Cloud Costs Match Payroll

Max discusses the significant infrastructure costs of running an AI writing assistant service and potential solutions through on-device processing. The core business challenge is managing processing costs while serving millions of free users.

Key Points:

  • Current Infrastructure Costs:

    • Cloud processing costs comparable to payroll costs (with ~800 employees)
    • Amazon Web Services is primary provider
    • Processing costs particularly challenging for free users who aren't monetized
  • On-Device Processing Solution:

    • Move calculations and suggestions to user devices (phones/laptops)
    • Benefits:
      • Better user experience
      • Lower latency
      • Higher performance
      • Enhanced privacy
      • Reduced cloud computing costs
      • No server processing required for basic functions
  • Business Model Challenge:

    • Tens of millions of free users generate costs without revenue
    • No alternative monetization (no ads, no data selling)
    • Cost burden especially high in developing countries where premium upgrades are less likely
  • Infrastructure Strategy:

    • Mix of cloud and in-house solutions
    • Some model training done in-house
    • Regular evaluation of bringing services in-house vs cloud
    • Amazon remains primary but not exclusive infrastructure provider
25:54 - 27:50
Full video: 30:53
ML

Max Lytvyn

Co-founded Grammarly, a leading writing assistance tool, in 2009. Ukrainian-born with Canadian citizenship, obtained through graduate studies in Toronto. Featured on the My First Million podcast, discussing potential IPO plans for Grammarly.

Twitter
Founder