- Anthropic
- 1. Anthropic expects to turn a profit faster than OpenAI
- 2. Anthropic invests $50B in AI infrastructure
- OpenAI
- 3. SoftBank divests Nvidia to fund OpenAI bet
- 4. Blue Owl invests $3B in OpenAI's Stargate data center
- Google
- 5. Microsoft, Google to invest $16B+ on AI infrastructure in Europe
- 6. Google launches Private AI Compute for secure cloud-based queries
- In other news:
- 2 new AI tools to try
- Recommended reading
The CFO AI Digest: November 12, 2025
This week’s stories focus on the massive capital flowing into AI infrastructure, and explore the different paths the two largest AI companies in the world have taken to profitability. Plus, Google makes a major privacy upgrade.
Let’s get into it:
Anthropic
1. Anthropic expects to turn a profit faster than OpenAI
Anthropic is projected to break even by 2028. That same year, OpenAI expects $74B in operating losses, according to financial documents obtained by The Wall Street Journal. Both companies are burning about 70% of revenue in 2025, but by 2027, Anthropic expects to cut its burn rate to 9% while OpenAI’s is expected to stay at 57%. Anthropic generates 80% of its revenue from corporate customers, and is growing through demand for Claude, especially Claude Code. It is avoiding compute-heavy areas like image and video generation. OpenAI, in contrast, is spending heavily on chips and data centers, committing up to $1.4 trillion over the next eight years on deals and stock-based compensation for top researchers.
CFO takeaway: Anthropic’s focus on paid enterprise API customers is paying off. As one product leader noted on Substack, Anthropic is targeting customers with a higher willingness to pay than ChatGPT, while serving far fewer nonpaying users than ChatGPT. Meanwhile, OpenAI is pursuing a path to build a multi-trillion-dollar company and become the AI company of choice through aggressive spending on infrastructure, consumer, and advertising.
2. Anthropic invests $50B in AI infrastructure
Anthropic announced a $50 billion investment to build custom data centers in Texas and New York in partnership with Fluidstack, with more sites to follow. The infrastructure is designed to support frontier AI research and growing enterprise demand for Claude. The buildout will create 800 permanent jobs and aligns with the Trump administration’s AI Action Plan to strengthen domestic AI leadership.
CFO takeaway: This investment follows Anthropic’s revenue projections we just discussed, which emphasize a more measured approach relative to OpenAI’s aggressive infrastructure spending. It’s a notable shift, and a signal that even the more profit-conscious players in AI now see owning infrastructure as critical to staying competitive at the frontier.
OpenAI
3. SoftBank divests Nvidia to fund OpenAI bet
SoftBank sold its Nvidia stake for $5.8B to help fund its $30B commitment to OpenAI. SoftBank fulfilled $7.5B of that commitment earlier this year, resulting in quarterly profits of $16.2B — more than double 2024’s results — driven by OpenAI's valuation. SoftBank CEO Masayoshi Son is aggressively pursuing AI investments, raising over $11B from asset sales and taking on new debt, according to The Wall Street Journal.
CFO takeaway: SoftBank's aggressive capital repositioning is Son’s bet on AI infrastructure maturity. As SoftBank's former CFO explained a few months ago, today’s AI investments are different from the 2016 $100B Vision Fund, when there were not AI-related opportunities of this size to invest in. "It's now realistic to think in terms of investing $100B and potentially a lot more in AI infrastructure,” the former CFO said.
4. Blue Owl invests $3B in OpenAI's Stargate data center
Private credit giant Blue Owl Capital is investing approximately $3B in exchange for equity in a New Mexico data center that's part of OpenAI's Stargate project, with banks arranging an additional $18B in loans. The 4.5-gigawatt facility is one of the largest data centers being developed. Blue Owl reports a pipeline of over $100B in potential data center deals and has already invested in other Stargate and Meta facilities, according to The Information.
CFO takeaway: By taking the risky $3B equity slice in the $21B data center, Blue Owl enables banks to participate in the safer debt layers while capturing high returns. This structure reflects how private credit can unlock mega-scale AI infrastructure that banks can't finance alone, though few firms have the size or mandate to play this role.
5. Microsoft, Google to invest $16B+ on AI infrastructure in Europe
Microsoft will invest over $10B in a data center hub in Portugal that it plans to build with NVIDIA, AI infrastructure startup Nscale Global Holdings, and data center builder Start Campus. The center is expected to have one of the strongest AI computing capacities in Europe. Google will inject $6.4B into AI infrastructure in Germany through 2029, including a new data center in Dietzenbach, an existing facility in Hanau, and office expansions in Berlin, Frankfurt, and Munich.
CFO takeaway: We previously explained that the EU’s AI Act requires companies to meet strict standards for AI models. Microsoft and Google’s new data centers are designed to handle these regulatory requirements, keeping data local and building tools for compliance. For CFOs, this shifts regulatory complexity to cloud providers and makes Europe a more viable base for scaling AI.
6. Google launches Private AI Compute for secure cloud-based queries
Google launched Private AI Compute, a secure space that lets its Gemini models run in the cloud without exposing user data — not even to Google. Private AI Compute runs on one seamless Google stack powered by Tensor Processing Units. It is already enhancing use cases for on-device features such as Magic Cue, which now delivers more timely suggestions while retaining privacy protection.
CFO takeaway: Google’s launch mirrors Apple’s Private Cloud Compute and builds on Google’s broader push into privacy-first AI, including Vault Gemma. The common thread: cloud AI architectures are being reengineered for sensitive data use. For finance teams, this is yet another step to safely apply LLMs to internal data in regulated, high-stakes environments.
In other news:
- Meta chief AI scientist Yann LeCun plans to exit and launch own start-up (Financial Times)
- AMD predicts accelerating sales growth on data center demand (Bloomberg)
