Anthropic opens grants for rare disease projects

Where AI could actually help medicine

The announcement feels practical.

Anthropic is offering up to $50,000 in Claude credits over six months to researchers and early-stage biotechs working on rare genetic diseases. The program has two tracks: one focused on basic science and another aimed at speeding up clinical development.

This stood out because the company isn't promising breakthroughs. It is targeting the slow work that holds research back: reading huge volumes of literature, connecting scattered datasets, drafting regulatory documents, and finding biological links that researchers might otherwise miss. You can see the office lights still on when you think about how much of this work is done today.

That feels like a more believable use of AI. Faster paperwork and better pattern finding will not cure a disease, although they can give scientists more time to focus on experiments that matter.

Anthropic also admits AI has limits. Missing data, manufacturing delays, and access to care remain problems that software alone cannot fix. That honesty makes this program more credible than most AI healthcare announcements.

The question now is whether these tools can consistently shorten research timelines without creating more work for scientists to verify.

The AI chip race shifts to efficiency

Power is becoming the performance metric that matters.

According to a report from The Information, Google is developing a new AI chip called Frozen v2, expected in 2028. The chip could generate six to ten times more tokens per unit of power than Google's current AI processors, although the company has not confirmed the project.

The timing makes sense. The first wave of AI infrastructure was about building the biggest clusters possible. Now companies are under pressure to show that all of that spending can produce AI services at a lower operating cost.

That also explains why nearly every major AI company is investing in custom silicon. OpenAI, Anthropic, Microsoft, Amazon, and Google are all trying to reduce their dependence on Nvidia while tailoring hardware to their own models and workloads.

The next milestone in AI infrastructure may not be a faster model. It could be one that delivers the same results while using much less power.

Azure adds more AMD chips for AI

One chip no longer fits every AI job.

Microsoft is expanding Azure with three new AMD-powered offerings aimed at different workloads: HDv2 for AI data systems, HXv2 for chip design and scientific computing, and ND MI455X v7 for large-scale AI inference. The company is also bringing AMD's latest Helios AI platform and next-generation EPYC processors into Azure.

What matters is the direction, not the hardware specs. Microsoft is moving away from the idea that one GPU cluster can handle everything. Instead, it's building separate infrastructure for data preparation, inference, engineering simulations, and agent workloads because those jobs have different performance and cost requirements.

That makes sense. As AI deployments grow, customers will care less about buying the fastest hardware and more about matching the right infrastructure to each workload.

The next competition between cloud providers may come down to who gives developers the widest range of optimized systems instead of who offers the biggest AI cluster.

Prohuman team

Free Guides

Explore our free guides and products to get into AI and master it.

All of them are free to access and would stay free for you.

Feeling generous?

You know someone who loves breakthroughs as much as you do.

Share The Prohuman it’s how smart people stay one update ahead.