Anthropic bets on efficiency over scale, co-founder Amodei says

Anthropic is focusing on algorithmic efficiency and disciplined spending, contrasting with rivals like OpenAI that are committing large sums to computing infrastructure, President and co-founder Daniela Amodei said.

The San Francisco-based AI firm has roughly US$100 billion in compute commitments, far less than OpenAI’s reported US$1.4 trillion.

Anthropic prioritizes developing models that deliver more capability per dollar, rather than pursuing the largest pre-training runs.

Its Claude model is available on major cloud platforms, allowing flexibility in infrastructure choices.

Anthropic said its revenue has increased tenfold year-on-year for three consecutive years.

Both Anthropic and OpenAI have not announced IPO timelines, but are taking steps that suggest preparation for potential public offerings while continuing to raise capital.

🔗 Source: CNBC

🧠 Food for thought

Implications, context, and why it matters.

Anthropic’s efficiency claims need hard data to be credible

Series F news cites rapid run-rate revenue growth (annualized from recent results) but omits absolute revenue and margin data. The case for an efficiency-first strategy with better unit economics (profitability per unit of usage) over scale rivals like OpenAI stays unproven 1.
Run-rate revenue rose from $1 billion in early 2025 to over $5 billion by August 2025. Projected gross profit margins (revenue minus direct costs as a percentage) jump from negative 94% last year to an expected 50% in 2025 1 2.
OpenAI expects $13 billion revenue in 2025 with API revenue of $1.8 billion, while Anthropic projects $3.8 billion in API revenue, yet without standardized totals and costs any efficiency comparison is guesswork 2.

Third-party AI gateways gain ground as enterprises hedge provider risk

Enterprise IT teams adopt multi-model setups. Gateways are software layers that route requests between applications and multiple AI model providers. Examples include Kong (an API gateway vendor), OpenRouter (a model-routing marketplace), plus Atlassian’s internal solution (from an enterprise software company). They route across providers including Claude to tune cost, latency (response time), and compliance 3 4.
Founders and investors face a split market. Kong targets large enterprise workloads with security features. OpenRouter favors model flexibility over raw performance, creating room for specialized gateways in specific verticals 3.
Atlassian’s AI Gateway supports 100+ use cases across 8+ apps with 20+ models from 5+ providers. That signals demand for vendor-neutral orchestration layers (tools that sit between enterprise applications and AI providers to dynamically select models) 4.