AI news, Nov 15, 2026: Anthropic Compute Dilemma and Apple AI Cuts
Anthropic's CEO stated that the company's near $901 billion in compute commitments poses its biggest commercial dilemma before its IPO, with it remaining unclear if future revenue can support these costs. Concurrently, Apple has officially initiated internal AI budget cuts, reducing the daily AI spending cap for engineers from $300 to $200, a 33% decrease. These developments indicate that the tech industry is reassessing AI investments, shifting from large-scale expansion to a more cost-effective,精细化管理 approach.
Lead written by AI from the day’s stories
Top stories
- Open source
datasette-llm supports OpenAI-compatible endpoints
Simon Willison released datasette-llm 0.1a11, adding support for OpenAI-compatible endpoints. The update allows Datasette users to configure LLMs hosted locally via LM Studio or Ollama, utilizing a unified web protocol interface to enhance development toolchain flexibility.
- Funding
Anthropic CEO Details Compute Spending Dilemma Pre-IPO
Anthropic CEO Dario Amodei stated that buying compute is the company's deepest business dilemma ahead of its record IPO. The company's total commitments for compute resources are nearly $901 billion, and it remains unclear if future revenue will justify these costs. The IPO serves as a financial buffer against this uncertainty.
- Models
Gemini 4 Carbon Tops MathArena Benchmark
Google's Gemini 4 Carbon and Anthropic's Fable 5.2 surpassed OpenAI's models on the MathArena AI math benchmark. Gemini 4 Carbon showed strong performance but cost 15 times more than GPT-6.1 Sol, often ignoring time limits during testing.
- Products
Anthropic updates Claude Opus 5.5 guide
Anthropic has set Claude Opus 5.5 as the default model for all paid plans and published a new guide. The guide advises users to discard outdated prompting habits and start at medium effort to optimize performance.
- Open source
Cmpnd Releases dspy-wsp: Weight-Space Programming Comes to DSPy
Cmpnd released dspy-wsp, a prototype DSPy optimizer enabling frontier LLMs to read and edit small model weights. The method requires no gradients and costs about $5 to $10 per run, offering a new approach to model tuning.
Products1
- Products
LG U+ to launch AI Token service next year to cut enterprise costs
South Korean telecom carrier LG U+ announced it will enter the AI Token business next year, using proprietary optimization technologies and a chip combination strategy to help enterprises reduce AI transformation costs. The service improves inference efficiency via smart routing and model lightweighting, utilizing a heterogeneous semiconductor architecture.
Research1
- Research
Surge AI test shows AI misses key tax questions
Surge AI released results for the DAYJOB: Tax benchmark simulating US tax consultation. The study found that while AI provides answers, it often overlooks critical questions that should be asked, highlighting limitations in professional tax work.
Other1
- Other
Apple cuts internal AI budget, slashing engineer quotas by 33%
Apple has initiated an internal AI budget reduction to control costs. The daily AI spending cap for engineers was cut from $300 to $200, a 33% decrease, with plans to lower some quotas to $100. Non-technical departments also face cuts, signaling a shift toward precise ROI accounting for AI usage.
All news → · All rights belong to the original creators