The CFO AI Digest: November 7, 2025

This week, we track the accelerating enterprise adoption curve, from OpenAI crossing 1 million business customers to Anthropic projecting $70 billion in revenue by 2028. And Google, Apple, and Snap each lean into their distribution advantages, embedding models into products already used by billions.

Let’s get into it:

OpenAI

1. OpenAI reaches 1 million business customers, reports 75% enterprise ROI

OpenAI announced that 1 million business customers now use its platform, either through ChatGPT for Work or its developer platform. The company now has more than 7 million ChatGPT for Work seats, with ChatGPT Enterprise seats growing 9x year-over-year. To support claims of business value from AI deployments, OpenAI cited a Wharton study showing 75% of enterprises report positive ROI on GenAI tools. The company also announced a partnership with Databricks to bring OpenAI models to enterprise data platforms.

CFO takeaway: OpenAI's enterprise traction and reported ROI metrics validate the business case for significant AI spending. Fidji Simo's observation that "the most successful businesses use AI to multiply, not just cut" reinforces the idea that AI investments should be evaluated not just on cost savings but on revenue expansion and competitive positioning through faster product launches and better intelligence. This creates urgency for CFOs: competitors that secure enterprise agreements now may build operational advantages that become costly to replicate later.

Read more

Google

2. Google launches Ironwood TPU, Anthropic to deploy up to 1 million chips

Google will make Ironwood, its seventh-generation Tensor Processing Unit, widely available in the coming weeks after initial testing began in April. TPUs are Google's custom-built chips for AI workloads, serving as an alternative to standard GPUs. Ironwood can link up to 9,216 chips in a single configuration and is more than four times faster than the previous generation. Anthropic plans to use up to 1 million of the new TPUs to run its Claude model.

CFO takeaway: Google's Ironwood TPU availability adds another data point to the chip diversification trend we have observed across AI leaders. After OpenAI's 6GW AMD commitment, Anthropic's plan to deploy up to 1 million Google TPUs confirms that multi-vendor compute strategies are becoming favorable. Nvidia’s pricing power, and the premium investors assign to its dominance, could face increasing pressure.

Read more

3. Apple taps Google’s Gemini to power Siri 2.0 in $1B annual deal

Apple is finalizing a $1 billion annual deal with Google to license its 1.2 trillion parameter Gemini AI model to power a major overhaul of Siri. The upgrade will modernize Siri’s reasoning and task-planning systems, with much of the processing moving to Apple’s Private Cloud Compute servers for greater speed and privacy. Gemini, eight times more complex than Apple’s current 150 billion-parameter model, will serve as the backbone for Siri’s new summarizer and planner features when the assistant launches in 2026.

CFO takeaway: Apple’s Gemini deal mirrors OpenAI’s alignment with Microsoft in its goal: accelerating AI capability through collaboration rather than full vertical control. For CFOs, these alliances might be emerging as the fastest route to durable AI value.

Read more

4. Google adds Gemini-powered conversational navigation to Maps

Google Maps is introducing Gemini-powered conversational navigation that lets drivers ask multi-step questions and get answers using only voice. Users can ask "Is there an affordable vegan restaurant nearby?, "check parking conditions,” then say "Take me there.” The update includes landmark-based navigation using recognizable buildings instead of distance measurements ("turn right past the corner restaurant"), proactive traffic alerts before trips begin, and a camera-based Lens feature that identifies places through voice interaction. The feature is rolling out on Android and iOS where Gemini is available.

CFO takeaway: In a previous Digest, we noted that Google’s enterprise AI strategy is powered by distribution. The Maps rollout extends that playbook to consumers: using existing products to drive AI adoption. Unlike Amazon’s Alexa+ push, which relies on dedicated hardware, Google is layering Gemini onto an already ubiquitous app, deepening engagement without the friction of new tools or platforms.

Read more

Anthropic

5. Anthropic projects $70B Revenue by 2028, eyes $300-400B valuation

Anthropic projects up to $70 billion in revenue by 2028, up from approximately $5 billion expected in 2025. The company raised its most optimistic revenue forecasts by 13 to 28% over the next three years (2025–27) compared to earlier projections. It anticipates becoming cash flow positive by 2027 and generating $17 billion in free cash flow by 2028 — a faster timeline than OpenAI, which doesn't expect positive cash flow until 2030. The company's business focuses primarily on API sales to enterprises, which are projected to account for over 80% of revenue through 2028. Anthropic raised $13 billion in September at a $170 billion valuation and is considering a future funding round targeting a $300 to 400 billion valuation.

CFO takeaway: Anthropic’s updated forecast reinforces that the company’s aggressive enterprise push is becoming its growth engine. With API-based enterprise revenue already exceeding 80% of total sales, and expected to remain dominant through 2028, the AI behemoth is positioning itself as an enterprise infrastructure provider. Its recurring, usage-based model, where revenue scales with customer adoption, offers the predictability and margin leverage investors increasingly favor.

Read more

Perplexity

5. Snap partners with Perplexity in $400M deal to bring AI search into Snapchat

Snap Inc. and Perplexity announced a partnership to integrate Perplexity's AI-powered answer engine directly into Snapchat starting in 2026. Perplexity will pay Snap $400 million over one year through a combination of cash and equity. The integration will allow Snapchat's 943 million monthly active users to ask questions and receive conversational answers with verifiable sources within the Snapchat interface.

CFO takeaway: Snap’s partnership with Perplexity and the 9% jump in its stock price following the announcement underscore how markets are rewarding AI-led revenue diversification amid a pullback in digital ad spending. The move also reflects that consumer platforms are evolving into AI distribution ecosystems, where partnerships can drive long-term value creation.

Read more

Cognition

6. Cognition launches Windsurf Codemaps to visualize and navigate complex codebases

Cognition announced Windsurf Codemaps, an AI tool that creates annotated maps of codebases to help engineers understand code structure before modifying it. Engineers prompt the tool with questions about specific functionality and it identifies relevant code across the codebase with clickable links to lines and explanations of how they relate to each other. The maps can be referenced within Cascade, Cognition's coding agent, to improve agent performance on specific tasks. Codemaps is powered by SWE-1.5 and Claude Sonnet 4.5.

CFO takeaway: Cognition is targeting the root cause of AI slop in engineering: poor understanding of code. It calls this the largest constraint in engineers’ ability to code. Better code comprehension means fewer costly errors and less rework, leading to shorter delivery cycles and higher return on existing software investments.

Read more

See you next week.

2 new AI tools to try

1. Kosmos
2. Windsurf Codemaps

Recommended CFO reading

1. Don’t blame AI for your job woes (The Economist)
2. Will quantum be bigger than AI? (BBC)