Samsung’s chip outlook separates HBM4 sales from NPU and Taylor plans

Samsung says it significantly scaled up HBM4 sales and responded to growing demand for server SSDs during the third quarter ended September 30, 2026. Its October 29 results release places those memory developments alongside plans for a Q4 ramp in 4nm NPU mass production and the expected start of operations at its Taylor, Texas, fab in 2027.

For a reader tracking the supply behind AI systems, the dates and verbs matter. A report that a product’s sales grew describes business already recorded. A production ramp and the start of a factory’s operations are manufacturing milestones. Combining all three into a claim that new chip capacity is available today would lose the distinctions in Samsung’s announcement.

The timeline below keeps Samsung’s reported quarter separate from its outlook. Each row describes the company’s own statement, not an independently verified customer allocation or production audit. [Source: Samsung](https://news.samsung.com/global/samsung‑electronics‑announces‑third‑quarter‑2026‑results) | Workstream | Period and stated status | What remains separate | | --- | --- | --- | | HBM4 and server SSDs | Q3 ended September 30: Samsung reports HBM4 sales scale‑up and growing server SSD demand | The release does not specify a particular customer’s allocation or product availability | | 4nm NPUs | Q4 2026: the foundry business plans a mass‑production ramp‑up | A planned ramp is not confirmation that every resulting product has shipped | | Taylor fab | 2027: Samsung expects operations to commence | The expected opening is not capacity already operating in Q3 | The memory row concerns sales and demand, the NPU row concerns a planned manufacturing ramp, and the Taylor row concerns a facility. A buyer comparing supplier statements needs to identify which of those milestones a claim actually describes before treating them as interchangeable evidence of supply.

Samsung says it expects AI competition to support continued infrastructure investment in Q4 and forecasts further demand growth in 2027. It plans to increase sales of HBM, server DRAM and SSDs, as well as supply of high‑speed server DDR5 products and QLC SSDs. Those are the [memory business’s outlook and strategy](https://news.samsung.com/global/samsung‑electronics‑announces‑third‑quarter‑2026‑results), rather than disclosures of guaranteed deliveries to every buyer. Its foundry outlook separately points to advanced‑node activity and the Taylor facility. That is another reason to keep the product, manufacturing business and time window attached to a statement. An expectation about industry investment does not identify which customer will receive which component or when a finished system will be available.

A practical follow‑up to a supplier’s AI‑capacity announcement is to ask what has been completed: a sale, a production ramp or a facility milestone. Then ask which product and delivery commitment apply to the intended system. Samsung’s release supplies a useful high‑level timeline, but it does not replace a specific procurement commitment. The current development is the reported expansion of HBM4 sales alongside dated NPU and foundry plans. No customer allocation, chip shipment, factory operation or performance claim was independently tested for this report. The timeline should be read as a separation of reported business and expected milestones, rather than a guarantee of supply or a comparison of chip performance. Archival photograph of a Samsung SO‑DIMM memory module, taken November 12, 2015. It provides memory‑supplier context and does not show HBM4, a 4nm NPU or the Taylor fab. Photo: [© Raimond Spekking / CC BY‑SA 4.0 (via Wikimedia Commons)](https://commons.wikimedia.org/wiki/File:Samsung_1GB_1Rx8_PC3‑8500S_SoDimm_Memory_Module‑2720.jpg), [license](https://creativecommons.org/licenses/by‑sa/4.0/). Downsampled for display; responsive crops and other image adaptations retain the same license.