The big players burn money to build AI, but all the funds end up in memory storage warehouses—this storyline is way more exciting than a PowerPoint presentation.

null

Silicon Valley tech giants are footing an extremely unusual bill.

With every new product launch, the PowerPoint slides repeatedly emphasize: how many times the number of transistors has increased, how much computing power has surged. Everyone is celebrating the intelligence of this “brain.”

But upon carefully examining this computing power bill, you’ll notice a noteworthy misalignment: the total industry expenditure has more than doubled, yet the budget share allocated to the “brain” has hardly increased.

According to the latest breakdown data from the renowned research firm Epoch AI, from early 2024 to the end of 2025, the total expenditure on AI chip components worldwide skyrocketed from $22 billion to $52 billion. However, the cost share of the “computing brain” (the main logic bare chip), representing the most advanced manufacturing process, has remained steady at around 13% to 14%.

This means that what truly drains the pockets of tech giants is the “memory” warehouses—