Andrey Golub writes on LinkedIn: The current AI ecosystem is unsustainable, driven by a heavy reliance on Nvidia and its GPU technology. While Nvidia thrives, with smaller manufacturers attempting to carve out a share of the $75 billion infrastructure market, the companies that build AI applications generate significantly less revenue—around $5 billion. This imbalance reveals a structural flaw: applications, not infrastructure, should drive economic growth to maintain a healthy tech ecosystem.

Historically, applications were the primary source of revenue, which, in turn, fueled infrastructure development. Today, the inverse has occurred, stalling innovation and making the system fragile. To address this issue, AI developers must reduce the costs associated with inference and create their own vertically integrated technology stacks, much like how Apple revolutionized the smartphone industry with its custom chips.

Some companies have already taken steps in this direction. Li Kaifu’s startup, 01.ai, for example, trains its AI models on custom hardware, slashing costs significantly. Such innovations are essential for AI to transition from being a niche automation tool to becoming the “brain” that drives core business strategies.

If AI is to remain economically viable and sustainable in the long term, companies must prioritize the development of specialized, energy-efficient chips and a restructuring of the current revenue model to focus on applications rather than the supporting infrastructure.

https://lnkd.in/djmbmNvH