Stanford researchers found that Meta's newest AI model can reproduce 42% of Harry Potter word-for-word—ten times more than earlier versions. The findings complicate copyright lawsuits and reveal a troubling trend in AI development.
Anthropic says multiple AI agents working together beat single models by 90%. The catch? They use 15x more computing power. This trade-off between performance and cost might reshape how we build AI systems for complex tasks.
AI models typically learn by memorizing patterns, then researchers bolt on reasoning as an afterthought. A new method called Reinforcement Pre-Training flips this approach—teaching models to think during basic training instead.
Oracle just agreed to spend $40 billion on chips for OpenAI's first independent data center. The deal marks OpenAI's break from Microsoft after years of relying on its cloud. Why now, and what does this mean for the AI race?
Oracle just agreed to spend $40 billion on Nvidia chips to power OpenAI's new data center in Texas. The deal covers 400,000 of Nvidia's latest GB200 processors and marks a major step in OpenAI's plan to break free from Microsoft's computing stranglehold.
The facility in Abilene will gulp down 1.2 gigawatts of power when it opens in mid-2026. That's enough electricity to run about one million homes. Oracle will lease the site for 15 years and rent the computing power to OpenAI.
This Texas operation represents the first real piece of Stargate, the $500 billion data center network that OpenAI and SoftBank unveiled in January. President Trump announced the project with great fanfare, positioning it as America's answer to the global AI arms race.
The numbers behind the Texas site tell their own story. Crusoe Energy Systems and Blue Owl Capital own the facility and raised $15 billion to build it. JPMorgan provided most of the debt financing through two loans totaling $9.6 billion. The remaining $5 billion came from equity investors.
Eight buildings will house Oracle's mountain of Nvidia chips. Each GB200 processor combines two of Nvidia's newest Blackwell graphics cards with a 72-core central processing unit. These chips run AI models using 25 times less electricity than their predecessors. Built-in AI models even predict when the hardware might fail.
The timing matters. OpenAI ended its exclusive cloud computing deal with Microsoft last year after growing frustrated with supply constraints. Microsoft had invested nearly $14 billion in OpenAI, much of it in the form of cloud computing credits. Now OpenAI wants its own infrastructure.
Oracle's massive chip purchase puts it in direct competition with other data center plans. Elon Musk wants to expand his Colossus facility in Memphis to house about one million Nvidia chips. Amazon is building a data center in northern Virginia that will exceed one gigawatt of power. The race to build AI infrastructure has become a high-stakes game of who can accumulate the most computing power.
But Texas is just the beginning. OpenAI and its partners have much bigger plans through Stargate. The project aims to spend $100 billion on data centers in its first phase, with that figure potentially reaching $500 billion over four years.
The ownership structure reveals the true scope of ambition. OpenAI and SoftBank each committed $18 billion for majority control. Oracle and MGX, an Abu Dhabi sovereign wealth fund, each pledged $7 billion. This isn't just about building data centers. It's about reshaping the global AI landscape.
The international expansion started quickly. During Trump's recent Middle East tour, the partners announced plans for a massive AI campus in the United Arab Emirates. This facility will span 10 square miles in Abu Dhabi and consume 5 gigawatts of power. That's more than four times the size of the Texas operation.
The UAE campus involves Emirati firm G42 as a local partner. OpenAI and Oracle will manage a 1-gigawatt compute cluster within the larger facility. Nvidia will supply the chips while Cisco handles connectivity infrastructure. A 200-megawatt cluster should come online next year, with the full facility operational by 2026.
These international partnerships matter for more than just computing power. They help OpenAI access global markets while keeping the U.S. government happy about maintaining technological leadership. The UAE deal reinforces America's AI infrastructure while giving allies access to advanced technology.
The chip technology itself represents a major leap forward. Taiwan Semiconductor Manufacturing Company produces the GB200 processors using its enhanced N4P process. This manufacturing technique requires fewer steps involving photomasks, the panels that filter laser light during chip production. Fewer steps mean faster production, which should help Nvidia meet the crushing demand for its processors.
OpenAI's strategy makes sense from multiple angles. The company gets guaranteed access to massive computing power without depending on Microsoft's goodwill. It can scale its AI training and inference workloads without worrying about capacity constraints. And it positions itself as a true infrastructure player, not just a software company renting someone else's computers.
The financial mechanics work for everyone involved. Oracle gets a massive long-term customer and positions itself as a major player in AI infrastructure. Nvidia sells billions of dollars worth of its most advanced chips. The data center owners get stable, long-term tenants with deep pockets.
For OpenAI, the deal solves multiple problems at once. The company can train larger AI models without hitting Microsoft's capacity limits. It can offer more reliable service to ChatGPT users. And it reduces the risk that a single cloud provider could cut off access to computing resources.
The broader implications extend beyond these specific companies. America's AI infrastructure is becoming a matter of national security and economic competitiveness. China has made massive investments in AI computing power. The Stargate project represents America's answer to that challenge.
Other AI companies are watching closely. If OpenAI can successfully build its own infrastructure, competitors like Anthropic and Google might follow similar paths. The cloud computing oligopoly of Amazon, Microsoft, and Google could face new competition from companies that control both AI software and hardware.
The timeline for all this infrastructure remains ambitious. The Texas facility should be fully operational by mid-2026. The UAE campus will start with a smaller cluster next year before reaching full capacity. OpenAI is also considering data center locations in 16 U.S. states as part of the broader Stargate initiative.
Success isn't guaranteed. Building data centers at this scale involves enormous technical and financial risks. Supply chain disruptions could delay chip deliveries. Regulatory hurdles might slow construction. Energy costs could spiral higher than expected.
But the stakes are too high for these companies to move slowly. The AI revolution is happening now, and the companies with the most computing power will likely emerge as winners. OpenAI's bet on its own infrastructure represents a fundamental shift in how AI companies think about their competitive advantages.
Why this matters:
Q: How much electricity will the Texas data center actually use?
A: The Abilene facility will consume 1.2 gigawatts of power when fully operational - equivalent to the electricity usage of about one million households. To put this in perspective, that's enough to power a city the size of Dallas for several hours.
Q: When will OpenAI start using the new data center?
A: The Texas facility is expected to be fully operational by mid-2026. However, OpenAI will likely begin using portions of the data center as they come online throughout 2025 and early 2026, rather than waiting for complete construction.
Q: What makes Nvidia's GB200 chips so much better than previous models?
A: Each GB200 chip combines two Blackwell B200 graphics cards with a 72-core CPU and uses 25 times less electricity than predecessors. They also include built-in AI models that predict hardware failures before they happen, reducing downtime.
Q: Who actually owns the Texas data center?
A: Crusoe Energy Systems and Blue Owl Capital own the facility. Oracle has signed a 15-year lease and will rent the computing power to OpenAI. This arrangement allows OpenAI to access massive computing resources without the upfront capital investment.
Q: How does this compare to other major AI data centers?
A: At 1.2 gigawatts, it's smaller than the planned UAE facility (5 gigawatts) but competitive with Elon Musk's Colossus expansion in Memphis, which aims to house one million Nvidia chips versus Oracle's 400,000.
Stanford researchers found that Meta's newest AI model can reproduce 42% of Harry Potter word-for-word—ten times more than earlier versions. The findings complicate copyright lawsuits and reveal a troubling trend in AI development.
Meta users think they're chatting privately with AI. Instead, they're broadcasting medical questions, legal troubles, and relationship problems to the world through a public feed that many don't realize exists.
Disney and Universal sued AI company Midjourney for using their characters without permission to train image generators. It's the first major Hollywood lawsuit against generative AI, testing whether copyright law protects creators in the age of artificial intelligence.
OpenAI cut its o3 model prices 80% while launching o3-pro—a reasoning AI that takes over 3 minutes to respond but outperforms rivals on complex tasks. The move intensifies AI pricing wars and splits the market between fast chat models and slow thinking ones.
Get tomorrow's intel today. Join the clever kids' club. Free forever.
