Facing an AI bill that looks like their velocity chart, Intercom deliberately absorbs the cost. They encourage universal use of the most powerful models, viewing the immediate gains in speed and innovation as an investment that outweighs near-term cost concerns.

Healthcare AI firm Abridge is building its own models with NVIDIA not primarily for cost savings, but to achieve the real-time performance and low latency required for in-workflow clinical tools. This shows that for specialized, high-stakes applications, controlling the model stack for speed is more critical than just reducing token costs.

Newer AI models may have low per-token prices but are often "token hungry," requiring more tokens to complete a task. This can make them more expensive overall. The true measure of economic viability is the final cost-per-task, not the misleading per-token price.

The era of using the most powerful AI model for every task is ending. Companies are now focused on the trade-off between quality, cost, and latency. The key question is no longer "Which model is best?" but "Which model is good enough for this task at the lowest price point?"

Superhuman's CEO advises against simply tracking AI costs, a practice he calls 'token maxing'. Instead, they evaluate the ROI of internal AI tools by measuring developer productivity metrics like feature delivery pace. This output-focused approach has doubled engineering velocity, justifying the AI spend.

The high operational cost of using proprietary LLMs creates 'token junkies' who burn through cash rapidly. This intense cost pressure is a primary driver for power users to adopt cheaper, local, open-source models they can run on their own hardware, creating a distinct market segment.

In response to budget blowouts from agentic AI, enterprises are moving beyond simple adoption to active cost management. A new "token efficiency" stack is emerging, featuring tactics like model routing to cheaper alternatives (e.g., DeepSeek) and custom post-trained models to reduce reliance on expensive foundation models.

Parser's AI costs are lower than its server costs. They achieve this by intentionally avoiding the most powerful, expensive LLMs which are often slow and rate-limited. Instead, they find a balance, prioritizing speed and cost-effectiveness to process high volumes affordably.

AI-native companies grow so rapidly that their cost to acquire an incremental dollar of ARR is four times lower than traditional SaaS at the $100M scale. This superior burn multiple makes them more attractive to VCs, even with higher operational costs from tokens.