OpenAI's half-trillion dollar Stargate project

PLUS: Google opens up its Gemini ecosystem, and a tiny model beating giants

Author

Gen AI Art
October 10, 2025 • Estimated Reading Time: 8 minutes

Sponsored by

OpenAI is undertaking a massive spending initiative to build the infrastructure for future AI, reportedly signing deals worth hundreds of billions for its ambitious Stargate supercomputer project.

These colossal investments underscore the immense resource requirements for building frontier systems. As the race for AI dominance intensifies, does this level of spending create an insurmountable barrier for anyone but the largest tech giants?

Today in AI:

  • OpenAI’s half-trillion dollar Stargate project

  • Google opens up its Gemini ecosystem

  • A tiny AI model beats giants on reasoning tasks

PRESENTED BY SUPERHUMAN AI

Go from AI overwhelmed to AI savvy professional

AI will eliminate 300 million jobs in the next 5 years.

Yours doesn't have to be one of them.

Here's how to future-proof your career:

  • Join the Superhuman AI newsletter - read by 1M+ professionals

  • Learn AI skills in 3 mins a day

  • Become the AI expert on your team

What’s new? OpenAI is committing over $400 billion to secure its computational future, signing massive infrastructure deals with Oracle, NVIDIA, and AMD to power its next generation of AI models and the Stargate supercomputer.

What matters?

  • The largest component is a staggering $300 billion agreement with Oracle, which provides the foundational cloud infrastructure required for OpenAI's most ambitious AI projects.

  • The company is also deepening its ties with NVIDIA through a $100 billion strategic partnership, ensuring access to its cutting-edge GPUs that dominate the AI training market.

  • In a significant strategic move, OpenAI inked a multi-billion dollar deal with AMD, signaling a clear intent of diversifying its hardware supply and reducing dependence on a single chip provider.

Why it matters?

These unprecedented investments highlight the astronomical and growing cost of training and operating frontier AI models. This massive capital outlay shows that the race for AI dominance is now a battle fought with compute power, making access to scaled infrastructure the ultimate competitive advantage.

What’s new? Google is opening up its Gemini ecosystem, allowing anyone to build and release extensions without a formal approval process. This move sharply contrasts with OpenAI's curated and controlled app store model.

What matters?

  • Unlike OpenAI's walled garden, Google is removing the gatekeepers by allowing developers to publish extensions directly without a review.

  • The strategy bets on rapid, permissionless innovation, believing that letting users build whatever they want will lead to a richer and more diverse set of tools.

  • This open approach empowers developers to add new capabilities to Gemini instantly, bypassing the lengthy approval cycles common in other AI platforms.

Why it matters?

This open model could accelerate Gemini's feature growth and attract a large community of independent developers. For users and builders, it signals faster access to a wider variety of specialized AI tools directly within the Gemini interface.

PRESENTED BY MISO ROBOTICS

Elon Musk: “Robots Will…Do Everything Better”

And it’s already happening. Just look at fast food. Miso’s kitchen robots already fried 4m+ baskets of food for brands like White Castle. With 144% labor turnover and $20/hour minimum wages, they’re not alone. Partnered with NVIDIA and Uber’s AI team, Miso’s new robot sold out initial units in one week. Invest before Miso’s bonus shares change on 10/9.

This is a paid advertisement for Miso Robotics’ Regulation A offering. Please read the offering circular at invest.misorobotics.com.

What’s new? A new paper from a Samsung researcher introduces the Tiny Recursion Model, a 7M-parameter AI that outperforms models thousands of times its size on complex reasoning tasks using a self-improvement loop.

What matters?

  • It achieved a 45% score on the notoriously difficult ARC-AGI-1 benchmark, surpassing much larger models like DeepSeek R1 and Gemini 2.5 Pro.

  • The model drafts initial solutions and then refines them through up to 16 cycles of internal reasoning and revision on a separate scratchpad.

  • This approach shows that smart architectural tweaks can level the playing field, a trend also seen in models like Sapient's HRM.

Why it matters?

This research demonstrates that model performance isn't just about scale and compute power. Smart architectural designs can unlock impressive capabilities in small, resource-efficient models.

Everything else in AI

OpenAI unveiled its new AgentKit toolkit, designed to help developers build, deploy, and refine AI agents using a visual canvas, a connector registry, and integrated chat interfaces.

Zendesk claims its new AI agent can autonomously resolve up to 80% of customer support issues, aiming to free up human agents for more complex problems.

Anthropic lost a prominent AI researcher to Google, who cited the company's policy labeling China an "adversarial nation" as a key reason for his departure.

Meta began testing new AI-driven functionalities for its Ray-Ban display glasses, exploring enhanced augmented reality experiences.

Essential AI Guides - Reading List:

Let us know!

Work with us

Reach 100k+ engaged Tech Professionals, Engineers, Managers and decision makers. Join brands like MorningBrew, HubSpot, Prezi, Nike, Ahref, Roku, 1440, Superhuman, and others in showcasing your product to our audience. Get in touch now →