Altman’s Dealmaking Gambit Building OpenAI’s Infrastructure Kingdom

When Sam Altman speaks of OpenAI’s deal strategy these days, it reads less like corporate PR and more like a blueprint for empire building. In recent interviews (on The A16z Podcast and with Ben Thompson), he described the company’s relationships with AMD, Nvidia, Oracle, CoreWeave not as discrete contracts but as strands in a sprawling infrastructure web

Altman confirmed that OpenAI’s new deal with AMD  promising to deploy as much as six gigawatts of AMD GPUs and granting the option to acquire up to 10 percent of the chipmaker is not an isolated move but a deliberate acceleration of its vertical integration. He admitted that earlier he resisted the idea of controlling supply chains, but now sees it as essential to fulfilling OpenAI’s mission These deals are risky. OpenAI posted a net loss of $13.5 billion in early 2025 versus $4.3 billion in revenue. Some analysts warn that certain deals are “circular” that is, OpenAI’s deals are partly financed by its own partners (for example, Nvidia investing in OpenAI while also selling it chips) Still, Altman insists the math will work out. He argued that OpenAI’s revenue can absorb these deals and that the company even plans to help fund infrastructure investments for others in the ecosystem. He sees a future where OpenAI isn’t just a software and model provider, but also a backbone of cloud-scale AI compute The broader narrative becomes clearer when you look at OpenAI’s alliance map. With Nvidia, OpenAI secured a $100 billion commitment tied to chip access. With Oracle, OpenAI is collaborating on data center infrastructure under massive long-term contracts. CoreWeave, a GPU cloud provider, has a multiyear contract plus equity investment from OpenAI

What’s happening is this: OpenAI is turning itself into not just a buyer of compute, but a stakeholder in compute supply. It’s repositioning its role in the ecosystem, forcing chip makers, cloud providers, and hardware vendors to lock arms with it. If this succeeds, OpenAI won't just compete it may control key parts of the stack

From a journalist’s vantage, a few points stand out

1. Ambition on a grand scale This is not incremental expansion. OpenAI is chasing scale in hardware, data centers, finance, and partnerships that rival national infrastructure projects

2. Fragility under scrutiny With large losses and heavy leverage, these deals will be under pressure. Delays in chip rollout, regulatory roadblocks, or slower than expected demand could unravel parts of the strategy

3. Regulation as a wild card Governments will scrutinize companies that combine model development, infrastructure control, hardware ownership, and capital flows. Antitrust, data privacy, export controls: all could complicate this vertical approach

To bring an analogy: it’s as if a leading software company decided to own the server farms, cable lines, chip factories, and even some ISPs. The advantage is full control the risk is overextension and regulatory backlash

Altman is placing a huge bet. If OpenAI can pull off integrating compute, capital, and AI models under one roof, it will redefine tech hierarchy. If parts of it crack  say, AMD deployment misses schedule or costs balloon the fallout could be severe. In short: we’re watching the architecture of the next tech superpower being drafted before our eyes