AI inference service provider Baseten secures $1.5 billion in funding, with a double-layer valuation reaching up to $13 billion, and Wellington Management entering the AI inference track for the first time. But the gross profit margin risk from renting and reselling GPUs is the metric that this track should be most closely watched.

Baseten builds a software layer on top of the computing power from 20 cloud service providers, with clients including Cursor and Mercor. Its annualized recurring revenue in Q1 2026 skyrocketed from $200 million to $600 million, a 20-fold increase year-over-year. The CEO states that using open-source models for specific tasks costs only 30% of closed-source solutions.
But gross profit margin is the Achilles' heel for compute intermediaries. Renting and reselling GPUs is essentially about profit margin arbitrage; if upstream cloud providers cut prices or tighten terms, profit margins can be quickly squeezed. Wellington’s entry is seen as a signal of the sector’s durability, but institutional backing does not mean the business model is invulnerable.
The explosive demand for AI inference is certain, but whether compute intermediaries can maintain profits depends on whether they can establish sufficiently deep software barriers, making customer switching costs higher than their bargaining power. Otherwise, high growth could be accompanied by thin margins or even losses.
For the crypto market, the capital influx into the AI inference track is diverting funds that might have otherwise flowed into DePIN and compute tokenization. The success or failure of compute intermediaries also indirectly affects the credibility of decentralized compute network narratives.
$ai #rwa #AI #区块链 #Crypto Market