TrendForce projects memory to reach 68% of AI hyperscaler capex by 2027
TrendForce projected that memory could take 68% of AI hyperscaler capex by 2027, up from 47% this year, putting Micron in focus as DRAM, NAND and HBM prices surge. NVIDIA separately confirmed that tighter memory supply is being driven by the AI buildout itself and that memory commitments grew from $119B to $279B.
Detected & updated continuously · Source: Nebula
Story subjects
Sources
@Benzinga
TrendForce says memory could take 68% of AI hyperscaler capex by 2027, up from 47% this year. That puts Micron ($MU) in focus as DRAM, NAND and HBM prices surge, even with the stock still 26% below its June high.
@gaasych
$NVDA $MU $SKHY We want to be direct about this rather than let it linger as an open question. Memory scarcity today is being driven in large part by the AI build out itself and unlike a component that simply raises our cost with no offset benefit. Tighter memory supply is a
@Basat098
Nvidia Could Beat Earnings Today and Still Be a Sell — I’m bearish on Nvidia’s earnings reaction today, because the cost of sustaining that demand is quickly rising. Wall Street already expects about $92B in Q2 revenue, roughly $104B in Q3 guidance, and gross margins near 75%, so simply beating estimates may not be enough. The supply chain is constrained by SK Hynix. Although its memory demand remains extremely strong and customers are locking in long-term supply, Nvidia customers have reportedly been warned that AI server prices could rise more than 15% in early 2027 because of memory costs. Strong demand is bullish but the shortages make the finished systems materially more expensive. As we all know, power is a very quickly growing bottleneck. PJM, the largest U.S. power grid operator, came 6.8 GW short of its required capacity target for 2028/29 even though its latest auction cleared at the maximum price, while only 317 MW of new generation cleared. At the same time, the large tech companies are expected to spend more than $730B on data centers this year, and Nvidia (alongside PE firms) is helping support financing platforms targeting $500B+ in AI infrastructure while also providing major guarantees tied to OpenAI data-center projects. The bearish thesis is therefore simple: Nvidia can still report a huge quarter tonight and disappoint if management talks about rising memory costs, margin pressure, slower deployments, power constraints, or Q3 guidance that is only near rather than well above $104B.