Inference is the fastest-growing market in technology, projected to hit $250 billion within seven years, three times the size of the database market. Anthropic booked $9 billion and $10 billion in consecutive months. Google Cloud is growing 63% at an $80 billion run rate. The companies winning are either selling inference directly or indexed tightly to it.
Two pre-AI public companies have found the second position: Twilio and Datadog. Datadog CEO Olivier Pomel reported LLM Observability spans nearly tripling quarter-over-quarter, and 6,500 AI-integrated customers, just 20% of the total base, account for 80% of ARR. Twilio CEO Khozema Shipchandler confirmed AI-native companies and enterprises are using voice as their entry point to the platform. Neither sells inference. Both benefit because anyone running AI systems needs observability and communication infrastructure.
The original is worth reading for the power law argument buried in the middle: a small number of AI-native customers are driving outsized revenue concentration across the entire software stack. Tunguz frames this as the defining pressure on every legacy SaaS company, resell inference, index to it, or fade. The piece calls it the Saaspocalypse. That framing is blunt, and the earnings quotes make it concrete.
[READ ORIGINAL →]