South Korean brokerage Meritz: Middle Eastern sovereign AI funds are entering to purchase South Korean memory chips; Kimi K3 is a positive, not a negative impact.

```

Middle Eastern sovereign AI funds have officially entered the memory procurement market, reshaping the supply-demand landscape and driving prices to accelerate upwards. Meanwhile, the recently released Kimi K3 model, a hot topic in the AI community, has been interpreted by Korean securities firms as a mid-term positive for AI hardware, rather than the market’s feared demand shock.

According to a research report released by Meritz Securities in Korea on July 19, Middle Eastern sovereign AI investors such as Saudi Arabia have recently begun negotiations with Korean memory manufacturers including Samsung Electronics and SK Hynix regarding mid- and long-term procurement plans, inquiring about supply capability and volume for the next several years. The report expects that the quarter-on-quarter price increase of server DRAM contracts in Q3 2026 will exceed the previous market forecast of approximately 15%. Suppliers whose pricing was relatively relaxed in Q2 will see especially prominent price hikes in the latter half of the year.

Meanwhile, the spot market has already shown clear signals of price increases. The spot price of 64GB DDR5 server DRAM has surged rapidly since mid-July, with recent quotes reaching the range of $3,100 to $3,400, about 146% higher than the contract price of approximately $1,380 at the end of June. High-end products with a bus rate of 6400Mbps are seeing particularly significant increases.

As the supply-demand structure of the memory market undergoes dramatic changes, the investment strategy team at Meritz Securities released another report on the same day, characterizing the launch of the Kimi K3 model from Moonshot AI as a mid-term positive for AI hardware demand. The team recommends buying semiconductor, memory, and other AI component companies, and reducing holdings in large tech stocks like Google and Microsoft. Meritz believes that the market’s view that "Kimi K3 will reduce AI hardware demand" is based on two major misjudgments, and that K3 will instead bring demand growth for hardware manufacturers from a mid-term perspective.

Middle Eastern sovereign AI funds enter the market; a single buyer can sway supply-demand balance

Analysts at Meritz Securities point out in their report that the core logic of Middle Eastern sovereign AI investment is national security and data sovereignty—each country hopes to build data centers within its own territory and operate AI systems autonomously in its own language and data. As this investment plan becomes more concrete, the supply shortage expected for 2027 is already emerging ahead of schedule.

The current structural changes in the memory market make the entry of a single large buyer highly impactful on prices. The construction of AI data centers is highly concentrated; just one new procurement player can instantly shift the market demand satisfaction rate by 3 to 5 percentage points. The demand satisfaction rate refers to the ratio of actual supply to market needs. Even minor fluctuations in this indicator can trigger significant price jumps. Currently, large customers account for more than 70% of memory manufacturer revenue, and pricing power is heavily concentrated among a handful of buyers.

This is completely different from the decentralized procurement structure of the era of smartphones and PCs. Back then, many manufacturers purchased in small batches, and changes in orders from a single client had limited impact on the overall market. Today, the centralized procurement model for AI data centers means every new sovereign-level entrant could become a decisive variable in pricing.

Supply shortages have spread to end products; quantity, not price, is the core issue

Meritz points out that the impact of the memory shortage has spread from the data center side to consumer electronics terminals. Apple and Chinese smartphone manufacturers are both facing gaps in their Q4 production plans due to an inability to fully procure memory supplies before the high-demand season in the second half of the year.

He emphasized that the essential problem currently facing the market is "allocation of quantity" rather than merely "price"—"In short, there simply isn’t enough stock, and all parties are scrambling to buy." Regarding market rumors of cyclical concerns, he believes this view is somewhat off; the current tight situation is rooted in structural supply shortages, not in short-term demand fluctuations.

The report also states that some suppliers adopted relatively flexible low pricing strategies in Q2 to accommodate customer needs, meaning these manufacturers will have greater room to raise contract prices in Q3 and Q4, with increased pricing elasticity.

Kimi K3's release seen as a positive for AI hardware demand, not a shock

On the same day, the Meritz Securities investment strategy team released a report characterizing the launch of Moonshot AI's Kimi K3 model as a mid-term positive for AI hardware demand, recommending investors buy semiconductor, memory, and other AI component companies, while reducing positions in large tech stocks such as Google and Microsoft.

Meritz clearly distinguishes this event from last year’s DeepSeek impact. DeepSeek's core narrative was completing AI training at an extremely low cost of $6 million, whereas Kimi K3 did not disclose training costs or GPU usage, and officially stated that running K3 requires a large cluster composed of at least 64 high-performance chips—a model that cannot be supported by low-cost hardware. In terms of usage cost, according to AI performance evaluation firm Artificial Analysis, K3’s cost per task is $0.95, similar to GPT-5.6 Sol ($1.04) and Claude Fable 5 ($2.75), and an order of magnitude higher than DeepSeek V4 Pro ($0.04).

The analysis further points out that the logic behind "K3 will reduce AI hardware demand" is based on two misjudgments: First, AI infrastructure investment is not solely led by large tech companies—taking Nvidia’s data center chip sales as an example, the share of large cloud providers is declining while the share of mid-sized AI cloud services and ordinary enterprises is rising; second, Chinese AI models' inference services actually run heavily on US hardware, with providers like Fireworks AI and Together AI using US chips as their foundation to deliver enterprise services.

He concluded that, "Regardless of whether it’s closed or open source, all AI business participants ultimately need to use hardware companies, which will gain mid-term demand growth from the launch of K3."

Risk warning and disclaimerThe market involves risks; invest with caution. This article does not constitute personal investment advice and has not taken into account the specific investment objectives, financial situation, or needs of individual users. Users should determine whether any opinions, views, or conclusions herein suit their particular circumstances. Investment is at your own risk. ```