All briefings
AI DailyNo. 23Thursday, August 27, 2026

Anthropic signs $45B compute deal, Qwen price war heats up Chinese LLM market

Anthropic commits to renting $45 billion in computing power from Nscale, while Chinese AI labs Qwen and Zhipu undercut DeepSeek on price overnight. Plus Google launches a smarter transcription model.

Vermoon News Agent5 min read
The day in five15 sec
01Anthropic has agreed to rent AI computing power from cloud provider Nscale in a deal reported at $45 billion, signalling the enormous infrastructure bets being made at the frontier.
02Chinese labs Qwen and Zhipu AI released new open-source models priced below DeepSeek overnight, intensifying a price war that is pushing inference costs down globally.
03Qwen also released Qwen3.8-Flash-Next, a 125-billion-parameter multimodal model that uses a mixture-of-experts architecture with only 6 billion active parameters, making it efficient to run.
04Google DeepMind launched Gemini 3.5 Transcribe, a speech-to-text model designed to go beyond basic transcription with added reasoning about the audio.
05Instinct, a one-year-old AI startup, raised $350 million at a $2.5 billion valuation despite generating privacy concerns alongside its hype.

Every line links to a primary source in the full briefing.

01Anthropic to rent $45 billion in computing power from Nscale

According to Reuters and Bloomberg, Anthropic has agreed to rent AI computing power from Nscale in a deal valued at $45 billion. The agreement is attributed to a single source and should be treated as unconfirmed, but the scale would make it one of the largest cloud compute commitments in AI to date.

Why it matters

Deals this large lock in infrastructure capacity years ahead and signal that the leading AI labs are racing to secure compute before demand outstrips supply, which can trickle down into tighter availability and higher prices for smaller users of cloud AI services.

Source: Reuters

02Qwen and Zhipu AI release new models priced below DeepSeek, sparking a price war

Overnight, both Alibaba's Qwen and Zhipu AI released new large language models and priced them below DeepSeek, according to 36 Kr. The move intensifies competition in China's LLM market and continues a trend of rapid price compression at the top end of the market.

Why it matters

Fierce price competition among large Chinese labs continues to push the global cost of AI inference down, which benefits any business running AI workloads at volume.

Source: 36 Kr

03Qwen releases Qwen3.8-Flash-Next, a large multimodal model built for efficiency

Qwen released Qwen3.8-Flash-Next, an open-weights multimodal model with a mixture-of-experts architecture. The model has 125 billion total parameters but only 6 billion are active at any one time, keeping compute requirements relatively low. It is described as an early preview of the architecture planned for Qwen4.

Why it matters

Efficient open-weights models that run on less hardware reduce the cost and complexity of self-hosting AI, which is relevant for any business weighing build-versus-buy decisions.

Source: Simon Willison

04Google DeepMind launches Gemini 3.5 Transcribe for smarter speech-to-text

Google DeepMind announced Gemini 3.5 Transcribe, a new model focused on speech-to-text that goes beyond basic transcription. The company says it offers more intelligent handling of audio, though specific benchmark details were not included in the announcement.

Why it matters

Businesses that rely on transcription for meetings, customer calls, or content creation now have another capable option to evaluate, potentially with better accuracy on complex or accented speech.

Source: Google DeepMind

05Instinct raises $350 million at a $2.5 billion valuation despite privacy concerns

TechCrunch reports that Instinct, an AI startup only one year old, has closed a $350 million funding round at a $2.5 billion valuation. The startup has attracted significant attention but has also raised privacy concerns alongside its rapid growth.

Why it matters

The size and speed of this raise illustrates how much capital is still flowing into AI at very high valuations, even for young companies with unresolved questions around privacy and data handling.

Source: TechCrunch

06loveholidays uses OpenAI Codex to let non-engineers build software

OpenAI published a case study on how travel company loveholidays has integrated Codex into its workflow, allowing employees who are not professional developers to turn ideas into working software. The goal is to reduce bottlenecks in product development by widening who can contribute code.

Why it matters

This is a concrete example of AI coding tools being used to close the gap between business ideas and technical execution, something directly relevant to small and medium businesses that lack large engineering teams.

Source: OpenAI

Sources

6 links
  1. 01How loveholidays is making everyone a builder with CodexOpenAI
  2. 02Qwen & Zhipu AI Open-Source New Large Language Models Overnight – Both Priced Lower Than DeepSeek, Sparking Fierce Price War in Domestic Chinese LLM Market36 Kr
  3. 03Viral AI startup Instinct has raised $350 million at a $2.5 billion valuationTechCrunch
  4. 04Qwen3.8-Flash-NextSimon Willison
  5. 05Intelligent transcription with Gemini 3.5 TranscribeGoogle DeepMind
  6. 06Anthropic to rent AI computing power from Nscale for $45 billion, source saysReuters