OpenAI and AMD announced a multi-billion dollar partnership on 6 Oct that sent AMD’s stock soaring 35%.
The deal: OpenAI will deploy 6 gigawatts of AMD’s Instinct AI chips over multiple years, starting with 1 gigawatt of the upcoming MI450 GPUs in late 2026. In exchange, AMD has issued OpenAI warrants for up to 160 million shares, roughly 10% of the company. But, this comes on condition if shares of the AMD rise to 3x of the current price.
This deal, significant beyond the headline numbers: kind of adds one more cog to a circular investment loop in AI infrastructure that’s starting to look less like a supply chain and more like a self-sustaining AI cycle.
The AI cycle, in which Nvidia, which recently invested $100 billion in OpenAI. Through its $100 billion OpenAI investment and OpenAI’s subsequent AMD deal, Nvidia now indirectly holds a small stake in AMD, one of its primary GPU competitors.
We also saw recently Nvidia investing $5 billion in Intel and roughly taking 4% of the company.
The Deal: What OpenAI Actually Gets
The headlines scream “10% stake in AMD,” but the reality is far more conditional:
The actual structure:
- AMD issued OpenAI a warrant for up to 160 million shares (roughly 10% of the company) at one cent per share
- These shares vest in tranches tied to two conditions: deployment volume AND AMD’s stock price
- First tranche: 1 gigawatt of AMD Instinct MI450 GPUs deployed (starting late 2026)
- Final tranche: 6 gigawatts deployed AND AMD stock hits $600 (currently around $207)
Translation: OpenAI doesn’t get 10% of AMD just for signing a contract. They get it by:
- Actually deploying 6 gigawatts worth of AMD AI chips over multiple years
- Waiting for AMD’s stock to triple from its current price
- Building out infrastructure that won’t be fully operational until years from now
OpenAI commits to buying billions of dollars worth of AMD chips over multiple hardware generations, and AMD compensates them with deeply discounted stock warrants that only pay out if the deployment succeeds and the stock performs.
OpenAI’s Infrastructure Build out
The AMD deal comes less than two weeks after OpenAI announced a landmark $100 billion equity and supply agreement with Nvidia. That arrangement combined capital investment with long-term hardware supply, with Nvidia taking an ownership stake in OpenAI while committing to supply 10 gigawatts of computing infrastructure.
Together, the Nvidia and AMD deals account for 16 gigawatts of OpenAI’s broader 23-gigawatt infrastructure roadmap. At an estimated $50 billion in construction costs per gigawatt, OpenAI has committed roughly $1 trillion in new build out spending across the past two weeks. As a result of these agreements, Nvidia may indirectly be able to own ~1% of AMD through OpenAI
OpenAI also recently signed a $300 billion cloud deal with Oracle, under which it will purchase computing power over approximately five years. Additionally, the company is in talks with Broadcom to develop custom chips for its next generation of models.
Market Impact and Industry Context
Nvidia shares fell 1% in premarket trading Monday following the AMD announcement. The partnership could help ease industrywide pressure on supply chains and reduce OpenAI’s reliance on a single chip vendor, though Nvidia remains OpenAI’s primary GPU supplier.
The technical angle that matters: This circular structure where major chip manufacturers, AI companies, and cloud providers are simultaneously customers, suppliers, and equity partners creates a tightly wound ecosystem where each participant’s growth depends partly on the others’ success.
Supply Diversification or Ecosystem Risk?
OpenAI’s move to secure chips from multiple suppliers reflects broader industry concerns about GPU availability and supply chain concentration. By partnering with both Nvidia and AMD, OpenAI gains negotiating leverage and reduces dependence on any single vendor.
However, the equity-linked structure of these deals where chipmakers take ownership stakes in exchange for supply agreements ties the financial health of suppliers directly to their customers’ growth. If AI compute demand slows, it affects not just OpenAI’s revenue, but also the value of the equity stakes that AMD and Nvidia hold.
Industry analysts note that while diversification typically reduces risk, the circular nature of these arrangements may concentrate it instead. Every major participant in the AI infrastructure stack is now both invested in and dependent on the continued growth of AI compute demand.
The Contrast: Open Source Is Thriving
For contrast, open weight AI models (dominated by Qwen, Z.ai, and more) are doing great as well.
- GLM-4.6 just matched Claude Sonnet 4 on coding benchmarks
- Chinese video generation models are approaching commercial quality
The irony: open-source models keep closing the performance gap with frontier models, but the closed AI economy keeps concentrating capital in circular deals that assume only a handful of companies can compete.
If GLM-4.6 can deliver 70%-80% of Claude Sonnet 4 performance at a fraction of API cost, what happens to OpenAI’s revenue projections? If open models keep improving at current rates, who’s paying $300 billion for Oracle cloud infrastructure?
What’s Next
The first AMD chips under this agreement won’t arrive until late 2026, with the full 6-gigawatt deployment spanning multiple years and future chip generations beyond the MI450. OpenAI’s ability to fully exercise its warrant depends on both executing this massive infrastructure build out and AMD’s stock performance.
AMD’s stock would need to sustain significant growth for OpenAI to realize the full 10% stake, requiring not just successful chip deployments but continued market confidence in both AMD’s roadmap and AI infrastructure demand.
The partnership marks a significant win for AMD in the AI chip market, where Nvidia has maintained dominant market share. For OpenAI, it represents a multi-billion-dollar bet on AMD’s ability to deliver competitive AI hardware at scale while diversifying its supply chain beyond a single vendor.
Leave a Reply
You must be logged in to post a comment.