Nvidia blew past profit expectations on Wednesday, proving its absolute dominance in the technology sector. Its graphics processing units (GPUs) currently power most of the artificial intelligence boom. However, the landscape is shifting. Major tech companies are busy building their own alternatives to break free from Nvidia’s grip.
Google, Amazon, and even OpenAI are actively designing custom chips known as ASICs. These “application-specific” circuits offer a smaller, cheaper, and more accessible option for specific tasks. Daniel Newman, an analyst at the Futurum Group, predicts these custom chips will soon grow even faster than the traditional GPU market.
For years, GPUs were primarily for video games. That changed around 2012 when researchers realized these chips could handle parallel math much better than standard processors. This capability made them perfect for training AI models to learn from massive datasets. Nvidia seized this opportunity and now sells massive server racks—costing roughly $3 million each—to cloud giants and governments. The company says it currently ships about 1,000 of these powerful systems every week.
But GPUs are generalists. Think of them as a Swiss Army Knife; they do a lot of things well. As AI models mature, companies need chips that do one thing perfectly. This process, called inference, is where the AI actually applies what it learned to answer questions or solve problems. ASICs are hard-wired for this. They run efficiently and cost less over time because they don’t carry the extra baggage of a general-purpose chip.
The market is expanding in other ways, too. Companies like Qualcomm and Apple are pushing chips that run AI directly on your phone or laptop, rather than in the cloud. Meanwhile, Nvidia’s main rival, AMD, is fighting for market share by offering open-source software solutions to attract developers. While Nvidia remains the king of the hill, the race is no longer a one-horse show. Big Tech wants to own the hardware that powers its future, and for many, that means building it themselves.
