Fewer Chips, Bigger Bucks

The locus of value creation in the semiconductor industry is shifting.

Unit growth has largely stalled, yet revenue continues to accelerate. Pricing power and technology intensity are pushing total semiconductor sales to record highs.

For decades, more chips meant more dollars. But that no longer holds true.

The 2025 data from the Information Network and World Semiconductor Trade Statistics (WSTS) confirm that industry growth is being driven not by how many chips are made, but by how advanced—and how valuable—each chip has become.

Memory units stagnate

As shown in Chart 1 below, memory unit shipments have remained essentially flat since 2019, fluctuating in a narrow range. Despite the cyclical volatility of the memory business, the long-term trend shows no structural increase in output.

This stability has two sources.

• Memory manufacturers are intentionally controlling wafer starts to protect pricing and profitability, rather than pursuing unit growth
• The advent of High Bandwidth Memory (HBM)

As HBM consumes more wafer capacity and requires longer processing cycles, suppliers have shifted resources away from conventional DRAM (Dynamic Random Access Memory) and NAND (Non-Volatile Memory) output. And for HBM, packaging and hybrid bonding constraints have become the dominant bottlenecks. That’s keeping unit production constrained, even as global demand for AI and high-performance computing climbs. The outcome: value creation depends not on producing more chips, but on producing higher-value memory stacks.

Chart 1: Memory Units (Three-month moving average)

As shown below in Chart 2, memory revenues have risen sharply during the same period, climbing from roughly $10 billion in mid-2019 to more than $17 billion per month by mid-2025. The trendline shows consistent acceleration, with revenues diverging from stagnant unit shipments. The core driver of memory industry growth is now pricing and product mix, not volume.

Higher ASPs (average selling prices) are being sustained by the structural shift to HBM and DDR5 (Double Data Rate, a souped-up version of DRAM). Each HBM stack commands several times the price of a conventional DRAM module, and the limited number of qualified suppliers—primarily SK hynix, Samsung, and Micron—has reinforced pricing stability. Even as total unit shipments remain constant, the dollar value of each chip sold has increased substantially due to higher bandwidth, density, and energy efficiency.

Chart 2: Memory Revenues

It’s more than memories

As shown in Charts 3 & 4, the same pattern holds for all other integrated circuit (IC) categories—including logic, analog, power, and microcontrollers. Shipments (Chart 3) remain flat to slightly up since 2019, maintaining a range near 23 billion units. Yet, IC revenues (Chart 4) excluding memory have doubled from approximately $20 billion to $43 billion since 2019.

This pattern confirms that overall semiconductor demand has shifted from mass-market consumer products toward specialized, high-value components. Automotive electronics, industrial automation, and AI infrastructure have replaced smartphones and PCs as the primary growth drivers. These sectors consume fewer units overall but require far higher-value chips.

Chart 3: IC Units (Ex-Memory)

The Chart 4 revenue line climbs consistently through multiple industry cycles, suggesting that high-ASP products and advanced-node migration are the dominant forces in the broader IC sector. AI accelerators, data-center processors, automotive power devices, and advanced analog circuits all carry dramatically higher ASPs than traditional components. This shift toward richer product content and advanced manufacturing nodes has permanently altered the economics of the sector.

Chart 4: IC Revenues (Ex-Memory)

Winners and Losers in the High-ASP Era

The industry is being reshaped around technology intensity rather than production scale. Companies with the ability to produce at advanced nodes, supply high-performance memory, or provide essential process-control tools are winning. Those tied to mature nodes, consumer end markets, or wafer-volume dependence are falling behind. This divide is structural, not cyclical, and will continue as the industry’s center of gravity shifts toward AI, high-performance computing, and advanced automotive semiconductors.

Table 1 summarizes the emerging winners and losers in the high-ASP era. The ability to command pricing power, leverage advanced manufacturing, and enable process complexity now define success in the semiconductor value chain.

Table 1: Winners and Losers in the High-ASP Era

The winners are positioned where complexity and performance justify premium pricing. TSMC leads among foundries through its dominance in sub-5 nm processes, while Samsung and SK hynix (and Micron coming on) control the high-bandwidth memory segment that powers AI servers. KLA and ASML continue to benefit from the growing need for inspection, metrology, and extreme ultraviolet (EUV) lithography tools that make these technologies possible. Companies whose revenue models depend on wafer volumes or commodity devices—such as mid-tier foundries, consumer System on Chip (SoC) vendors, and deposition or etch tool suppliers—are seeing slower growth as the economic leverage of volume weakens.

Analysis and Outlook

The four charts together present a unified picture of an industry defined by maturity in volume and acceleration in value. The global chip market has transitioned from quantity-based growth to value-based growth, where each wafer and package carries more embedded technology, more functionality, and far greater pricing leverage.  At least through the remainder of 2025 and into 2026, I expect this decoupling of units from revenues to persist, driving a new era in which the sophistication and profitability, not the sheer number of chips produced, defines the health of the industry.