Key Takeaways

  • Anthropic confidentially filed for an IPO on June 1st, with some analysts like Gavin Baker predicting it would trade at $3 trillion if public today.
  • Chamath Palihapitiya warns of a coming “reckoning” for frontier AI companies, citing enterprise token costs “doubling every 45 days” with actual ROI for customers between zero and 2%.
  • Investor Brad Gerstner, a backer of both OpenAI and Anthropic, remains optimistic, pointing to Anthropic’s rumored trajectory of over $100 billion in revenue and the “intelligence is the largest TAM” argument.
  • The core tension for founders is between explosive top-line revenue fueled by early investment and the long-term sustainability of high AI spending for actual enterprise customers.
  • Post-IPO, founders and early investors should temper expectations; Gerstner notes that even these giants won't likely offer the “50 to 100% durable bounce” seen in past tech offerings.

The Disagreement: Growth at All Costs vs. the ROI Reckoning

The conversation around frontier AI companies like OpenAI and Anthropic often feels like two separate realities colliding. On one side stands Chamath Palihapitiya, the blunt venture capitalist, who sees a storm brewing. He points to unsustainable economics, particularly for enterprise customers adopting AI. “Right now, our token costs are doubling every 45 days,” Palihapitiya stated, revealing that the “actual ROI was somewhere between zero and 2%” for businesses utilizing these powerful models. For Palihapitiya, the question is simple: “But at some point, you'd have to be an idiot not to ask, well, who is paying you this? And can they sustain paying it to you?” He predicts a “reckoning” when the cost of running these models outstrips the demonstrable value they provide.

Contrasting this stark outlook is Brad Gerstner, an investor with significant stakes in both OpenAI and Anthropic. Gerstner remains bullish, arguing that the sheer market opportunity justifies the sky-high valuations. He cites incredible revenue growth, mentioning Anthropic’s rumored trajectory: “Anthropic's rumored to be, you know, trending over a hundred billion in revenue compared to the 35, right? If they exit the year at 100, that means their gap revenue next year could be well over a hundred.” For Gerstner, the narrative is about the “intelligence is the largest TAM” (Total Addressable Market), suggesting the potential for AI is so vast that current spending is merely a fraction of what’s to come. He believes there’s “a lot of meat on the bone” in these companies, even if he cautions against expecting massive IPO pops like the old days.

Who's Right (and When They're Wrong)

Both Palihapitiya and Gerstner offer critical insights for founders, and the truth likely sits in their uncomfortable middle ground. Palihapitiya is spot on about the coming pressure on enterprise ROI. As AI moves from novelty to utility, companies won't indefinitely absorb token costs that double every few weeks without clear, measurable business value. His warning is crucial for founders building AI products: if your business model relies on customers ignoring high operational costs, you're on thin ice.

Gerstner, conversely, correctly identifies the gargantuan market potential of artificial intelligence itself. The "largest TAM" argument holds weight because intelligence underpins nearly every industry. Early growth might be fueled by a mix of genuine value and speculative investment, but the long-term trajectory for truly impactful AI applications is massive. Gerstner's perspective is vital for founders looking to the horizon: don't let current inefficiencies blind you to the scale of the prize.

The challenge for today's founders is discerning between the hype-fueled spending of early adopters and the sustainable, value-driven adoption that will power long-term growth. The market will correct for the current ROI imbalances, but it won't diminish the fundamental shift AI represents.

What to Do With This

If you're building an AI product, stop optimizing for growth metrics alone. Pull your top 10 enterprise customers and deeply audit their actual, measurable ROI from your AI. Can they quantify savings, revenue gains, or efficiency boosts that far outweigh their token spend? If not, pivot your product or pricing model now before Palihapitiya's "reckoning" arrives.

If you're integrating AI into your existing business, don't just chase the latest model. Mandate a strict ROI analysis for every AI project. If your internal token costs are climbing without a proportional, tangible return (like Chamath’s 0-2% finding), halt the project. Re-evaluate your use case, find a more cost-effective solution, or build a smaller, bespoke model internally that actually delivers value.