AMD records third-quarter results on Tuesday, driven by surging data center accelerator sales. AMD also confirmed it has entered into multi-year long-term supply agreements (LTAs) with memory suppliers to cushion the financial impact of rapidly rising High Bandwidth Memory (HBM) prices.

Although AMD refrained from explicitly naming its primary counterparties, industry sources point to Samsung Electronics (via BiggoFinance) as the largest beneficiary of the arrangement.

Following high-level discussions between Su and Samsung boss Jay Y. Lee earlier this year, the Korean tech giant has secured a dominant position supplying AMD with both sixth-generation HBM4 and standard commodity DRAM.

“We have long-term contracts with some memory suppliers,” CEO Lisa Su said, noting that AMD is also working closely with major cloud customers to share incremental cost increases.

Samsung recently revealed it has locked in multi-year HBM supply commitments extending into 2027 with ten major global customers, while negotiating similar terms with ten additional buyers.

Read more: Samsung expands long-term memory agreements to 20 clients, seeks to lock in up to 70% of output through 2030

For Samsung, securing long-term supply agreements with key AI chipmakers guarantees predictable, high-margin output through 2027.

As memory shortages persist into 2027, the cost structure divergence between chip designers backed by multi-year LTAs and those relying on spot-market pricing will likely dictate hardware margins across the semiconductor industry.

You may like:

• Apple CEO warns “unprecedented” memory hike surge won’t ease in 2027

Samsung has reportedly decided to completely stop its 8-inch wafer chip manufacturing business by the end of 2030. The decision has raised concerns among several semiconductor companies that depend on Samsung to manufacture their chips.

According to the report, Samsung has informed customers of the planned shutdown. The company will gradually reduce production at its 8-inch facilities and focus more on advanced chip manufacturing using 12-inch wafers.

Samsung’s 8-inch production lines manufacture chips using older technologies ranging from 65nm to 180nm. These chips are widely used in power management ICs (PMICs), display driver ICs (DDIs), microcontrollers (MCUs) and image sensors.

Source – Samsung

The decision could affect several Korean fabless companies, including LX Semicon, Dongwoon Anatech, Silicon Mitus, Above Semiconductor and Images. These companies will now need to find other manufacturers to continue producing their chips.

However, finding alternative suppliers will be difficult. South Korean foundries DB HiTek and SK Hynix already have limited production capacity. Foundries in Taiwan and other countries are also facing capacity challenges.

Moving production to 12-inch wafers is another option, but it could increase manufacturing costs. Companies may also need to redesign their chips and test them again before starting production.

Some fabless companies have started building additional chip stocks to manage the situation. However, storing more chips will increase their costs.

Samsung Chairman Jay Y. Lee conducted an on-site inspection of the company’s Advanced Package Lab (APL) in Yokohama, Japan, targeting advancements in next-generation chip packaging crucial for AI hardware.

The visit marks Lee’s first direct review of the facility since its official opening in September. Samsung is committing 40 billion yen through 2028 to build out the Yokohama hub. The site focuses on evaluating key materials and equipment, alongside developing new packaging architectures such as wafer bonding and glass substrates.

Advanced packaging has become a central bottleneck in AI hardware execution. Integrating GPUs, HBM, and specialized accelerators onto a single package dictates processing bandwidth, thermal limits, and overall power efficiency.

By embedding R&D in Yokohama, Samsung aims to bridge its in-house design and foundry operations with Japan’s dense ecosystem of specialized material suppliers and equipment makers.

The strategic push signals Samsung’s broader intent to secure supply chain integrations and technical parity in high-performance memory and contract chip manufacturing as global competition intensifies.

Chinese memory manufacturer CXMT has claimed a major architectural milestone, announcing engineering-sample validation for its sixth-generation DRAM process built on a 4F² structure. The company frames the achievement as a world-first that could accelerate its commercial timeline against market incumbents.

According to Chosun, CXMT plans to ship DDR5 registered DIMM samples to server customers by the end of the year, targeting high-density data center applications before scaling to broader product lines.

The announcement comes alongside stellar financial momentum, with CXMT reporting third-quarter revenue of 118.7 billion yuan ($17.84 billion), a 613% year-over-year surge that pushed the firm back into profitability.

While Micron abandoned 4F² over these physical constraints, Korean leaders Samsung Electronics and SK Hynix are deliberately prolonging their mature 6F² platforms to preserve margin stability and yield predictability.

• The great 4F² DRAM split: Samsung and SK Hynix’s panic, CXMT’s aggressive bet, and Micron’s clever rejection

Crucially, CXMT has omitted silicon yield data for the 4F² prototype. Unverified sample validation rarely equates to volume manufacturing parity. Current market estimates place CXMT’s HBM3E production yields around 20% to 30%, far trailing SK Hynix’s established 70% threshold.

Holding a 10% global DRAM revenue share as of Q2 2026, CXMT has proven its ability to disrupt legacy nodes, but breaking the duopoly of Samsung and SK Hynix on advanced architectures will require proven wafer yields rather than self-reported milestones.

Memory costs continue to spike, and the budget and mid-range mobile segment is taking the biggest hit. As smartphone sales declined, mobile chip titans slipped as well, but Apple and Samsung outperformed them in Q3 2026.

Counterpoint Research reported in the latest report that global mobile chip shipments dropped 20% year-on-year in Q3 2026. Rising memory costs are the primary reason, which took the sales of non-premium devices down.

As smartphone prices hike, consumers intend to keep their phones longer. The era of frequent smartphone upgrades is over. Mobile owners now prefer keeping their existing device until it lasts or a tech breakthrough lands.

While chip titans like Qualcomm and MediaTek declined, Apple and Samsung improved their presence in the Mobile chip market in Q3 2026.

MediaTek remained the market leader with 28 percent market share, with Qualcomm running up with 22 percent. UNISOC also managed to secure 13 percent market share globally.

Apple’s A-series chips seized 21 percent of the worldwide mobile SoC market. This was driven by a decline in the sales of affordable Android phones, handing the iPhone maker the opportunity to capitalize.

Samsung also expanded the adoption of its in-house Exynos chip across segments. Galaxy S26 series did the heavy lifting in surging the shipments and market share of the Exynos 2600 chipset.

In Q3 2026, Samsung’s Exynos stood as the fifth-largest mobile SoC vendor. The company secured 8 percent market share in the third quarter, which went up from 6 percent in Q3 2025.

Looking ahead, Samsung is expected to use Exynos 2700 in its next flagships. The Galaxy S27 Pro could also be bringing Exynos inside in most regions, potentially uplifting Samsung’s mobile application process share further.

Global smartphone SoC shipments are expected to decline by 20% YoY in 2026 and another 3%-4% YoY in 2027. Memory prices are expected to maintain an upward trajectory, but price volatility is likely to moderate next year compared with 2026.

Counterpoint noted that flagship SoCs are expected to remain broadly stable in 2027 despite higher prices due to migration to 2nm process nodes, while rising memory and component costs will continue to add to cost pressures.