A global memory shortage driven by AI demand has been pushing DRAM chip prices up since early 2026. In practice, this means your next PCs, servers and cloud subscriptions could cost more. Here is what this hardware crunch means for an SMB's IT budget, in plain language.

In Brief

• Manufacturers Samsung, SK Hynix and Micron have redirected much of their production toward high-bandwidth memory (HBM) used by AI data centers, at the expense of consumer-grade memory (source: specialized tech press, June 2026).
• GoPro filed a regulatory warning with the SEC (filing dated June 1, 2026) citing substantial doubt about its ability to continue as a going concern, after memory prices jumped 80 to 115% in a single week in late March 2026 (source: GoPro SEC filing, reported by Bloomberg).
• Research firm IDC expects a 3 to 8% price increase on smartphones and 4 to 8% on PCs in 2026 under its moderate scenario, with the PC market potentially shrinking by up to 8.9% in the pessimistic scenario.
• CNBC notes this memory shock hits companies unevenly: Apple and Microsoft can absorb it through long-term supply contracts, while it has become an existential threat for smaller manufacturers.
• For an SMB, the right response is not panic but anticipation: audit your hardware fleet, plan purchases ahead, and favor arrangements that shift price risk onto the vendor.

Why memory prices are surging

DRAM memory is a component found in every computing device: PCs, servers, smartphones. Since 2025, cloud giants (Microsoft, Google, Meta, Amazon) have been building AI data centers that consume massive amounts of a specialized variant, high-bandwidth memory (HBM), essential to the server chips used to train AI models.

HBM is significantly more profitable to manufacture: its margin exceeds 70%, compared with 20 to 30% for standard DRAM used in PCs and smartphones (source: analyses cited by specialized tech press, June 2026). The logical result: Samsung, SK Hynix and Micron, which control most of the world's supply, have redirected their factories toward the most lucrative production. Consumer memory becomes scarce, and therefore more expensive.

Share of global memory supply absorbed by servers and data centers, before and after the rise of generative AI.

Who bears the cost: the GoPro example

GoPro's case illustrates the scale of the shock. In a regulatory filing submitted to the SEC (US Securities and Exchange Commission) on June 1, 2026, the company cited substantial doubt about its ability to continue as a going concern, partly due to soaring memory costs needed to manufacture its cameras (source: GoPro SEC filing, reported by Bloomberg).

This is not an isolated case. According to CNBC (June 27, 2026), large tech groups like Apple or Microsoft absorb the shock better thanks to multi-year supply contracts and substantial cash reserves. Smaller players, by contrast, buy memory at market prices, often with quotes valid for only a few days.

For an SMB buying IT hardware (PCs, servers, network equipment), the effect is indirect but real: your hardware vendors will sooner or later pass this increase on to their prices.

What this means for hardware purchases at your SMB

How long will this last

A quick return to normal should not be expected. New memory plants take years to build; supply adjustment is structurally slow relative to AI demand that keeps growing.

A shortage that also confirms a positive trend

It would be unfair to focus only on the cost. This shortage is the direct consequence of real, massive AI adoption by the world's largest companies, which are investing enormous sums in the compute needed for increasingly useful models. It is a sign that generative AI has moved past the experimentation stage to become a full-fledged economic infrastructure.

For an SMB, the best response remains the same as for any cost fluctuation: measure, plan ahead, and don't just absorb it. As shown in our article on cheaper AI reasoning thanks to Microsoft's in-house models, AI-related costs move in both directions depending on the technical layer involved: software gets cheaper, hardware gets pricier. Active budget management lets you capture one trend while absorbing the other.

FAQ

Why is computer memory getting more expensive because of AI?

Memory manufacturers (Samsung, SK Hynix, Micron) have redirected production toward high-bandwidth memory (HBM) for AI data centers, which is more profitable than standard memory. This reduces the supply available for standard PCs and servers, pushing prices up.

Should my SMB buy hardware urgently right now?

Not in a panic, but with a plan. If a PC or server renewal was already scheduled for 2026, it is reasonable to move it forward rather than wait for a price drop, which is not expected before late 2027 according to SK Hynix.

Does the cloud let me avoid this price increase?

The cloud doesn't eliminate the cost, but it shifts it to the vendor, who negotiates large volumes and can better absorb volatility. It's worth considering for computing needs rather than investing in your own servers right now.

Will this shortage also affect the AI software I already use?

Indirectly, yes: cloud and SaaS providers sometimes pass on their infrastructure costs. But this effect is gradual and not guaranteed across every tool: watch your invoices rather than anticipating a blanket increase.

Conclusion

The global AI-driven memory shortage is not an abstraction reserved for tech giants: for an SMB, it will translate into pricier PCs, servers and subscriptions throughout 2026 and likely into 2027. The right response is neither urgency nor denial, but simple budget management: audit your fleet, prioritize purchases that can't wait, and shift to the cloud what can be shifted. To build that approach, explore our other LUWAI Mag resources or see how other leaders have managed their AI investments in our success stories.