Broadcom reportedly won't build OpenAI's custom chip unless Microsoft buys 40 percent of them

Broadcom Won't Build OpenAI's Custom Chip Unless Microsoft Buys 40% – Here's Why This Shifts the Future of AI

Imagine you want a custom superfast racing car built, but the manufacturer says they will only build it if a big sponsor agrees to buy almost half the production line. That is exactly what is happening in the AI chip world right now, according to a startling report from the-decoder.com published on May 9, 2026.

The story is huge: Broadcom reportedly won't build OpenAI's custom chip unless Microsoft buys 40 percent of them. This single demand sends shockwaves through the entire AI industry. It tells us that the companies that make the physical brains of AI are now in the driver's seat, and they are demanding financial guarantees before they spend billions on factories. What does this mean for the future of AI, for your business, and for the tools we all use? Let's break it down.

The Story in Simple Terms

OpenAI, the company behind ChatGPT, wants a custom chip designed just for its AI models. This is a common trend in AI – companies like Google and Amazon already build their own custom chips to save money and get better performance. OpenAI wanted to hire Broadcom, one of the world's biggest custom chip designers, to make this dream a reality.

But Broadcom reportedly pushed back hard. The condition, as reported, is stark: Broadcom will not build OpenAI's custom chip unless Microsoft agrees to buy 40 percent of the chips before they are even made. This is a huge demand. It means Broadcom sees OpenAI's request as risky. They want a sure-fire buyer – Microsoft – to guarantee that the chip will actually be used and paid for before the manufacturing lines even start running.

What This Tells Us About the Future of AI Hardware

This single news article reveals several deep truths about where AI is heading. It is not just about one chip deal; it is about the entire balance of power in the AI world moving from software to hardware.

1. Chipmakers Hold the Power Now

For years, AI companies acted like the kings of the hill. They built the best models, attracted billions in funding, and seemed unstoppable. But they needed computer chips – lots of them. Companies like NVIDIA, Broadcom, and TSMC make the chips. The demand for these chips is so incredibly high that the chipmakers can now set serious terms.

Broadcom's demand shows they are no longer just a supplier. They are a partner with veto power. They are basically saying, "We will only help if you show us the money and the market demand." This flips the script. Instead of AI companies telling chipmakers what to build, chipmakers are telling AI companies, "Prove your business works, or we will not build for you."

2. The Cost of Custom Chips Is Astronomical

Why does Broadcom need a 40% pre-purchase guarantee? Because building a custom chip is insanely expensive. It costs hundreds of millions of dollars just to design it, and then billions to fabricate it in a factory (called a fab). If OpenAI's chip fails, or if Microsoft does not like it, Broadcom would be left holding worthless silicon. That is a risk no company wants to take.

By demanding Microsoft buy 40% of the chips upfront, Broadcom is essentially looking for a co-sponsor. Microsoft is already deeply invested in OpenAI (having put billions into the company). This demand makes Microsoft an even bigger stakeholder in the chip itself, not just the models. This ensures that the chips will find a home inside Microsoft's Azure data centers, powering Microsoft's AI services for its own customers.

3. The "Build or Buy" Decision Is Changing

For a long time, the debate was, "Should we use NVIDIA's chips or build our own?" But build-your-own is becoming so hard and risky that only the biggest companies can even consider it. And even then, they need a partner like Broadcom. This demand signals that the barriers to entering the custom chip business are getting higher, not lower.

This will slow down the move to custom chips. Many AI startups that dreamed of having their own chip will now have to think twice. They might have to rely on off-the-shelf chips from NVIDIA or AMD for much longer than they hoped. The custom chip dream is becoming reserved for the absolute giants – companies like Microsoft, Google, and Amazon – and even they face tough negotiations.

What This Means for the Business and Society

The ripple effects of this single negotiation will be felt by every business that uses AI, every developer who builds AI apps, and eventually, every person who uses an AI tool.

For Businesses Using AI Tools

If custom chips become harder to build, then the cost of AI computing (often called compute) might not drop as fast as we hoped. Custom chips are often far cheaper to run than general-purpose ones. If OpenAI cannot get its custom chip made, and Microsoft is forced to buy 40% of them anyway, the cost savings might not materialize for a long time.

This means your AI usage costs might stay high. If you are a business that uses a lot of AI (customer support bots, content generation, data analysis), you should expect that the price you pay per hour of AI computation might remain expensive. The golden age of cheap AI compute might be delayed because the hardware supply chain is demanding guaranteed profits.

For the Future of OpenAI and Microsoft

This deal, if it goes through, ties Microsoft and OpenAI together even tighter. Microsoft already has a huge investment in OpenAI. Now, with Microsoft being forced to buy 40% of OpenAI's custom chips, Microsoft becomes a direct customer of OpenAI's hardware. This could lead to a situation where Microsoft has a huge say in how the chip is designed and used.

This is good for stability but bad for competition. It concentrates power. OpenAI gets its chip, Microsoft gets its guaranteed supply, and Broadcom gets its risk-free factory run. But for everyone else – any other AI company that wants to compete with OpenAI – they are now locked out of a potentially better, cheaper chip. The AI economy becomes a "walled garden" owned by a few giants.

For Society and Regulation

When a few companies control both the AI models (like GPT) and the AI chips (like this custom one), it raises serious questions about fairness and monopoly power. Regulators are already watching Microsoft and OpenAI. A deal like this, where Microsoft demands a 40% chip pre-purchase, could be seen as anti-competitive.

Imagine a world where the most advanced AI models can only run on chips that are pre-sold to a single cloud company. It would mean you have to use Microsoft's cloud to access the best AI. This would stifle innovation from smaller cloud providers and AI startups. Society would lose out on the diversity of ideas that comes from many players competing.

Actionable Insights for You

How should you react to this news? Whether you are a business leader, a developer, or just an AI enthusiast, here are practical steps to take:

Diversify your AI providers. Do not put all your eggs in the OpenAI basket. Because if the chip deal falls apart, OpenAI's ability to serve customers cheaply might be hurt. Consider using models from Google (Gemini), Anthropic (Claude), or open-source models that run on any hardware.
Watch the cost of compute. If you are building a product that relies heavily on AI, start budgeting for higher computing costs for the next 12-18 months. The chip shortage is not over, and this negotiation shows that making chips is getting harder, not easier.
Bet on flexibility over custom hardware. For most businesses, building a custom chip is a fantasy. Focus on using standard chips (like NVIDIA or AMD) and optimizing your software to run well on them. The real advantage will come from smart software, not from having a rare custom chip.
Watch for regulatory action. If you care about open markets, keep an eye on the Federal Trade Commission and European Union. Large deals like this often trigger investigations. If you are a competitor, this is a good time to complain about unfair practices.
Invest in efficiency. The best way to handle high chip costs is to use less of them. Focus on making your AI models smaller, faster, and more efficient. Technologies like pruning and quantization (fancy ways to shrink AI models) will become more important than ever.

Conclusion: The Age of Hardware Reality Is Here

Broadcom's demand that Microsoft pre-buy 40% of OpenAI's custom chips is a wake-up call. It marks the end of an era where AI companies could simply demand the hardware they want. Now, the hardware companies demand proof of demand, guaranteed revenue, and powerful partners before they will lift a finger.

The future of AI will be shaped by these chip deals, not just by clever algorithms. The companies that can afford to guarantee chip sales (like Microsoft) will have a massive advantage. The rest of us will benefit from more stable, reliable hardware, but we will also face higher costs and less choice.

For the next few years, the smartest move is to stay flexible. Use multiple AI providers, invest in making your code efficient, and prepare for a world where the most advanced AI chips are only available to a select few. The AI revolution is no longer just about writing great code – it is about who can afford to build the physical brains that run it.

TLDR: Broadcom reportedly refuses to build OpenAI's custom AI chip unless Microsoft commits to buy 40% of them upfront. This shifts power from AI software companies to chip manufacturers, raises barriers to custom hardware, and signals that AI computing costs may remain high for many businesses. The move ties Microsoft and OpenAI even closer together, concentrates market power, and highlights that the future of AI will be increasingly determined by hardware supply chain dynamics rather than just model innovation.