Skip to content
LIVE
The Executives BriefThe Executives Brief

Anthropic's $1.4T IPO rests on engineers who spend six figures a month on tokens

Four engineering leaders at Grafana Labs, Theory Ventures, Menos AI and Harness explain the Claude budgets powering Anthropic's pre-IPO surge, and why they still call it cheaper than hiring.

ByLama Al-RashidTechnology Correspondent, The Executives Brief
·4 min read
Anthropic's $1.4T IPO rests on engineers who spend six figures a month on tokens
Executive summary

Bryan Bischof, who runs AI at Theory Ventures, has personally spent six figures on Claude tokens in a single month, while Grafana Labs CTO Tom Wilkie says his company's token bills have climbed into the tens of millions a year, and Harness engineers burned 265 billion tokens in one recent week. Those budgets are the demand curve behind Anthropic's targeted $1.4 trillion IPO valuation, which means every CFO now has to decide whether token spend is a software line item or a headcount substitute.

Bryan Bischof, who runs AI at the San Francisco venture firm Theory Ventures, has personally racked up six figures of token spend in a single month, mostly on Anthropic's Claude models. He told Business Insider he could easily crack $1 million if he spent more time running simulations. That is one person, at one firm, leading a team of six engineers. Multiply that appetite across thousands of businesses and you have the engine behind Anthropic's initial public offering, which the company is targeting at a $1.4 trillion valuation this week. Claude Code, Anthropic's coding tool, is the juggernaut product behind the pre-IPO growth, and the customers Business Insider spoke with are not flinching at the bill.

The prospectus numbers explain the gap between casual users and the whales. Business accounts that did not use Claude Code spent less than $75 a month on average, while the top 5% of users of the coding tool averaged around $1,500 a month. Anthropic reported 11,900 business customers that had spent $100,000 over the previous year, and 250 that had spent $10 million or more. Those 250 accounts are what make a $1.4 trillion valuation arguable rather than absurd, and they are the cohort public-market investors will be underwriting.

Grafana Labs, the observability startup last valued at $9 billion, sits in that upper tier. Chief technology officer Tom Wilkie said the company's token bills have jumped into the tens of millions a year, with about two-thirds going to Claude and the rest mostly to OpenAI's Codex. "We're not worried about the spend," Wilkie said. "We're not worried about controlling it, just yet." The returns are visible. One of his teams spent tens of thousands of dollars on tokens in a single month to improve the performance of one Grafana service, which produced a six-figure cut to the company's Amazon Web Services bill. A non-engineer routed customer feedback to Anthropic's Fable model, spent more than $30,000 on tokens, and landed a cluster of successful changes in the product.

Grafana's guardrails are light by design. After a data analyst accidentally racked up tens of thousands of dollars in token spend by leaving a model running, the company started granting access to its most powerful AI models in 24-hour increments. Even so, Wilkie described the policy as "no questions asked." His reasoning: "We really want to be careful with the chilling effect of budgets and access control to models because you just don't know where the next big piece of innovation is going to come from."

Bischof's team at Theory Ventures builds custom software for the firm's investors, heads of talent and relationships, and intelligence leads. He and his engineers sometimes drop down to open-weight models, which are far cheaper than Anthropic's top-end options, but time is the scarcer resource. "If a very intelligent model is going to dramatically reduce the cycle time, then that's absolutely worth it," he said. Bischof said the firm may also invest in its own always-running AI agents, similar to Meta's Muse and OpenAI's Dots, a move that would convert token spend from a project expense into standing infrastructure.

At Menos AI, which runs AI-powered research software for investors, cofounder and chief AI officer Chris Yang said his engineering team uses Claude Code or Claude models through the coding tool Cursor. "Because Claude has been evolving so, so well in terms of coding capabilities, I don't even have much opportunity or willingness to try out Codex," he said. The investor software consumes around 90% of the company's token costs, routing user inquiries to different models depending on their difficulty. Yang calls himself one of the company's "hero users," spending $3,000 to $4,000 a month on Claude tokens while running several AI agents at once. His justification is a labor comparison: "We do spend a lot of expenses on the Claude Code use, but when we compare it to a human resources hire, it's still a big benefit. It's still much cheaper than human labor."

Harness, a software development startup worth $5.5 billion with more than 500 engineers, shows what scale looks like. Kiran Gangadharappa, a senior director of software engineering, said that in a recent week his staff used 265 billion AI tokens, with Claude Code and Cursor the most popular tools among his engineers. "We definitely see it improving and taking a bigger and bigger chunk of the manual work of writing code," he said. The biggest token-guzzlers are gigantic one-off tasks, like rebuilding an existing software architecture or conducting security research. Two levers pulled spending down: using Sonnet, one of Anthropic's cheaper models, and creating a soft $125-a-week budget for engineers. Gangadharappa expects the company has reached a "steady" level of AI expenses. "Agents never get tired," he said.

The strategic read for anyone running an engineering org is that token budgets are now a real line item with real governance questions, and these four leaders have landed on four different answers. Grafana buys optionality with 24-hour access windows and no questions asked. Theory Ventures buys cycle time and is weighing permanent agents. Menos AI buys against headcount. Harness buys with soft caps and cheaper models. What they share is a willingness to keep spending while the productivity math holds, which is precisely the assumption Anthropic's IPO pricing depends on. If that assumption cracks, or if a cheaper open-weight model closes the capability gap, the top 5% of users averaging $1,500 a month and the 250 accounts above $10 million become the most scrutinized cohort in the company's story. For now, the buyers are not blinking.

Executive ActionsLocked

This story's Key Insights and Take-aways are locked.

Create a free account to unlock Executive Actions for one credit.

Register to Unlock

Always free for Executives Club members. Join the Club

More in Technology