OpenAI wants $100B to build its own chip empire

If you cannot buy enough GPUs from Nvidia, why not just raise $100 billion and build your own chip foundries? That sounds wild, but Sam Altman might actually try to pull it off.

According to a Bloomberg report, OpenAI CEO Sam Altman is currently in discussions with venture capital firms and manufacturing partners to raise roughly $100 billion. The goal is to build a global network of semiconductor fabrication plants specifically to supply chips for generative AI workloads.

Project Tigris and the Middle East connection

Building semiconductor fabrication plants is notoriously expensive. Advanced lithography tools cost fortunes, and setting up cleanrooms requires staggering capital.

Altman has reportedly met with Abu Dhabi-based AI group G42 as well as Japan’s SoftBank Group. OpenAI is eyeing between $8 billion and $10 billion from G42 alone, following an earlier partnership established in October to expand AI services across the Middle East.

Back in December, reports surfaced that OpenAI was seeking new funding at a massive scale, though its intended use was unclear at the time. Now we know it likely ties back to this chip initiative, codenamed Tigris, aimed directly at breaking Nvidia's stronghold on AI hardware.

Going far beyond Big Tech's playbook

Other tech giants are already designing custom silicon to dodge the ongoing GPU crunch. Microsoft, AWS, Google, and IBM all have proprietary AI accelerator chips in production.

However, all of them strictly follow the fabless model. They design the chips in-house and leave the actual manufacturing headaches to established foundries like TSMC.

OpenAI, on the other hand, wants to do it all. The reported blueprint spans chip design, manufacturing, and global distribution. According to xResearch, the AI chip market was valued at $14.9 billion in 2022 and is projected to expand at a 40.5% CAGR to reach $227.6 billion by 2030, meaning existing foundry lines may never keep up with appetite.

Why it matters

We think this is one of the most audacious hardware gambits in modern tech history, and it carries enormous risks alongside huge ambition.

On the smart side, OpenAI knows that compute is its ultimate bottleneck. Depending entirely on Nvidia's margins and allocations puts OpenAI's entire future at the mercy of outside suppliers. Securing dedicated foundry capacity would give OpenAI an unassailable defensive moat and drastic cost advantages.

Yet the practical obstacles are daunting. Manufacturing cutting-edge silicon takes decades of institutional know-how, precise yield optimization, and specialized talent that raw cash cannot simply purchase overnight.

For the broader industry, this sends an unmistakable signal: foundation model builders are unwilling to stay passive buyers forever. Even if Altman's venture takes years to deliver a single functional wafer, it alters the leverage between hyperscalers and chipmakers permanently.