Samsung Electronics Co. (SSNLF) is doing what any sensible company does when it has something everyone desperately wants: charging more for it. The Korean tech giant is negotiating sharply higher prices for its next-generation High Bandwidth Memory 4 (HBM4) AI memory chips, and the numbers tell a story about who holds the cards in the AI hardware race right now.

According to local media reports, Samsung is seeking around $700 per unit for HBM4, which represents a roughly 20% to 30% jump over the prior generation. That's not just inflation—it's a signal that Samsung believes it has regained serious leverage in a market where AI memory remains scarce and precious.

Back in the Game After Playing Catch-Up

The aggressive pricing push comes as Samsung announced it has started mass production of HBM4 and begun shipping commercial products to customers. This marks an important milestone for a company that initially found itself trailing SK Hynix when the AI memory boom kicked into high gear. Samsung is now working to claw back market position, and pricing its premium chips at a premium is part of that strategy.

Charu Chanana from Saxo Markets told Bloomberg the pricing discussions highlight Samsung's renewed "pricing power," suggesting the company has successfully positioned itself at the high end of the AI memory market where margins are fattest and competition is fiercest.

Samsung's higher HBM pricing strategy isn't happening in a vacuum. Broader memory prices are rising across the board, with DRAM—a key input for HBM production—also climbing sharply. That strengthens Samsung's hand not just in specialized AI memory but also in general-purpose DRAM used in smartphones and PCs. The company can now juggle production across both categories to maximize returns, shifting capacity wherever the economics look best.

What Goes Up for Chips Goes Up for Phones

Here's the twist: the same memory inflation that's padding Samsung's semiconductor profits is also inflating the company's own smartphone manufacturing costs. Reports indicate Samsung is considering a price increase of 100,000 to 200,000 won for Galaxy S26 models compared to the previous generation. The Galaxy S26 Ultra with 512GB storage could potentially exceed 2 million won, a threshold that would have seemed eye-watering not long ago.

Samsung reportedly plans to soften the blow by using its in-house Exynos 2600 chip in domestic Galaxy S26 models and by emphasizing new AI features that might justify the higher sticker price. Whether consumers will bite is another question entirely, but at least the company is trying to give them something new to pay for.

Nvidia's Shadow Looms Large

The HBM4 pricing news lands just as Nvidia Corp. (NVDA), one of Samsung's key customers, gained additional demand visibility. Meta Platforms Inc. (META) recently agreed to deploy "millions" of Nvidia AI processors over the next few years, which translates directly into voracious appetite for high-bandwidth memory.

Samsung's HBM4 is also expected to feature in Nvidia's next-generation "Vera Rubin" AI chip platform, according to Bloomberg reporting based on the Chosun Ilbo newspaper. That's a meaningful design win for Samsung, validating both its technology and its production capabilities.

The Supply Chain Shakeout

According to SemiAnalysis, Nvidia plans to source HBM4 almost entirely from SK Hynix and Samsung, effectively cutting Micron Technology Inc. (MU) out of the mix. Analysts expect SK Hynix to provide about 70% of Nvidia's HBM4 supply with Samsung handling roughly 30%, according to the Chosun Daily.

Industry observers say Micron fell behind because it struggled to meet Nvidia's stricter technical specifications. TrendForce and other sources noted that Nvidia raised HBM4 data-transfer speed requirements above 11 gigabits per second last year, and Micron apparently couldn't keep pace with those tougher targets.

The competitive dynamics are fascinating. SK Hynix reportedly set HBM4 pricing for Nvidia in the mid-$500 range last August, but may now move to match Samsung's higher pricing level. When you're one of only two suppliers for something Nvidia desperately needs, you don't leave money on the table.

The Margin Math

Bloomberg Intelligence analyst Masahiro Wakasugi wrote that a $700 price point could deliver operating profit margins of 50% to 60% for Samsung's HBM4. Those are the kind of margins that make semiconductor executives very happy. Wakasugi also suggested the pricing gap between Samsung and SK Hynix could narrow in 2026 if Samsung succeeds in supplying more HBM to Nvidia at these premium price points.

Reports indicate Samsung plans to begin HBM4 manufacturing by February and prepare initial shipments after passing qualification tests for both Nvidia and Advanced Micro Devices, Inc. (AMD). Qualification is the last major hurdle before volume shipments begin, and clearing it with multiple customers gives Samsung flexibility and negotiating leverage.

The big picture here is that the AI gold rush has flipped the traditional power dynamics in the chip supply chain. Memory makers who can deliver cutting-edge HBM at the speeds and capacities AI chips demand suddenly have pricing power they haven't enjoyed in years. Samsung is clearly planning to use it.