Will RAM and PC price maintain the current high?

Memory chip prices hold their current high through the end of 2026 and into 2027: the direction is still up, with the climb slowing rather than the level falling. Personal computer prices follow the same path, because memory is a rising share of what a machine costs to build and makers are absorbing it by trimming other parts. What keeps the high in place is data-centre demand that is funded and barely sensitive to price; what would end it is that funding being cut, or consumers buying so much less that memory capacity frees up.

The newest figures point up, not down. On 30 September 2026 TrendForce projected that DRAM contract prices — the prices set in bulk supply deals for the working memory a computer uses while it is switched on — would rise 10–15% in the fourth quarter of 2026 compared with the three months before (Trendforce). NAND flash, the chips that hold data when the power is off, is projected to rise 15–20% over the same quarter. Server memory is expected to remain undersupplied through the quarter, constrained by component supply and by limits on packaging, testing and wafer production.

Memory contract price growth, 3Q26 to 4Q26 (% QoQ; Evertiq, 2 Oct 2026, on TrendForce figures)

The chart plots the midpoint of each published range.

The price direction follows where makers put their factories. DRAM suppliers continue to give their most advanced production capacity to high-performance server products, which keeps the overall market short (Trendforce). PC memory supply could fall in 2027 as more capacity moves to server work. PC makers are still buying aggressively because they expect supply to stay tight in 2027, which supports prices. Consumer memory prices are expected to keep rising, though more slowly, with the big suppliers holding back production.

The money behind that demand is committed and mostly indifferent to the memory price. Evercore and Bank of America both put AI capital spending in 2027 above $1 trillion, with 2026 estimates rising to between $800 billion and $900 billion, CNBC reported on 30 April 2026 (Ai Boom Big Tech Capital Expenditures Now Seen Topping 1 Trillion In 2027). Moody's Ratings projected $785 billion of spending by the largest data-centre operators in 2026 and close to $1 trillion in 2027, up from about $700 billion in its March 2026 estimate, RCRTech reported on 18 May 2026 (Moodys Hyperscaler Capex). Meta raised its 2026 range to $125–145 billion from $115–135 billion on 29 April 2026 and attributed most of the increase to higher component costs, particularly memory prices.

Cheaper AI computing does not shrink this demand; it grows it, because a lower cost per unit of work tends to increase how much work gets done. One vendor advertises $42 per billion input tokens for its AI service, or $0.042 per million (Ai). That price is the vendor's own commercial claim and has not been independently checked.

Where supply falls short of what buyers need, the high-performance end takes the scarce chips and ordinary buyers get what is left. That is why enterprise solid-state drives — the storage inside data centres — are the only memory category whose price growth is expected to speed up, to 23–28% in the fourth quarter (Trendforce). Demand for them is projected to grow more than 80% in 2026 (Trendforce).

On the computer side the evidence is about cost pass-through rather than a published price tag. PC brands are cutting storage capacity in mainstream models to lower the total cost of parts per machine, which reduces both how many drives they buy and how large those drives are, so buyers get less storage at the same price (Trendforce). Notebook inventories are increasingly carrying higher component costs into the fourth quarter, which is weighing on sales.

Two things push the other way, and both are already visible. Rising memory costs and used-up cheap inventory are holding back smartphone production and reducing how much memory that market needs (Trendforce). NAND wafer prices have reached records that retail channels are increasingly unable to absorb. The funding is not unlimited either: Meta's free cash flow fell to $1.2 billion in the first quarter of 2026 from $26 billion a year earlier, CNBC reported (Ai Boom Big Tech Capital Expenditures Now Seen Topping 1 Trillion In 2027). Higher spending is reducing operating cash flow while debt and lease liabilities rise, RCRTech reported (Moodys Hyperscaler Capex).

What would end the high is consumers stopping: if demand falls far enough to free capacity, the consumer market reprices and the squeeze unwinds (Trendforce). It would end faster if the buying is cut rather than funded, since the increases are being carried on falling free cash flow and rising debt (Moodys Hyperscaler Capex). The level is the firm part; the span past 2027 is not.

Nothing here is financial, legal or investment advice. Every figure is published by its named source and carries the date it was published; figures may since have changed.