Samsung Says AI Chip Shortage Will Last Until 2028

Samsung Sees No End to the AI Memory Squeeze Through 2028

Samsung's latest earnings have delivered a blunt message to the tech industry: the memory chip shortage fueling the artificial intelligence boom is not a temporary blip. The company says demand from cloud providers building out AI infrastructure is so intense that supply constraints could persist until 2028, with major hyperscalers now locking in multi-year contracts just to secure the chips they need 1. That timeline reframes what many investors had hoped was a cyclical crunch as something closer to a structural reordering of the memory market, driven directly by the scale of cloud infrastructure investment.

A Boom With Winners and Skeptics

The scramble for AI capacity is visible across the hardware stack, not just in memory. AMD used a recent event to lay out its own infrastructure ambitions, unveiling the Helios rack — its first full AI infrastructure system — built around sixth-generation Epyc CPUs, alongside a broader vision spanning cloud, client devices, and robotics 2. The message from chipmakers is consistent: demand for AI compute and memory shows no sign of cooling, and companies are racing to offer complete rack-level solutions rather than individual components.

Yet Wall Street's patience with the spending required to keep pace is thinning. Meta's latest results exposed a growing split among Big Tech firms: those with established cloud businesses are already turning AI infrastructure into revenue, while companies still in the build-out phase are being punished for costs without corresponding payoff 3. Meta's stock slide illustrates that investors are no longer content to fund AI ambitions on faith alone — they want to see monetization.

Google has offered the counter-example. Its cloud division posted record profits, which the company points to as proof that heavy AI infrastructure spending is translating into real business results, giving it cover for continued massive capital outlays 4. That contrast — Google's justified spending versus Meta's questioned spending — captures the market's current mood: infrastructure investment is tolerated only when it visibly converts into cloud revenue.

The $400 Billion Question

That tension is set to intensify as Amazon and Microsoft, the two dominant cloud providers, prepare to report on a combined $400 billion in AI-related spending. Coming just after Alphabet was penalized by investors for its own aggressive AI outlays, the reports will test how much more scrutiny the market is willing to apply to the biggest spenders in cloud infrastructure 5.

Taken together, the reporting suggests an industry at an inflection point: chip supply is tightening well into the future, infrastructure investment is accelerating across CPUs, memory, and rack systems, and investors are increasingly sorting winners from losers based on whether that spending shows up in cloud earnings.