Broadcom lines up over $50B to finance OpenAI's custom chip

Share
Broadcom lines up over $50B to finance OpenAI's custom chip

The AI buildout is increasingly being paid for with borrowed money — and today's biggest example puts OpenAI's in-house silicon at the center of it. Also: Nous Research banks a $1.5 billion valuation, and an AI-built Adobe clone suite declares "software is over."

Broadcom has been working to arrange more than $50 billion in financing for OpenAI's custom AI chip, with Oracle in separate talks on financing for a large chip purchase, the Wall Street Journal reported. Apollo, Blackstone and Goldman Sachs are among the lenders in talks on what the Journal describes as blockbuster debt deals worth tens of billions of dollars apiece, aimed at closing by the end of the year. The report adds real detail to OpenAI's internal chip program — internally called Nexus and planned to scale to multiple gigawatts through 2029 — and describes Oracle's version as an off-balance-sheet entity that buys the chips and leases them back to the business. Bloomberg separately relayed the more-than-$50-billion figure, attributing it to the Journal, so the specific numbers remain single-newsroom until another outlet confirms them independently. What matters is the shape of it: this is now a debt market story, with GPUs as collateral, not an equity story. We saw the same machinery earlier this week when SpaceX reportedly seeks $40B, led by Apollo, to buy Nvidia chips — and Broadcom itself already lends Anthropic up to $42 billion against the chips it designs for them. When chip purchases are collateralized debt rather than cash burn, the AI race starts to look like an infrastructure finance race, with the lenders — not just the labs — deciding how fast it runs.


Nous Research confirmed a $90 million Series B at a $1.5 billion valuation, and is putting the money into an enterprise push called Hermes for Businesses. Robot Ventures led the round, with Nvidia, Union Square Ventures, Menlo Ventures, Samsung and 1789 Capital participating, bringing the three-year-old lab's total funding to $158 million. Nous says its open-source Hermes agent has been cloned more than 24 million times and drives roughly 2.5% of global AI token usage — a company estimate that, if it holds, makes this one of the most-deployed agent runtimes few enterprises have heard of. The Wall Street Journal reported the lab was at about $36 million in annualized revenue by mid-September and expects to pass $100 million before year end; the enterprise product brings SSO, auditing and workspace controls to companies that want to run the agent on their own data.


A single developer used Claude Opus 5.5 to build open-source, clean-room clones of seven Adobe apps — and declared "software is over" in the process. Brandon Thomas's Artcraft project shipped PhotoCraft, Filmcraft and five other Rust apps (with browser versions) that mimic the interfaces of Photoshop, Illustrator, Premiere, Lightroom, After Effects, InDesign and Acrobat Pro, all under MIT/Apache licenses. Thomas claims 100% feature parity "within a month," then walked it back to months rather than years — early testers on Hacker News have already cataloged the gaps, and the alpha state is obvious. The business model is the interesting part: the clones are free, funded by token sales for Artcraft's pre-existing visual AI model, and reverse engineering is generally legally sound — but if the trade dress looks too much like Adobe's, the copies could still land in court. Either way, AI-assisted clean-room reimplementation has gone from stunt to distribution.

What to watch: whether the lenders actually commit the more than $50 billion by year-end — and whether Adobe responds to clones that ship before they're finished.

Is debt-financed chip buying the sign of a healthy buildout or the top of a bubble? Tell us in the comments.

Read more

Trump's AI science summit banks $1B in Genesis Mission pledges

Trump's AI science summit banks $1B in Genesis Mission pledges

The White House is putting real money behind its AI-for-science push today, and the chipmaker feeding the buildout just posted another record quarter. The White House's AI science summit is bringing more than $1 billion in industry commitments from AMD, OpenAI, Anthropic and others for the Genesis Mission. According to Axios' exclusive preview, Trump will attend the summit — focused on using "super intelligence" to accelerate scientific breakthroughs — where the administration will announce th

Australia plans to regulate AI like banks and airlines

Australia plans to regulate AI like banks and airlines

Canberra is putting teeth around model accountability this morning, Google just turned its AI-content detector into a public utility, and Nvidia's dealmaking reportedly reached all the way to OpenRouter. Australia wants AI companies supervised the way banks and airlines are — and it plans to legislate that by 2027. Assistant Minister for Science, Technology and the Digital Economy Andrew Charlton laid out a "systems-based" regime in a speech at the Sydney Trust and Safety Festival on October 8

GPT-6 safety report: fewer refusals, more regressions

GPT-6 safety report: fewer refusals, more regressions

GPT-6 reaches ChatGPT's free tier today, an open-source agent just found its price tag, and world models drew a high-profile new contestant. GPT-6 is rolling out to free ChatGPT users today — and OpenAI's 24-page deployment safety report shows exactly what the model traded to become chattier. Plus, Pro, Business and Enterprise tiers started getting GPT-6 Sol on October 7; free and Go users get GPT-6 Luna from October 8, replacing the GPT-5.6 line in ChatGPT's main chat (Work and Codex models a

Stuart Russell: current training may make AI alignment impossible

Stuart Russell: current training may make AI alignment impossible

A safety-obsessed week just found its second heavyweight: after Hinton asked for an FDA of AI, the man who gave the field the word "alignment" says the current road may not get there at all — while memory markets show exactly where the AI money is going. Stuart Russell says he regrets coining the word "alignment," because the field read it as an engineering target — and he now thinks the way models are trained today may make avoiding misalignment impossible. The Berkeley professor made the cas