Anthropic projects an operating profit of $559 million on $10.9 billion in the June quarter, a roughly 130% year‑over‑year revenue jump driven by enterprise adoption of coding tools and agentic use of Claude;
Anthropic expects an operating profit of $559 million on $10.9 billion in revenue for the June quarter, representing roughly 130% year‑over‑year growth, according to recent reporting. That pace of expansion — likened in the reporting to the early days of Zoom and pre‑IPO Google and Facebook — would bring the company close to its first profitable quarter after previously warning investors last summer not to expect annual profitability before 2028. If sustained, the revenue surge will force customers and procurement teams to factor higher token and capacity costs into budgets.
Two product trends account for the leap: rapid enterprise uptake of Anthropic’s coding tools and a growing pattern of “agentic” use of Claude, where the model autonomously executes multi‑step tasks over extended periods. Those agentic workloads and heavy coding demand have at times pushed utilization beyond Anthropic’s available compute, prompting the company to throttle some users and to sign additional data‑center capacity agreements, including a reported deal with SpaceX.
Technical adjustments to Anthropic’s flagship family are also changing user economics. Opus 4.7 keeps the same per‑token price as its predecessor but uses a new tokenizer that can split identical text into as much as 47% more tokens, increasing per‑request costs. Developer Abhishek Ray estimated roughly a 20 — 30% cost rise for an 80‑round session under Opus 4.7, while an OpenRouter analysis recorded token‑cost increases of 12 — 27% for prompts over 2,000 tokens.
Anthropic says its compute efficiency is improving even as token consumption climbs: compute spend fell from 71 cents per dollar of revenue in Q1 to an expected 56 cents this quarter. The company attributes part of that improvement to investor‑facilitated access to lower‑cost chips via capacity deals with major cloud providers and to not subsidizing a large free consumer tier, a strategy that has helped margins despite rising model‑level token usage.
The broader market is showing similar pricing pressure. Reported list prices for other major models have risen, and third‑party analyses point to real‑world cost increases across vendors. For engineers and procurement teams building on these models, practical steps include budgeting for token inflation, implementing rate limits or retries to control bursty agentic workloads, and negotiating capacity or dedicated‑compute agreements as coding and agentic use cases scale.
Sources
Replies (0)
No replies in this topic yet.
