Welcome to AI Decoded, Fast Company’s weekly newsletter that breaks down the most important news in the world of AI. I’m Mark Sullivan, a senior writer at Fast Company, covering emerging tech, AI, and tech policy.
This week, I’m focusing on OpenAI’s gigantic new funding round and valuation. I also look at a recent leak around Anthropic’s models, and at backlash to ads placed in GitHub Copilot.
Sign up to receive this newsletter every week via email here. And if you have comments on this issue and/or ideas for future ones, drop me a line at sullivan@fastcompany.com, and follow me on X (formerly Twitter) @thesullivan.
OpenAI closes $122 billion funding round at $852 billion valuation
OpenAI has closed what may be the largest private funding round ever, raising $122 billion (well more than the $110 billion target it announced in February.) The company is now valued at $852 billion, making it more valuable than most publicly traded companies.
Investors include some of the biggest (and richest) names in big tech and big finance. Amazon reportedly committed around $50 billion, Nvidia put in $30 billion, and SoftBank contributed roughly $30 billion. Andreessen Horowitz, D.E. Shaw Ventures, TPG, T. Rowe Price, and Abu Dhabi’s MGX also participated, along with Microsoft and other existing backers.
The AI company is clearly laying the groundwork for a 2026 IPO, and it may come sooner rather than later, depending on market conditions. Currently, about a third of the total value of the stock market is being buoyed up by faith that the AI models and agents will transform the way business operates.
OpenAI says it’s now generating about $2 billion per month in revenue, mostly from ChatGPT subscriptions, enterprise contracts, and API usage. (OpenAI was quick to point out that 40% of its revenue comes from enterprise sales, hoping to dispel the narrative that rival Anthropic is capturing the enterprise market.) Still, the company remains unprofitable, burning through cash on computing power, data centers, chips, and talent. OpenAI will use the additional funding to continue scaling up its models so that they can support the advanced agentic systems that are capable of autonomously handling challenging business tasks.
Analysts’ reaction to the gigantic new round has been mixed. Winning the race to the smartest AI models depends in large part on access to abundant computing power, so some analysts read Nvidia’s and Amazon’s involvement in the round as a sign that OpenAI may have the inside track in that regard. Bernstein and others said the new funding is good news for Microsoft, Amazon, and Oracle, since it shows that OpenAI actually has the funds to pay for all the infrastructure capacity it’s contracting.
Others are nagged by doubts about OpenAI’s potential for becoming profitable anytime soon. Developing and deploying AI models is a hugely expensive business, and the biggest line item is computing costs. HSBC and Deutsche Bank analysts believe that OpenAI and other general-purpose model developers may yet need to spend hundreds of billions more for compute alone before they reach profitability, while in the meantime reporting unprecedented losses.
The current AI boom is often compared to the dot-com bonanza of the late 1990s, in which tech companies raked in large investments despite little evidence of a business model. Some suspect that the same “irrational exuberance” may be driving the record-breaking investments in AI companies. OpenAI’s new $852 billion valuation won’t help matters. A Bloomberg report says demand for OpenAI shares has cooled on secondary markets, while demand for shares of Anthropic, which has a far lower valuation of $380 billion, has been strong.
So the new funding round should position OpenAI as the frontrunner in the AI race for now, but everybody will be watching for signs that the AI startup is cranking up profits to offset its frightening cash burn.
Why the leak of Anthropic model information matters
Another day, another Anthropic Claude Code source material leak. A debugging file was accidentally bundled into a routine Claude Code update and pushed to the public code package repository, and as a result we got information about the tool’s architecture and some performance data. The leak revealed no customer information or anything about the code or weights of Anthropic’s model, but it did provide a look at how Anthropic is scaling its Claude models, including new features that appear to be fully built and ready for release.
Anthropic is steadily turning Claude into a more proactive assistant. The company appears to be developing features that let users schedule tasks to run automatically at set times, alongside expanding voice interaction so Claude can be spoken to and respond out loud. At the same time, it is improving memory capabilities, enabling future models to revisit past chats, learn from them, and apply those insights in new sessions. All of this points toward a “persistent assistant” that runs in the background even without an active session, with control extending across devices.
Based on the direction of previous model releases, and on the agentic features already in Claude Code, the product road map direction revealed in the leak aren’t terribly surprising. They are all targeted toward getting Claude models ready to perform key tasks within the business operations of large companies. Anthropic’s continued growth in the enterprise market is crucial to the IPO the company hopes to hold this year.
What’s of greater concern is what the leak says about Anthropic’s own operational security. The company was founded on the idea of being a “safer” AI provider. To enterprises that likely means not only that the company’s models aren’t easily used to do harm, but also that they can be entrusted with sensitive proprietary information. Some may wonder if Anthropic can deliver on that promise when it can’t even keep its own product plans private.
What GitHub Copilot ad insertion goof says about AI monetization
GitHub’s AI coding assistant Copilot recently injected some promotional text into more than 11,000 pull requests (code change proposals) across thousands of code repositories. Developers were not amused.
The issue surfaced when developer Zach Manson noticed that a teammate had asked Copilot to correct a typo in a pull request description. Copilot fixed the typo but also quietly inserted a promotional message for itself and Raycast, a productivity app with a Copilot integration. “Quickly spin up Copilot coding agent tasks from anywhere on your macOS or Windows machine with Raycast,” it read. The Register found the same statement in 11,400 pull requests.
GitHub, which is owned by Microsoft, had been adding “helpful tips” to pull requests created by Copilot for some time. The company told The Register that a bug caused the tips to bleed into human-created PRs whenever someone mentioned Copilot. Now GitHub has removed agent tips from pull request comments.
Was GitHub soft-launching a form of ad monetization? Maybe. But the backlash may show how sensitive developers are to having ads, no matter how benign, injected into their professional work. “This is horrific,” Manson wrote. “I knew this kind of bullshit would happen eventually, but I didn’t expect it so soon.”
More AI coverage from Fast Company:
- The Navy’s AI bet to fix its submarine bottleneck
- Intuit thinks it’s found your company’s next CFO: AI
- 4 AI chatbots tried to fact-check Rubio on Iran. They couldn’t agree
- When to use human English vs. AI English
Want exclusive reporting and trend analysis on technology, business innovation, future of work, and design? Sign up for Fast Company Premium.