Stargate’s Next Frontier: Scaling AI’s Backbone

Dear Readers,

Sometimes it feels as if the world of AI is building faster than we can keep up with. Just a moment ago, we were talking about models, and now we're talking about gigawatts, data centers, and the question of who actually controls the foundations of our digital future. This issue takes you into the next phase: not just what models can do, but what foundations they run on. Between visions of ten gigawatts of computing power and tough questions about efficiency and access, an exciting picture is unfolding.

To this end, we look at how new research approaches are achieving amazing results with tiny data sets, how GPT-5 Codex is finding its way directly into the everyday lives of developers, and why Elon Musk is already denying Anthropic's victory. A look at Stargate shows how infrastructure itself is becoming the actual battleground, while fresh models such as Qwen3-Max and Suno v5 are pushing creative boundaries. So there's a lot to cover – and the crucial question: Which of these movements most strongly shapes your understanding of AI?

In Today’s Issue:

⚡ Stargate is expanding with five new mega data center sites

🚀 Alibaba's Qwen3-Max is redefining coding and agentic AI benchmarks

🎶 Suno v5 just dropped!

🛡️ Qwen3Guard sets a new standard for AI safety

✨ And more AI goodness…

All the best,

OpenAI and NVIDIA have unveiled a landmark partnership to build 10 gigawatts of AI infrastructure powered by millions of NVIDIA’s new Vera Rubin GPUs. This move gives OpenAI the muscle to expand its own data centers while still scaling with existing cloud partners.

Musk’s remark cuts deep - framing Anthropic as a company that was never built to “win” outright. It suggests their role is more about shaping the field and keeping balance than seizing the crown.

The Takeaway

👉 Five new Stargate sites expand capacity toward 7 GW.

👉 Regional distribution cuts latency and boosts efficiency.

👉 Infrastructure is now the real competitive edge in AI.

👉 Control over compute geography will shape access and influence.

OpenAI, Oracle, and SoftBank just unveiled five new Stargate data center sites across the U.S., marking one of the boldest infrastructure bets in AI’s short history. With locations in Texas, New Mexico, Ohio, and the Midwest, this expansion pushes Stargate’s planned capacity toward 7 gigawatts - on track to hit 10 gigawatts and $500 billion in investment by the end of 2025.

This isn’t only about raw compute. By distributing infrastructure regionally, Stargate lowers latency, improves energy efficiency, and positions AI services to scale more sustainably. The move signals a clear shift: the future of AI will be shaped as much by power grids, data centers, and fiber as by model architectures. For developers, startups, and enterprises, this infrastructure race will decide access, costs, and speed of innovation.

The big question: will such centralized mega-projects define the AI era, or can more open and decentralized models emerge to challenge them?

Why it matters: It matters because compute infrastructure will determine who can train, deploy, and benefit from advanced AI. It matters because energy and geography are fast becoming as critical as algorithms and datasets.

Sources:

🔗 https://openai.com/index/five-new-stargate-sites/

Qwen3-Max is now pushing coding and agentic skills to the limit: the Instruct variant already rivals top models on SWE-Bench, Tau2-Bench, SuperGPQA, LiveCodeBench, and AIME25. The Thinking model, boosted with tool use and heavy-mode deployment, delivers near-perfect results—driven by huge scale, vast data, and relentless compute in pre-training and RL.

Learn More ►