of expensive work.

Why does Amazon want them? This is about Amazon Web Services and its own AI chips. Amazon has been building its Trainium family of custom AI chips to compete with Nvidia, and the Journal reports Amazon gave Decart early access to its newest chip, Trainium 3, back in 2025. In early tests, Decart’s models on Trainium hit four times the frame throughput and twice the cost efficiency of leading GPUs. Amazon wants the software layer that makes its chips the easy choice.

And the backstory is juicy. Decart was valued at nearly $4 billion in May, and Bloomberg reported that Anthropic explored buying Decart for $6 billion in August, then walked away after due diligence. So Anthropic said no, and now Amazon is offering $7 billion. The Journal says a deal could come together soon if the talks hold. One to watch.

Nvidia Circles Open-Weight Startup Reflection AI

From one acquisition talk to another. Nvidia is reportedly in early talks to buy, or invest much more heavily in, an open-weight AI startup called Reflection AI. That came from the Financial Times on Saturday, via Reuters.

Reflection was founded in 2024 by two former Google DeepMind researchers, Misha Laskin and Ioannis Antonoglou. They build open-weight AI models that automate software development, and their first model, called Beam, launched just last week on October 5.

And Nvidia is already deep in. The FT says Nvidia has invested $800 million, and Reflection is raising at a pre-money valuation of $25 billion. The talks are early, and one option is an acqui-hire, where Nvidia hires the team and licenses the technology instead of buying the company outright — which could dodge a lengthy regulatory review.

Why does a chip company care about open-weight models? Two reasons. First, Nvidia wants to encourage a market for open models that businesses can customize and run on their own hardware, which can be much cheaper than the most advanced versions of Claude or ChatGPT. Second, the White House wants U.S. open-weight models to rival China’s, and Reflection is being talked about as the American answer to DeepSeek.

Put the two stories together — Amazon chasing Decart and Nvidia circling Reflection — and the AI infrastructure land grab is clearly still accelerating.

The October Funding Scoreboard

October is the month the AI money spigot went into overdrive:

  • OpenAI is still hunting for at least $30 billion in fresh funding at a valuation of about $1.4 trillion. Bloomberg reported last week that OpenAI is in talks with several major United Arab Emirates funds, including Abu Dhabi’s MGX, to anchor the round, with the Gulf funds discussing a combined investment of up to $10 billion. BlackRock is also in talks to join. In an unusual move, OpenAI set the valuation itself, with no lead investor negotiating terms. This comes as OpenAI pushed its IPO to next year to focus on safety work.
  • Moonshot AI, the company behind the Kimi chatbot, has closed its final private round at a valuation of about $50 billion, up from $31.5 billion this summer, according to Bloomberg. And Moonshot is heading for a Hong Kong IPO early next year that could raise up to $5 billion. Their revenue climbed from $300 million in June to about $1 billion now, heading for $2 billion by December, driven by the Kimi K3 open model.
  • And Anthropic is reportedly eyeing a November IPO near $2 trillion.

So $30 billion plus at $1.4 trillion for OpenAI, $50 billion for Moonshot, and Anthropic reportedly near $2 trillion. The frenzy isn’t slowing down. If anything, October is the month it went into overdrive.

Japan Declares a “State of Emergency in Cyberspace”

Now a darker story. Japan has declared what officials are calling a state of emergency in cyberspace, after cyberattacks hammered more than 20 major companies.

Originally reported by the Financial Times. The victims include convenience store chain Lawson, car rental operator Times Car, and rail operator JR East, which lost 1.67 million customer email addresses.

And South Korea is getting hit too. Reuters reports nine South Korean banks and two mega-churches are probing cyberattacks that may have involved AI tools. Here’s the kicker: CrowdStrike says a suspected China-based hacker behind the bank attacks used a Chinese-developed AI agent plus Anthropic’s Claude Code to steal data from tens of thousands of South Korean bank customers — and probably could not have pulled it off without AI assistance.

Security experts say AI is changing the math for attackers. Trend AI data shared with Reuters shows Japan recorded more cybersecurity incidents in the first nine months of this year than in all of 2025. One expert told the FT that AI lets hackers do the reconnaissance that used to take weeks, so smaller targets that were never worth the effort are now fair game.

The language barrier is gone too. Experts compared the attacks to carpet bombing, and Japan’s financial regulator has reportedly asked banks to only accept chipped identity documents after Times Car leaked troves of driver’s license images.

A sobering reminder that the same AI capabilities we celebrate are also making cybercrime faster, cheaper, and harder to detect.

Quick Hits

India’s HCLTech signed a pact with Anthropic today to accelerate enterprise adoption of agentic AI, focused on safety, cybersecurity, and governance. Reuters reports HCLTech will build a dedicated business unit with a center of excellence embedding Claude, Cowork, and Claude Code. HCLTech is India’s third-largest IT services firm, and they report second quarter results later today.

And one more market mover. IREN shares gained in premarket trading after The Information reported the cloud computing company struck a deal with Anthropic.

Sources

  • Amazon in talks to buy AI startup Decart for around $7 billion: Wall Street Journal
  • Nvidia in talks to invest further in Reflection AI or buy it, FT reports: Reuters
  • South Korea and Japan buffeted by hacks as AI lowers bar for cybercriminals: Reuters
  • India’s HCLTech signs a pact with Anthropic, to set up business unit: Reuters

Adapted from the October 12, 2026 episode of Daily Tech and AI.

Daily Tech and AI is AI-voiced and produced by Angelo Ruiz.

Angelo Ruiz Avatar

More about Angelo · LinkedIn

Leave a Reply

Your email address will not be published. Required fields are marked *