Anthropic, we have questions

1. Key Themes

Theme: The Rise of "Services-as-Software" as an Investment Category

Bain Capital Ventures is underwriting an entirely new category: AI-native companies that replace labor/services rather than sell tools.

"The firm has been swapping software for something venture capital has never underwritten: AI-native companies that sell the work itself rather than the technology behind it."

Theme: Anthropic's Revenue Growth Is Explosive but Concentration Risk Looms

Anthropic's run-rate revenue exploded nearly 7x in seven months, but the sustainability of that growth is unproven.

"Anthropic's run-rate revenue rose from about $9 billion to more than $65 billion in seven months." "More than 1,000 customers were spending at least $1 million annualized as of April. That sets a floor of $1 billion and says nothing about distribution."

Theme: True Margins Are Murkier Than Headline Numbers Suggest

There's a large gap between the marketed gross margin and the real, fully-loaded margin once model-training and revenue-sharing costs are included.

"Investor materials reportedly cite gross margins above 80% before partner revenue sharing and model-training costs. After those costs, we estimate 44%."

Theme: Compute Contract Structure Creates Hidden Duration Risk

Anthropic's asset-light compute strategy (contracts vs. owned infrastructure) creates a mismatch risk between customer churn speed and Anthropic's payment obligations.

"Anthropic secured diversified compute through contracts rather than owning data centers, power, or silicon. The risk is duration: if customers cut spending within months while Anthropic owes years of payments, a slowdown hits cash flow before costs can adjust."

Theme: Safety Commitments Are Becoming Competitive Table Stakes, Not Differentiators

Both leading labs converged on similar safety/transparency commitments almost simultaneously, suggesting this is becoming industry standard rather than a unique value proposition.

"On September 12, Dario Amodei committed to giving outside evaluators employee-like access to Anthropic's training pipelines and the right to publish findings, subject to limited redactions... OpenAI matched the evaluator commitment the same day."

2. Contrarian Perspectives

PitchBook Values Anthropic Well Below the Reported IPO Target

Despite the extraordinary growth story, PitchBook's own valuation is meaningfully below what's reportedly being floated for the IPO — a notable gap for a company this hyped.

"We value Anthropic at roughly $1.4 trillion to $1.6 trillion on current evidence, below the $2 trillion reportedly floated for the IPO."

Safety Rhetoric Shouldn't Move the Valuation Needle

Contrary to the market narrative that Amodei's safety pledge signals risk or slower growth, PitchBook argues it's a non-event for valuation absent concrete filing disclosures.

"Amodei's letter doesn't affect our valuation. A discount would need the filing to show who can delay a release, what triggers it, and whether the growth forecast already assumes a slower pace."

Betting on Unprecedented Business Models Without Underwriting History

Bain's Matt Harris is explicitly comfortable underwriting AI-native "services" businesses with no track record or valuation framework — a break from traditional VC diligence norms.

"Harris said BCV isn't troubled that these AI-native businesses have no underwriting precedent. He argues that these new businesses should be valued based on their ability to transform services businesses."

3. Companies Identified

• Anthropic — AI model developer preparing for IPO. Mentioned as the subject of PitchBook's pre-IPO diagnostic analysis. "Our latest research outlines what we'll look for in the company's IPO filing."

• Bain Capital Ventures (BCV) — VC firm, closed 11th flagship fund. Mentioned for pioneering a new "post-AGI" investment thesis. "Bain Capital Ventures closed its 11th flagship fund at $1.6 billion on Wednesday, betting big on a 'post-AGI world.'"

• Reserv — AI-native insurance claims administration startup, incubated by BCV. Cited as the proof-of-concept for the "sell the work" thesis. "The company hit $100 million in revenue in its second full year... raised a $125 million Series C."

• Legora, Crosby, Norm AI — AI-native legal services/agentic law firm startups. Mentioned as BCV portfolio examples of the services-as-software model.

• OpenAI — Competing AI lab. Mentioned for matching Anthropic's safety commitment and separately reported to be raising at a massive valuation. "OpenAI matched the evaluator commitment the same day." / "OpenAI is in talks to raise a new round at a $1.2 trillion valuation, Bloomberg reported."

4. People Identified

• Dario Amodei — CEO of Anthropic. Notable for a public commitment on AI safety transparency that set an industry benchmark. "Dario Amodei committed to giving outside evaluators employee-like access to Anthropic's training pipelines and the right to publish findings... He also called for a slower industry cadence."

• Matt Harris — Partner at Bain Capital Ventures. Notable for architecting and defending the firm's contrarian "post-AGI" investment thesis. "'What's it all worth in the end? We've ticked a couple of key boxes, but that's the box that I think only time will tell.'"

5. Operating Insights

1. For AI startups seeking investment: framing matters. Positioning your company as replacing an entire labor/services category (not just selling software into it) is attracting dedicated new capital pools like BCV's $1.6B fund — a distinct pitch from traditional SaaS framing.

2. For companies raising at scale, expect investors to scrutinize the "real" margin, not the marketed one. Anthropic's gap between reported (80%+) and effective (44%) margin illustrates that operators should proactively define cost of revenue clearly, since sophisticated investors will reconcile the two regardless.

3. Compute/vendor contract structuring is now a scrutinized risk factor, not just a cost line. Founders in AI infrastructure-heavy businesses should be prepared to explain contract duration and cancellation terms relative to customer churn risk — this is becoming a standard diligence question.

6. Overlooked Insights

1. The PE Barometer signals a broadly favorable return environment right now, which is easy to miss amid the AI headlines but relevant for portfolio-level allocation decisions: "The PE Barometer signals a strong underlying return environment, with a desmoothed return nowcasts of 6.8%... driven by gains in small-cap and value equities, falling implied equity volatility, low financial stress, and tight leveraged loan spreads."

2. Fusion energy startups have raised massive capital despite zero net-positive results to date — a potential bubble/overhang signal buried in the "Side Letters" section: "Fusion startups have raised $3.8 billion so far this year, but none have yet produced more electricity than their facilities require to operate."