Key Points:
- Tech giants are building their own chips to reduce reliance on Nvidia.
- The strategy mimics IBM’s vertical integration model from the 1960s.
- Google, Microsoft, and Amazon are also buying underground fiber cables.
- Analysts warn that the high cost of chip design limits success.
The world’s biggest technology companies are looking to the past to secure their future. Alphabet, Meta, Microsoft, and Amazon are currently racing to build their own custom artificial intelligence chips. This shift in strategy mirrors the business model that made IBM a global powerhouse in the 1960s.
Known as “hyperscalers,” these companies are rushing to fill their data centers with their own hardware. Google is currently leading the pack with its Tensor Processing Units (TPUs) and is reportedly talking about selling them to Meta. This puts the tech giants in direct competition with Nvidia, the current king of the chip market. Analysts at Bloomberg Intelligence predict this new market for custom AI chips could grow to $122 billion by 2033.
The push for control goes beyond just computer chips. Microsoft and Amazon are also investing heavily in “dark fiber,” which is underground fiber-optic cables that are not yet in use. By owning the chips, the cables, and the software, these companies are embracing “vertical integration.” This is the same strategy oil tycoons used in the 19th century, and IBM used to dominate the mainframe computer era.
The driving force behind this change is the high cost of doing business. Since ChatGPT launched in 2022, the demand for AI has skyrocketed. Nvidia’s chips are expensive and often in short supply. By making their own parts, tech giants can save money and tailor the hardware to their specific software needs.
Paul Roberts from Amazon Web Services noted that their new Trainium3 chip offers better performance for a lower price compared to standard options. He also highlighted that specialized processors are far more energy-efficient, which is critical as data centers consume massive amounts of electricity.
However, experts warn that this path is dangerous. Jay Goldberg, a market analyst, believes the trend is reaching its limit. Designing cutting-edge chips requires a massive amount of money. Goldberg warns that only a few companies can afford the expense, meaning not everyone who tries to copy IBM’s old playbook will succeed.
