KR — ZDNet Korea
AI Investment Rises and Memory Prices Jump, Infrastructure Spending Set to Surge 53% Next Year
AI1 min read
Research firm Gartner says global spending on AI infrastructure, mainly servers, will jump 53% next year to about $2.4 trillion. Rising memory prices are pushing the cost even higher.
What happened
Gartner expects companies to spend much more on AI infrastructure in 2027. That means the computers, chips and servers needed to run and train AI. Infrastructure is expected to make up about 59% of all AI spending. Total AI spending is forecast to rise 47.5% to about $4 trillion.
Key numbers
- AI infrastructure spending in 2027: $2.4 trillion, up 53% from the year before
- AI-optimized servers: $527.6 billion this year to $957.2 billion next year, up 81.4%
- AI services spending: up 29.5%
- AI software spending: up 42%
- Generative AI model spending: $51 billion, up 82%
Why spending is growing
Companies are using AI more, and each AI answer needs computing power. Training new AI models also needs more and more computing power. Big cloud companies and new 'neocloud' providers are building more data centers. Gartner calls the rise in memory prices 'Memflation,' and says it is making servers more expensive.
Why it matters
AI is becoming a very large business, and the money is going first into the hardware behind it. Companies that make servers and memory chips may benefit. Businesses that buy AI tools may face higher costs. Gartner also says general-purpose foundation models will still hold most of the market, even as specialized AI grows fast.
What to watch
- Whether memory prices keep rising and push server costs higher
- How fast cloud companies and neoclouds add new data centers
- Whether specialized AI models grow as fast as Gartner expects
Glossary
- Gartner
- A research firm that studies technology markets and forecasts how much money will be spent on them.
- Memflation
- A word for rising memory prices. Here it means memory costs are pushing up server prices.