As demand for AI infrastructure surges, memory chip prices are rising rapidly, increasing pressure on cloud providers’ profit structures. AI cloud company CoreWeave plans to respond to price volatility with a risk management strategy that combines long-term supply contracts and financial techniques.

Cryptopolitan, a blockchain media outlet, reported on July 15 that as the spread of AI services makes memory a key bottleneck resource, cloud companies must manage not only GPUs but also memory procurement and price volatility.

The shift is also evident in memory makers’ performance. Micron posted revenue of $41.46 billion in its fiscal 2026 third-quarter results announced on June 24. That was well above the previous quarter’s $23.86 billion and $9.3 billion a year earlier. Revenue from its cloud memory business was $13.77 billion, and the companywide gross margin was 84.6 percent.

Micron CEO Sanjay Mehrotra (산제이 메흐로트라) said, "In the AI era, memory has become one of the most strategic assets." He assessed that the expansion of AI infrastructure is fundamentally changing the memory industry’s profit structure.

Market research firm TrendForce also forecast that a supply shortage of high-bandwidth memory, or HBM, used in AI servers will persist for some time. In a recent report, TrendForce defined the shortage of memory for AI servers as a “New Normal”. It said the shortage is unlikely to be resolved easily as long as AI demand continues to outstrip production capacity. The market is also raising the possibility that Samsung Electronics’ next-generation HBM4 could expand its early market share as certification procedures speed up.

Morgan Stanley recently analysed that expanding AI demand is triggering “chipflation” in the memory industry. It said pricing structures themselves are changing as memory makers focus on producing higher-margin AI products rather than general-purpose ones. The change puts a bigger burden on AI cloud companies that sign multi-year contracts and provide services. Memory accounts for a large share of server costs, but it is difficult to pass sudden cost increases directly on to customers.

CoreWeave is putting forward a conservative investment strategy to respond to these risks. CEO Michael Intrator (마이클 인트라토어) said in a shareholder letter in April, "We were risk managers before we were cloud builders," emphasising a strategy of expanding infrastructure only after confirming actual customer demand.

The strategy has also led to growth in scale. CoreWeave said it is the fastest cloud services company to reach annual revenue of $5 billion. Revenue growth was 168 percent from a year earlier, and it currently operates more than 850 MW of infrastructure across 43 data centres. Its order backlog rose to $66.8 billion from $15 billion, and its average contract term lengthened to 5 years from 4 years.

It also improved its financial structure. CoreWeave raised about $18 billion in combined debt and equity over the past year and lowered its weighted average borrowing cost by more than 300 basis points, or 3 percentage points. The company estimated it could cut financing costs by about $700 million a year. Intrator said it built a new business model that combines long-term customer contracts with its financing structure.

A key example is its long-term contract with Meta. CoreWeave signed an AI infrastructure supply deal worth about $21 billion with Meta Platforms in April that runs through 2032. It also gained access to early volumes of Nvidia’s next-generation AI platform, Vera Rubin. The company said long-term contracts can increase the predictability of hardware investment plans.

CoreWeave is also considering ways to manage memory price swings themselves through financial techniques. Reuters reported that CoreWeave is reviewing a hedging strategy using derivatives such as put options while also signing long-term contracts with major suppliers including Micron and SanDisk.

Long-term contracts have the advantage of securing needed parts during supply shortages, but they also carry the risk of having to buy at higher prices if market prices fall later. CoreWeave is examining ways to reduce this price volatility risk through financial hedges.

Discussions are still at an early stage, however. It has not executed any actual hedging trades so far. Even so, the industry is assessing that a new era is opening in which AI infrastructure companies manage key component prices from a financial risk perspective, like airlines and energy companies.

Ultimately, memory procurement is expanding beyond simple parts purchasing into the realm of capital management strategy. As demand for AI memory continues to rise, the competitiveness of AI infrastructure companies is expected to be decided not only by technical capabilities but also by financial strategies for responding to price volatility.