Anthropic mentions "compute" 49 times in its 42-minute IPO roadshow video — more than any theme except its own products. That single count may say more about the company's public offering than the video's glossy graphics do.

Chief Executive Dario Amodei opens the pitch, released on Tuesday, and needs just over 30 seconds to raise the risk of AI-enabled bioweapons. He calls AI "the most promising technology in human history" and says it is "capturing the essence of cognition." The framing is deliberate: Anthropic wants Wall Street to see its safety mission as a feature of the investment, not a cost.

Business Insider ran a transcript of the video through ChatGPT and Claude separately, counting word frequencies and themes. The results sketch a company whose stated values and financial logic sit in a specific order — and the order matters.

What do the word counts actually show?

Strip out the expected references to "Claude" (65 mentions) and "model(s)" (58), and compute dominates. The full top ten:

• Claude — 65
• Model(s) — 58
• Compute/computing/computation — 49
• Customer(s) — 42
• Work/working/workers — 39
• AI — 39
• Frontier — 32
• Anthropic — 32
• Capability/capabilities/capable — 27
• Safe/safety/safeguards — 27

Compute outranks safety nearly two to one in raw word count. That gap is not accidental. OpenAI CEO Sam Altman likes to say compute is destiny; for Anthropic, compute is arguably the reason the company is going public at all.

Its amended S-1 filing shows Anthropic is set to spend $895 billion on compute and infrastructure, with nearly half of that expected to be spent by the end of 2030. (An S-1 is the registration document companies file with US regulators before listing.) Data centers and the chips that fill them are expensive, and a public listing is how Anthropic plans to help finance them.

The video's production follows a familiar IPO formula. Brightly colored graphics — a staple of Anthropic's published research — alternate with testimonials from top customers, including Bristol Myers Squibb and Accenture. The customer count of 42 mentions signals where the revenue story sits.

Does safety still anchor the pitch?

Anthropic has built its brand on the claim that it takes AI risk more seriously than its rivals. By theme, safety-related language becomes more prevalent than the raw word count suggests — but it still does not displace compute. The ranked themes:

• Models and training — 83
• Compute and infrastructure — 81
• Financial performance and growth — 77
• Efficiency and value — 69
• Customers and enterprise — 62
• Safety, alignment, trust, security, and risk — 55
• Research and discovery — 42
• Coding and software — 41
• Products and platforms — 34

Safety ranks sixth of nine themes. The 27 direct mentions of safety words also understate the theme in one direction: they exclude Amodei's bioweapons reference, which carries the safety message without using the vocabulary.

Anthropic anticipated the skepticism. The roadshow narrator states plainly: "We're finding safety doesn't hold us back." The video continues: "In fact, it's because we can trust Claude's work that we can take its discoveries and apply them in the real world in ways that will benefit everyone."

That is the pitch's core inversion — safety reframed as an accelerant rather than a brake. Whether investors accept the premise, or simply price the $895 billion compute bill, is the question the listing will answer.

What does this change for investors?

The roadshow is part of an expected pre-IPO blitz, and the word counts tell prospective shareholders what the company itself emphasizes when money is on the line. The hierarchy is clear: product, then compute, then growth, then customers, then safety.

Notably absent from the top rankings: any single customer segment dominating, and any language suggesting Anthropic sees safety as a constraint to be managed. What the video does not say is equally telling — it offers no explicit reconciliation between the trillion-scale infrastructure commitment and the company's risk-aware identity, beyond the narrator's assurance.

For a company that has marketed itself as the responsible AI lab, the arithmetic of its own pitch places that identity behind the machinery it needs to build. The offering, and the market's reception of it, will test whether both stories can be sold at once.