Lanceum — AI & Technology in Asia

Anthropic Walks Away From $6 Billion Decart Acquisition After Due Diligence

A Firmus AI Factory data center facility

OpenAI Becomes Anchor Customer of Nvidia-Backed Firmus for Malaysian AI Factories

OpenAI will contract dedicated compute from two Firmus AI Factory sites in Malaysia under a multi-year deal that pushes the Sydney-based firm's contracted capacity beyond 900 MW — and deepens OpenAI's Southeast Asia footprint.

Japan's AI Data Center Capacity Set to Quadruple to 4.9 GW by 2033 With $60 Billion Buildout

A Nikkei survey of 26 operators finds Japan's AI data center capacity could more than quadruple by 2033, led by NTT's 2 GW plan, Blackstone's $30 billion commitment and Microsoft's $10 billion pledge.

Kioxia's High Bandwidth Flash prototype module

Kioxia Bets on High-Bandwidth Flash to Crack the AI Memory Wall

The Japanese memory maker outlined plans to capture AI demand with a 5TB, 64 GB/s High Bandwidth Flash module that pushes NAND toward the GPU memory bus — offering 8-16x the capacity of DRAM-based HBM.

More Headlines

Shipping containers at the port of Busan, South Korea

South Korea Smashes Full-Year Export Record in 248 Days as AI Chip Boom Powers $709.4 Billion

Sep 8

Nvidia and Hugging Face logos following the announcement of the acquisition agreement

Nvidia Agrees to Buy Hugging Face for $13 Billion in Its Biggest Bet on Open AI Models

Sep 8

Newsletter

Weekly AI & tech briefing from Asia.

Subscribe →

News

Anthropic Walks Away From $6 Billion Decart Acquisition After Due Diligence

OpenAI Becomes Anchor Customer of Nvidia-Backed Firmus for Malaysian AI Factories

Japan's AI Data Center Capacity Set to Quadruple to 4.9 GW by 2033 With $60 Billion Buildout

Kioxia Bets on High-Bandwidth Flash to Crack the AI Memory Wall

Analysis

Sheet music and a gavel representing the music industry's copyright lawsuits against AI companies

Maya Santos•September 8, 2026

All Three Majors Are Now Suing Anthropic. The AI Copyright Endgame Is Taking Shape

Sony and Warner's new lawsuit — naming Dario Amodei personally — completes the music industry's encirclement of Anthropic, while the chaotic payout of the $1.5 billion book settlement previews what an AI licensing regime will actually look like.

Transmission lines and a data center under construction in Asia

The Backlash Has Arrived: Asia's Data Center Boom Is Colliding With the People Who Live Next Door

From preschool parents in rural Australia to grid politics in India and Malaysia, public resistance to AI data centers is emerging as the binding constraint on Asia's compute buildout — and operators are unprepared for it.

Nvidia CEO Jensen Huang presenting on stage

Groq, Nemotron, Now Hugging Face: Nvidia Isn't Buying Companies — It's Buying the Open AI Ecosystem

The $13 billion Hugging Face deal completes a vertical stack from silicon to model distribution. What Nvidia gains, what 'open' means when it has an owner, and why China's PyTorch play landed the same week.

Illustration of Chinese AI companies competing globally against US labs

Maya Santos•September 6, 2026

The Cost War: How Moonshot, Z.ai and DeepSeek Are Beating US Labs on Price — and Why It Works

Chinese frontier labs now ship models within striking distance of the US frontier at a tenth of the price. The strategy isn't charity — it's a deliberate play for the world's developer mindshare, and it's winning.

Weekly Briefing

Stay ahead on
AI in Asia

Join thousands of readers who get our curated weekly briefing. The most important AI and tech developments from across the region — no hype.

Free • Weekly • Unsubscribe anytime

Startups

Travis Kalanick, co-founder of Uber and founder of Atoms

Maya Santos•September 8, 2026

Travis Kalanick's Atoms Eyes Robotaxis — With Uber's $100 Million and His Old AV Team

Fresh off a $1.7 billion a16z-led round, the Uber co-founder's 'physical AI' startup is reuniting veterans of Uber's autonomous vehicle program and holding early talks to put its technology on Uber's network.

GPU server racks in a Fluidstack-operated data center

Maya Santos•September 8, 2026

Fluidstack Closes $1.5 Billion Led by Jane Street — Revenue Up From $1.8M to a Projected $660M, While Owning Zero Chips

The AI cloud that aggregates other people's GPUs raised one of the week's largest rounds, riding its $50 billion Anthropic infrastructure pact into hyperscale — an asset-light bet in an increasingly asset-heavy race.

Gimlet Labs branding and server hardware

Gimlet Labs Raises $300M at $3B Valuation to Build the First Multi-Silicon Inference Cloud

Andreessen Horowitz leads the Series B for a company that decomposes AI models across GPUs, CPUs, near-memory compute and dataflow chips — with Arm and Microsoft's M12 betting inference is about to escape the GPU.

Illustration of startup funding trends across Asia

Daniel Park•September 5, 2026

Asia's Startup Funding Hits a Multiyear Peak — and AI Is Taking More Than 60 Cents of Every Dollar

Investors poured $42.8 billion into Asian startups in Q2, with AI companies capturing over $26 billion. Southeast Asia's native AI funding has doubled to $4.1 billion — but one deal accounts for most of it.

Research

Google Ships Gemini 3.8 Flash and a 'Cyber' Twin That Hunts Vulnerabilities

The new Flash flagship jumps to 90.8% on Terminal-Bench 2.1 and beats larger frontier models on long-horizon coding, while Gemini 3.8 Flash Cyber debuts under Google's restricted Fairwind program for governments.

Meta AI's Muse Spark 1.3 model announcement

Meta's Muse Spark 1.3 Reaches the Frontier: 75.4% on DeepSWE, 1M Context, and a Data-for-Discount Endpoint

Meta's strongest model yet posts its biggest jump on coding and agentic benchmarks — and introduces a 'contributor' pricing tier that is 10-20x cheaper if you let Meta train on your data.

Qwen3.8-Flash-Next: Alibaba Previews the Qwen4 Architecture With Hybrid Attention and 'Engram' Embeddings

The experimental open-weights release activates just 6B of 125B parameters per token, pairs Gated DeltaNet with Qwen Sparse Attention, and adds a 51B n-gram lookup table — a blueprint for where China's most-downloaded model family goes next.