Welcome back to LLM Decode 👋
Today’s stories show that the AI race is expanding in two directions at once. OpenAI is securing the physical infrastructure needed to run AI at massive scale, while Anthropic has reportedly stepped away from what could have been a $6 billion acquisition- showing that compute and capital allocation are becoming just as important as model performance.
OpenAI’s Next AI Bet Is… Data Centers
OpenAI has signed a multi-year strategic partnership with Australian AI-infrastructure company Firmus for dedicated computing capacity at two AI Factory sites in Malaysia. OpenAI will become an anchor customer for the new capacity.
The agreement pushes Firmus’ total contracted capacity across customers to more than 900 MW. Firmus says its broader footprint now spans seven AI factories across Australia, Singapore, Indonesia and Malaysia, with five facilities still under development.
The hardware matters too. Firmus plans to deploy NVIDIA Vera Rubin NVL72 systems across its Asia-Pacific infrastructure, while OpenAI says the Malaysian facilities will help it meet growing product demand both in the region and globally.
Why it matters
Compute is becoming a competitive moat. Better models require not only research talent, but reliable access to enormous amounts of GPUs, electricity, cooling and networking.
AI infrastructure is globalizing. Malaysia is becoming part of the infrastructure layer supporting frontier AI products- not simply a market consuming them.
Watch the cost per token. Firmus explicitly frames its infrastructure around improving deployment efficiency and lowering token-production costs.
The AI stack is getting bigger. Competitive advantage increasingly stretches from models all the way down to chips, data centers, energy and connectivity.
How 2M+ Professionals Stay Ahead on AI
What’s the secret to staying ahead of the curve in the world of AI? Information.
Luckily, you can join 2,000,000+ early adopters reading The Rundown AI — the free newsletter that makes you smarter on AI with just a 5-minute read per day.
Anthropic Walked Away From a Potential $6B Acquisition
Anthropic has reportedly decided not to pursue an acquisition of Decart AI, according to Bloomberg reporting cited by multiple outlets. The companies had previously been discussing a deal valuing the startup at roughly $6 billion.
The potential acquisition was notable because Decart works on technologies including AI world models and highly efficient AI infrastructure. Earlier reporting described the proposed transaction as potentially Anthropic’s largest known acquisition.
For now, the important distinction is that Anthropic did not announce a $6B acquisition and then cancel it. Reporting indicates it was considering a deal and ultimately decided against pursuing it, a more accurate framing than saying a completed deal fell apart.
Why it matters
AI labs are becoming strategic buyers. Frontier labs increasingly need technology beyond foundation models, from infrastructure optimization to agents and world models.
$6B shows how valuable differentiated AI technology can become. Even an acquisition discussion at that level signals how aggressively strategic AI assets are being valued.
Building vs. buying is becoming a major decision. Labs must constantly decide whether acquiring specialized teams and technology is faster than developing capabilities internally.
Watch consolidation. As AI infrastructure and research costs rise, expect more partnerships, investments and acquisition attempts around frontier labs.
4 AI Workflows to Take From News
Compute-Cost Mapping: Before scaling an AI product, track model cost per request, latency and usage volume together. OpenAI’s infrastructure expansion is a reminder that AI economics matter as much as capability.
Build-vs-Buy AI Audit: When your team needs a new AI capability, compare three paths: build internally, integrate an API/tool, or acquire/partner. Anthropic’s reported Decart decision shows this trade-off happening even at frontier-lab scale.
Model Routing Workflow: Don’t automatically send every task to your most powerful model. Route simpler jobs to cheaper models and reserve frontier models for tasks where the additional capability creates measurable value.
AI Dependency Map: List the models, APIs, cloud providers, data sources and infrastructure your important AI workflows depend on. Then identify where one failure or price change could disrupt the entire workflow.
Ready to level up your AI skills?
Explore Our Courses →That’s it for today.
The AI space doesn’t slow down - and neither should your thinking.
See you in the next drop.