Meet 4 engineering leaders whose budgets explain how Anthropic has gotten so rich
According to Business Insider, Anthropic is aiming for a $1.4 trillion valuation in its IPO, driven largely by heavy spending from business customers and the popularity of Claude Code. The prospectus shows 11,900 business customers spent at least $100,000 over the past year and 250 spent $10 million or more, while the top 5% of Claude Code users average about $1,500 monthly versus under $75 for other business users. Engineering leaders, including Grafana Labs CTO Tom Wilkie, say token bills reach tens of millions annually, with roughly two-thirds going to Claude and the rest mostly to OpenAI's Codex.
Business Insider reports that Anthropic is pursuing an initial public offering this week at a targeted valuation of $1.4 trillion, and the article argues that the figure is plausible largely because of a base of unusually heavy-spending enterprise customers. The piece profiles engineering leaders at companies that have adopted AI-assisted coding aggressively and, for the most part, chosen Claude over rival options. Only the opening portion of the story was available in the text provided, so this summary covers the framing, the prospectus data, and the first profiled executive.
The spending picture from the prospectus
The article cites Anthropic's IPO prospectus to show how concentrated and intense usage has become. Business accounts that do not use Claude Code spend, on average, less than $75 per month. By contrast, the top 5% of users of the coding tool average roughly $1,500 per month. That gap suggests agentic coding is a very different consumption category from conventional chat or API usage, with token consumption scaling sharply when AI agents write, test, and revise code on engineers' behalf.
At the account level, the prospectus reportedly lists 11,900 business customers who each spent at least $100,000 over the previous twelve months, and 250 customers who spent $10 million or more. Those numbers indicate that a relatively small cohort of large enterprises and AI-forward startups accounts for a meaningful share of revenue, which is both a strength, given the depth of commitment, and a potential concentration risk for a newly public company.
Token-maxxing and rising bills
The article situates the current spending against the so-called token-maxxing trend of earlier in the year, when companies encouraged engineers to consume as many tokens as possible in pursuit of productivity gains. Several engineering leaders told Business Insider that, months later, individual projects can still generate Claude bills in the tens of thousands of dollars. Some managers also now accept that certain engineers will occasionally spend more on tokens than they earn in salary, a notable shift in how companies think about the cost of a developer. The piece also links to related coverage on AI companies raising prices and imposing internal token limits, and on OpenAI researchers' token spending, hinting that cost governance is becoming a broader industry theme.
Grafana Labs: tens of millions and no urgency to cap
The first profiled leader is Tom Wilkie, chief technology officer at Grafana Labs, an observability company last valued at $9 billion. Wilkie said the company's annual token spend has climbed into the tens of millions of dollars. Roughly two-thirds goes to Claude, with most of the remainder directed to OpenAI's Codex. He stated that the company is not currently worried about the spend or about controlling it. That posture is telling: rather than treating AI coding costs as a line item to minimize, Grafana appears to view them as an investment whose returns justify the outlay, at least for now. The article indicates that one of Wilkie's teams spent tens of thousands on a single effort, though the remainder of that passage and the other three leaders' profiles were not included in the available text.
Analysis and implications
Several takeaways emerge for the AI tools market. First, Claude Code appears to be a primary revenue engine for Anthropic, with usage-based pricing amplifying revenue from power users. Second, the split at Grafana, with Claude taking the largest share and OpenAI's Codex taking much of the rest, shows that large engineering organizations are multi-sourcing even while leaning toward one vendor, which keeps competitive pressure on both. Third, the willingness of leaders to tolerate spending that rivals or exceeds salaries signals that perceived productivity gains are still outpacing cost concerns, though the mention of price increases and internal limits elsewhere suggests that tolerance may not be permanent.
For investors weighing the IPO, the data points highlight both the upside of deep enterprise adoption and the risks tied to customer concentration, pricing sensitivity, and competition from OpenAI and others. For developers and tool buyers, the story reinforces that agentic coding workflows are becoming a significant budget category that will require governance, measurement, and vendor strategy as usage scales.