A new crack in the AI storage boom: customers are finding ways to break free from memory dependence

```

The memory chip industry is experiencing the strongest boom in history, but the same force driving record-breaking profits is also nurturing the next wave of technological transformation.

Micron's latest financial report shows that memory chip supplies have tightened further compared to three months ago. The company expects operating profit this quarter to exceed the highest annual revenue in its history and has signed long-term supply agreements with 15 new customers, extending expectations of supply-demand tension beyond 2027.

However, the continued rise in memory prices is altering incentives along the AI industry chain. As HBM becomes an increasingly expensive cost in AI systems, major customers such as Qualcomm, NVIDIA, and Cerebras have begun exploring new architectures and algorithms to reduce reliance on HBM.

This means that the high prices underpinning the current memory bull market may, in the long run, also serve as a catalyst for innovation on the demand side.

The More Expensive Memory Becomes, the More Incentive Customers Have to Reduce Usage

The supply-demand imbalance is continually strengthening memory manufacturers' bargaining power and forcing downstream customers to accept higher costs.

On Thursday, Apple unusually raised the prices of several Mac and iPad products between product release cycles, explicitly attributing the reason to rising memory chip costs.

Micron stated that most of its newly signed long-term agreements last up to five years and include price floors, meaning that even if the sector enters a downturn, prices will remain well above previous lows. Futurum analyst Rolf Bulk said that even at the minimum contract price, profitability exceeds peaks in past industry boom cycles.

Although historically, long-term chip supply contracts have not always been strictly enforced, the current market consensus is that supply is unlikely to improve significantly over the next two to three years, meaning customers have virtually no bargaining power in the short term.

However, precisely because supply remains tight and prices keep rising, more and more tech companies are considering another solution—not to compete for more HBM, but to minimize HBM usage.

Tech Giants Begin to Explore the “Less Memory” Route

This week, Qualcomm highlighted its "High-Bandwidth Compute" architecture at an investor meeting, aiming to reduce reliance on HBM in AI computation through new system designs. Previous reports indicate that NVIDIA is adjusting some designs in its next-generation Vera Rubin platform to lower total memory requirements.

AI chip firm Cerebras positions "not using HBM" as a key selling point. In its first earnings report since going public this week, the company stated that its wafer-scale chip completely eliminates the HBM constraint. CEO Andrew Feldman bluntly said: “HBM is in tight supply and expensive, but we simply don’t use it.”

At the same time, algorithm-level optimizations are also advancing. In March, Google released its TurboQuant research, which uses a new model compression method to significantly reduce memory consumption of AI models with minimal performance impact. When this was announced, Micron’s stock price temporarily plunged nearly one third, though it rebounded more than two-fold as the market reassessed. The incident showed that the market is highly sensitive to any technical breakthrough that might weaken HBM demand.

Prosperity Is Giving Birth to the Next Wave of Competition

In the medium to short term, HBM remains an irreplaceable key component for AI computing, with Micron’s observed supply-demand tension solidly supported by fundamentals.

But in longer industry cycles, price is often the most powerful driver of innovation.

When the world's largest AI customers are willing to sign unprecedented five-year long-term supply contracts, it also means they have both the time window and strong economic incentives to invest in new chip architectures, software algorithms, and system designs to do more computing with less memory.

For the memory industry, today’s strongest competitive advantage—tight supply, high prices, and unprecedented bargaining power—has reinforced the current boom, but may also become the biggest motivation for customers to break free of HBM dependence. Historically, every time core hardware prices kept rising, it spurred new technological directions; and this time, the industry’s prosperity itself may be the starting point for the next round of competition.

Risk Warning and DisclaimerThe market has risks and investments require caution. This article does not constitute personal investment advice, nor does it take into account individual users’ specific investment objectives, financial situations, or needs. Users should consider whether any opinions, views, or conclusions in this article are suitable for their circumstances. Invest accordingly at your own risk. ```