United States · Business news · AI & Tech · 07 Oct 2026
AI: Models and labs · Cloud and data centres · Enterprise software · Anthropic · Amazon (AWS) · Alphabet (Google) · Microsoft — more AI news
Anthropic released its small model Claude Haiku 5.5 on October 7 on its own platform and on Amazon Web Services, Google Cloud and Microsoft Azure. For requests of up to 100,000 tokens it costs $0.10 per million input tokens and $0.50 per million output tokens, against $1.00 and $5.00 for Haiku 4.5. Larger requests cost $0.50 and $2.50, a 50% reduction. Anthropic also halved the cache-read price of its mid-range Sonnet 5.5 model to $0.10 from $0.20 per million tokens.
The finance angle is the price of AI computing for the companies that buy it. Anthropic puts the overall cost of running a workload about 75% below Haiku 4.5, and says the Sonnet change takes roughly 20% off a typical agent task. Software companies and IT services firms that build products on these models pay less per task; the three cloud providers that resell the model earn a share of a lower unit price and depend on volume growing to make up for it. Anthropic itself gives up revenue per token while it prepares an IPO. Rival model sellers face the same price point in the small-model tier.
Reported by Anthropic — “Introducing Claude Haiku 5.5” ↗ on 07 Oct 2026.
Official series on 07 Oct 2026 or the last day before it, and the latest observation.
The finance angle is the contracted revenue: about $36 billion over the initial terms of the leases, paid by the AI cloud companies and hyperscalers that rent the capacity. That backlog is many times current revenue…
The finance angle is where AI work gets paid for. A model that runs on the laptop shifts spending from cloud usage to the device: Nvidia earns from the chip, Microsoft from the hardware and Windows, and the buyer pays…
The Dow fell about three times as much as the S&P 500 and the Nasdaq, so the day's weakness sat more with older industrial and financial names than with the large technology companies that dominate the other two…
Taiwan is where Micron makes much of its DRAM and the high-bandwidth memory that goes into AI servers, so a stoppage there would reach Nvidia-based server supply chains at a time when memory is already short. The money…