In partnership with

It’s go-time for holiday campaigns

Roku Ads Manager makes it easy to extend your Q4 campaign to performance CTV.

You can:

• Easily launch self-serve CTV ads

• Repurpose your social content for TV

• Drive purchases directly on-screen with shoppable ads

• A/B test to discover your most effective offers

The holidays only come once a year. Get started now with a $500 ad credit when you spend your first $500 today with code: ROKUADS500. Terms apply.

Get started today.

The AI Week That Rewrote the Rules: September 15-22, 2025

The past week delivered some of the most consequential AI developments of the year, fundamentally reshaping power dynamics across the industry. From unprecedented infrastructure deals to sweeping regulatory actions, here's what really happened behind the headlines.

The $100 Billion Handshake That Changes Everything

NVIDIA-OpenAI Partnership Transforms AI Infrastructure

The biggest story this week wasn't just about money—it was about strategic realignment. NVIDIA announced its intention to invest up to $100 billion in OpenAI, but the real significance lies in what this partnership represents for the future of AI computing.

The deal structures around deploying at least 10 gigawatts of NVIDIA systems for OpenAI's next-generation infrastructure, with the first gigawatt coming online in late 2026 using NVIDIA's Vera Rubin platform. To put this in perspective, Jensen Huang noted this equals roughly 4-5 million GPUs—essentially doubling last year's total output.

What makes this transformative isn't the dollar figure, but the vertical integration it creates. OpenAI gets guaranteed access to cutting-edge hardware at scale, while NVIDIA secures a massive customer for its most advanced systems. This bypasses the traditional cloud provider bottleneck that has constrained AI development.

The market responded predictably—NVIDIA's stock jumped nearly 4%, adding around $170 billion to its market cap. But the deeper implications extend beyond valuations. This partnership signals a shift toward AI companies building their own infrastructure rather than relying solely on traditional cloud providers.

Microsoft's Strategic Repositioning

The OpenAI Partnership Evolves

Parallel to the NVIDIA deal, Microsoft and OpenAI reached a tentative agreement restructuring their partnership, with OpenAI's nonprofit arm receiving over $100 billion in equity—roughly 20% of OpenAI's $500 billion valuation.

This restructuring is crucial for understanding AI's current power dynamics. Microsoft, despite investing over $13 billion in OpenAI since 2019, is preparing for a future where it's no longer the exclusive compute provider. The new agreement allows OpenAI to pursue its own data center initiatives, including the $300 billion Oracle deal announced earlier this month.

Microsoft's own AI independence strategy has been accelerating, with the launch of MAI-Voice-1 and MAI-1-preview models representing its pivot from OpenAI dependence. MAI-Voice-1 can generate a full minute of audio in under one second using just a single GPU, while MAI-1-preview ranks 13th on LMArena benchmarks.

As Microsoft AI chief Mustafa Suleyman put it: "The company has to be able to have the in-house expertise to create the strongest models in the world". This isn't about ending the OpenAI partnership—it's about building leverage within it.

California's AI Safety Bill: The Regulation Breakthrough

SB 53 Passes, Setting National Precedent

California's legislature passed Senate Bill 53, the nation's first comprehensive AI safety legislation targeting large-scale models. The bill now awaits Governor Newsom's signature, and the implications extend far beyond California's borders.

SB 53 requires companies with over $500 million in annual revenue to publicly disclose safety protocols, undergo independent audits starting in 2030, and report critical AI incidents to state authorities. The bill also establishes CalCompute, a public cloud computing cluster to democratize AI access.

What's significant here is the backing. Anthropic became the first major AI company to support the legislation, stating that SB 53 would formalize existing voluntary commitments while maintaining competitive dynamics. This represents a strategic calculation—better to shape regulation than fight it.

The bill emerged from recommendations by AI experts convened by Governor Newsom after he vetoed a broader safety bill last year. This iterative approach suggests California has found a regulatory sweet spot that balances innovation with accountability.

Senator Scott Wiener, SB 53's author, emphasized the bill's measured approach: "This legislation is a small but important first step in making AI companies accountable by making many of these voluntary commitments mandatory".

Federal Investigation Into AI Chatbots

FTC Launches Child Safety Inquiry

The Federal Trade Commission issued orders to seven major AI companies—including Alphabet, Meta, OpenAI, Snap, Character Technologies, and xAI—demanding information about how their chatbots affect children and teenagers.

This investigation follows several high-profile lawsuits, including a case where parents allege ChatGPT contributed to their 16-year-old son's suicide. The FTC wants to understand what steps companies have taken to evaluate chatbot safety, limit use by minors, and inform parents about risks.

FTC Chairman Andrew Ferguson framed the inquiry carefully: "As AI technologies evolve, it is important to consider the effects chatbots can have on children, while also ensuring that the United States maintains its role as a global leader in this new and exciting industry".

The timing is deliberate. OpenAI announced new parental controls for ChatGPT just days after the inquiry launched, including features that notify parents when the system detects their teen is in "acute distress". These controls allow parents to link accounts, disable features like memory and chat history, and set usage restrictions.

The Anthropic Copyright Settlement Rejection

$1.5 Billion Deal Faces Judicial Scrutiny

Judge William Alsup rejected Anthropic's record-breaking $1.5 billion settlement with authors over copyright infringement, calling the terms incomplete and expressing concern that class members would "get the shaft".

The settlement covered approximately 500,000 books at $3,000 per work—a fraction of potential statutory damages that could reach $150,000 per work for willful infringement. Judge Alsup wanted clarity on the exact scope of works involved, the claims process, and how class members would be notified.

This rejection has broader implications for the AI industry. While Anthropic's use of books for training was deemed "fair use," their downloading of pirated copies was not protected. This establishes a clear precedent: AI companies can use copyrighted material for training, but they must acquire it legally.

The settlement's failure suggests courts won't rubber-stamp deals that prioritize corporate convenience over plaintiff rights. For AI companies, this means higher legal costs and more scrutiny of training data acquisition practices.

Global AI Expansion Accelerates

Major Platform Updates Roll Out Worldwide

Several significant platform expansions happened this week, signaling AI's global acceleration:

Google AI Mode expanded to five new languages—Hindi, Indonesian, Japanese, Korean, and Brazilian Portuguese—opening access to hundreds of millions of new users. This follows AI Mode's expansion to 180 countries in English last month.

Apple Intelligence launched new features across iOS 26, iPadOS 26, and macOS Tahoe 26, including Live Translation in Messages, FaceTime, and Phone calls, plus enhanced visual intelligence for on-screen content. The update expanded language support to include Chinese, Danish, Dutch, Norwegian, Portuguese, Swedish, and Turkish.

Samsung AI Forum 2025 brought together leading researchers to discuss "Vertical AI Strategies and the Future of the Semiconductor Industry," with keynotes from Yoshua Bengio and Stefano Ermon. The forum highlighted Samsung's integration of AI across operations and semiconductor development.

These expansions represent more than feature updates—they're infrastructure buildouts for a globally integrated AI ecosystem.

What This All Means Going Forward

The New Power Structure Emerges

This week's developments reveal three fundamental shifts reshaping the AI landscape:

Infrastructure Verticalization: The NVIDIA-OpenAI partnership signals a move away from horizontal cloud providers toward vertical integration. Companies are building dedicated AI infrastructure rather than renting generic compute. This trend will accelerate as AI workloads become more specialized and demanding.

Regulatory Maturation: California's SB 53 and the FTC investigation represent a new phase of AI governance—targeted, expert-informed regulation rather than broad prohibitions. Expect similar approaches to spread nationally and internationally as policymakers gain technical sophistication.

Global Competition Intensifies: The simultaneous platform expansions by Google, Apple, and Samsung reflect an acceleration in the race for global AI dominance. Language support and regional deployment are becoming key differentiators as companies compete for international markets.

The week's events also highlight a more subtle but crucial development: the professionalization of AI risk management. Companies are moving beyond voluntary commitments toward formal compliance frameworks, driven by both regulatory pressure and competitive advantage.

Looking ahead, three trends bear watching: First, expect more vertical integration deals as AI companies seek guaranteed access to specialized infrastructure. Second, watch for additional states to introduce California-style AI safety legislation. Third, monitor international responses to U.S. AI infrastructure investments—other nations won't passively watch American companies dominate global AI systems.

The AI industry is no longer in its experimental phase. This week marked the beginning of its institutional phase, complete with mature partnerships, comprehensive regulation, and global competition. The companies and jurisdictions that adapt quickest to this new reality will shape the next decade of technological development.

The infrastructure is being built, the rules are being written, and the global competition is accelerating. What happens next will determine not just the future of AI, but the future of technological power itself.