AMD's Chip Deal With OpenAI Triggers Explosive Rally | Bloomberg Tech 10/6/2025

Key Concepts

• AI Compute: The computational power required to run AI models, including training and inference.
• GPU (Graphics Processing Unit): Specialized electronic circuits designed to rapidly manipulate and alter memory to accelerate the creation of images, now widely used for AI workloads.
• Inference: The process of using a trained AI model to make predictions or decisions on new data.
• Gigawatt (GW): A unit of power, equivalent to one billion watts, used here to quantify the scale of AI computational capacity.
• MI450 Chip: AMD's next-generation AI accelerator chip.
• AI Accelerator Market: The market for specialized hardware designed to speed up AI computations.
• Supply Chain: The network of organizations, people, activities, information, and resources involved in moving a product or service from supplier to customer.
• AI Stack: The entire technological infrastructure supporting AI, from hardware to software and applications.
• Deprecated American Chip: Less advanced or older generation AI chips.
• Audience of One: A marketing concept where AI enables hyper-targeted advertisements tailored to individual users.
• Advantage Plus Shopping Campaigns: Meta's AI-powered advertising tools designed to optimize campaign performance.
• Northstar: A long-term strategic goal or vision for a company or project.
• Circular Financing: A concern regarding deals where the funding for a purchase might indirectly come from the seller, often implying a lack of genuine external demand.

---

AMD-OpenAI Historic Partnership and Market Impact

The broadcast opens with a focus on AMD's significant surge, hitting a record high with a 28% to 37% increase in stock value, adding $70 million to its market cap in one day. This was driven by a historic deal with OpenAI, projected to generate tens of billions of dollars in new revenue. The agreement involves OpenAI deploying AMD GPUs, specifically focusing on inference workloads, and grants OpenAI the right to acquire up to 160 million AMD shares in tranches, contingent on operational and financial milestones.

OpenAI's Vision and Demand for AI Compute

Greg Brockman, President of OpenAI, emphasizes the "huge milestone" for AMD and the AI industry, highlighting the critical need for "more AI compute." He states that compute is the foundation for AI intelligence and that the world continues to "underestimate the amount of demand for AI compute." Citing ChatGPT's 800 million active users (a product that "practically did not exist three years ago"), he notes that the lack of computational power hinders the launch of new features and products. OpenAI is embarking on a "massive buildout of six gigawatts of AI compute," starting with the first gigawatt in the second half of 2026, utilizing AMD's next-generation MI450 chip. This is described as an "all-in partnership" requiring years of collaboration on hardware, software, and supply chain elements.

Industry-Wide Effort and Infrastructure Needs

Both Lisa Su (AMD CEO) and Greg Brockman stress that this is an "industry-wide effort." Greg highlights that compute requires the entire supply chain to "wake up and start building much more than people were planning on," starting with energy. OpenAI plans to deploy AMD chips in its own data centers and with cloud service providers like Oracle. Lisa clarifies that for such a massive amount of compute (six gigawatts), deployment will occur in "multiple locations" and likely involve "multiple providers."

OpenAI's Financing and AMD's Confidence

Regarding OpenAI's financing, Greg states that "AI revenue is growing faster than I think almost any product in history," justifying the investment in compute power. OpenAI is exploring various financing methods, including equity, debt, and "creative ways," to build the necessary compute for an "AI-powered economy." Lisa expresses "full confidence in OpenAI" and Greg, calling it a "massive opportunity" for AMD. She asserts that the deal is a "win for AMD, OpenAI, and our shareholders," as OpenAI's chip purchases directly boost AMD's revenue and earnings, creating a "virtuous positive cycle."

U.S. Prioritization and Supply Chain

The discussion touches on the geographical build-out. Greg states OpenAI's preference is to "build as much as possible in the U.S.," viewing compute as a "national security strategic resource." However, they are not limiting themselves, acknowledging global demand. Lisa confirms AMD's meticulous work on its supply chain, deeply partnered with TSMC, and reiterates the priority of "building in the United States" for the "U.S. AI stack."

AMD vs. NVIDIA and Workload Diversity

Sam Altman's statement on X that the AMD deal is "incremental to what is being done with NVIDIA" is addressed. Greg explains that while AI training has a "huge cost" and has primarily been done with NVIDIA, inference offers an "easier area of entry." He praises AMD's MI450 series as an "incredible chip" and notes that a "diversity of chips" accelerates progress due to a "diversity of workloads" with different balances of memory and computational power. Lisa adds that OpenAI is the "ultimate power user" of AMD chips, providing valuable testing and confidence in the technology.

Investor Perspective: Tony Wang (T. Rowe Price)

Tony Wang views the deal as "exciting for AMD," providing a "big league customer to scale" and attracting others. He believes the AI market has been consistently underestimated, with significant "productivity use cases" and strong ROI, especially with contracts "backstopped by blue-chip companies like Microsoft or Oracle." He finds the deal's mechanics, where OpenAI can buy AMD shares contingent on milestones, to be a "win-win for both companies," fostering "codevelopment and partnership." Wang believes this deal helps AMD's narrative, moving it from a "second place player" to a credible partner for OpenAI, signifying immense demand for AI compute and benefiting the country. He expresses confidence in the long-term AI story, seeing "no better time to be a tech investor." Regarding debt financing, he notes that the "useful life for a GPU data center" is longer than previously thought (4-5 years, potentially 7-8 years), making debt a sensible option given the expanding use cases in robotics, diagnostics, and simulation.

White House Perspective: David Sacks (AI Czar)

David Sacks describes an ongoing "AI boom" driving businesses to new highs, attributing it to "President Trump's pro-innovation, pro-export, pro-AI policy," which he links to a 3.8% U.S. GDP growth rate. He emphasizes the importance of energy infrastructure, citing Trump's "drill, baby, drill" stance and policies supporting "behind the meter power generation" (allowing AI companies to generate their own power), new oil, gas, and nuclear energy. Sacks confirms the White House does not get involved in private company deal-making.

As a self-proclaimed "China Hawk," Sacks states the U.S. must win the AI race against China, which he considers the "main competition." He advocates for a strategy of "pro-innovation, pro-infrastructure, pro-energy, and pro-exports" to ensure the "U.S. technology stack dominates the world." His metric for winning the AI race is "global market share." He criticizes previous administrations for restricting chip sales to countries like the Gulf States, arguing this forces them to "turn to China and adopt the Chinese tech stack," leading to a repeat of the 5G scenario where China gained market share. Sacks argues for selling "deprecated American chips" to China, rather than nothing, to avoid accelerating China's desire for independence from the American stack. He points out that the Biden administration approved the H20 chip for export to China in unlimited quantities, albeit with licensing and a surcharge, and criticizes the hypocrisy of those who then attacked Trump for a similar stance. He also addresses the talent pipeline from China, acknowledging that "half of the world's AI researchers are from China" and the need to be open to working with this talent, while defending Jensen Huang's nuanced comments on the matter.

Meta's AI Transformation in Marketing: Alex Schultz

Alex Schultz, author of a new book on AI's impact on advertising, discusses the balance between "incremental growth" and a "Northstar" long-term vision. His book aims to be a useful guide for businesses on advertising in the new digital landscape. He highlights the concept of "audience of one," where AI enables hyper-targeted, personalized ads. While acknowledging data privacy concerns, he believes the balance is better than a decade ago, with clear laws and platform policies. He warns that the "biggest risk is not taking a minute of it," citing Europe's regret over lack of growth partly due to advertising restrictions.

Schultz reveals that AI has "completely transformed" Meta's core business. Five years ago, Facebook and Instagram content was mostly connected; today, the majority is "unconnected content" ranked by AI based on semantic understanding. This has significantly boosted engagement and revenue, with "very high double-digit percentage uplift" for advertisers adopting tools like Advantage Plus Shopping Campaigns. He emphasizes that creativity in advertising now extends beyond pixels to data, targeting, and conversion rate optimization, focusing on "giving people amazing personalized experiences and high conversion rates."

Market Analyst Perspective: Ipek Ozkardeskaya

Ipek Ozkardeskaya, Senior Market Analyst, calls the AMD-OpenAI deal "very amazing news," especially for those questioning the "circular nature of the business." She notes AMD's proactive approach in striking deals with data centers and chipmakers to ensure supply meets demand. She believes NVIDIA's dominance is not threatened, as AMD chips will "complement" rather than replace NVIDIA's. The "pie is big enough to feed everyone."

Ozkardeskaya asserts that OpenAI is now a "macro-level factor" for public market investors, being the "biggest startup in the world" with a $500 billion valuation and making "market moving deals." She acknowledges that AI is a "very capital-intensive play" but believes OpenAI will not have funding problems, given NVIDIA's investment and the eagerness of private and public investors. She attributes OpenAI's strong revenue growth potential to its "reach," "first-mover advantage," and continued popularity despite competition from other models like Grok and Perplexity.

Synthesis and Conclusion

The discussions underscore the unprecedented and rapidly expanding demand for AI compute, driving historic partnerships like the AMD-OpenAI deal. This collaboration, focused on deploying six gigawatts of AI compute for inference, signifies a major validation for AMD's technology and a critical step in scaling AI infrastructure. The market views these developments with immense optimism, recognizing the long-term productivity gains and economic transformation promised by AI.

Key takeaways include:

1. Exponential Demand for AI Compute: The need for computational power is vastly underestimated and growing rapidly, necessitating massive infrastructure build-outs.
2. Strategic Partnerships and Ecosystem Development: The AI industry requires deep collaboration across hardware, software, and supply chains, with companies becoming increasingly "tied to each other" financially and operationally.
3. U.S. Leadership and Global Competition: The U.S. aims to dominate the AI race, prioritizing domestic infrastructure while navigating complex export policies to maintain global market share against competitors like China.
4. AI's Transformative Impact: AI is fundamentally reshaping industries from marketing (hyper-personalized ads) to core business operations (content ranking, engagement), driving significant revenue growth and efficiency.
5. Financing the Future of AI: The capital-intensive nature of AI infrastructure necessitates diverse and creative financing strategies, with strong investor confidence in the long-term ROI.
6. Diversity of Hardware: While NVIDIA remains dominant, there's a growing recognition for the need and benefit of diverse chip architectures (like AMD's MI450) to address varied AI workloads, particularly in inference.