The Price Check Nobody Wanted to See

Three major AI vendors. Three wildly different pricing strategies. One uncomfortable truth: the subsidized era of AI inference is over, and enterprise buyers are finally feeling it in their wallets. Tom Tunguz, a general partner at Theory Ventures and former Google PM, just dropped a data-driven breakdown on AI pricing that should make every CFO sweat.

The Numbers Don't Lie

Let's talk specifics. Google's Gemini 3.1 Pro runs $2.00 per million input tokens and $12.00 for output—a setup that's still less than half of what the competition charges. Anthropic's Claude Opus 4.7 sits at $5.00 input / $25.00 output per million tokens. And OpenAI's GPT-5.5? A cool $5.00 input and a staggering $30.00 output per million tokens.

Three Strategies, One Direction

Google has been tripling prices annually—an aggressive monetization play that signals confidence in demand. OpenAI appears to have exhausted its subsidy phase and is now chasing margins hard. Anthropic took the opposite path: holding luxe pricing steady until late last year, even dropping costs on their most powerful Opus models. That strategy worked when cash was plentiful and market share was the priority. That's no longer the case.

Why Prices Are Climbing Now

The common thread across all three vendors? CapEx spending is hitting record highs, and everyone's feeling the pressure to show profitability rather than just growth. Tunguz frames it perfectly: cuts happen when cash is abundant and share matters; increases happen when margins matter. All three are now in margin mode. The cheap inference era—the VC-subsidized period where startups got AI practically free—is officially dead.

Key Takeaways

  • Google's Gemini maintains the lowest absolute prices but triples annually—aggressive monetization, not generosity
  • OpenAI's GPT-5.5 charges 2.5x more for output than input tokens, their most profitable lever
  • Anthropic briefly played the discount card on top-tier models, but the industry trend is unmistakably upward
  • CapEx records and margin pressure mean enterprise buyers should expect continued annual increases

The Bottom Line

The AI industry sold developers a story about commoditized intelligence. What they're delivering is premium pricing with no ceiling in sight. If your startup's margins depend on cheap API calls, start hedging now—because Google, Anthropic, and OpenAI are done playing nice.