Semiconductor revenue reaches record $425B

A record quarter, powered by memory

Global semiconductor revenue surpassed $425 billion in the second quarter of 2026, setting another all-time high for the industry, Omdia reported in a release dated September 14, 2026 [1]. The headline number is striking on its own, but the pace is even more important: Omdia said the market expanded 31.4% quarter over quarter, beating the previous sequential growth record of 29.2% recorded in the first quarter of 2026 [1]. The same release, syndicated through MarketMinute, puts first-half 2026 semiconductor revenue at $752 billion, meaning the industry has already generated more in six months than in almost every full year before 2025 [2].

The central force behind that leap is memory. Omdia said AI demand and strong memory prices remained the leading market dynamics in 2Q26, with memory integrated circuits accounting for more than half of all semiconductor revenue during the quarter [1]. That is the key change in the cycle: the boom is not only about the processors that run AI models, but also about the high-bandwidth memory, DRAM, NAND and related components that make those processors useful at scale [1].

In older semiconductor upturns, memory was often the volatile sidecar: it amplified profits in good years, then punished suppliers when supply caught up. In this cycle, memory has moved to the center of the machine. Omdia said DRAM, NAND and NOR each recorded their highest sequential growth for a second quarter since the firm began tracking the market, and all three reached their highest recorded quarterly revenue in 2Q26 [1].

AI has changed the shape of the cycle

The semiconductor market has always been cyclical, but Omdia’s numbers suggest the current expansion is behaving differently. Since Omdia began tracking the market in the first quarter of 2002, only 10 of 97 quarters have produced sequential revenue growth above 10% [1]. The last four quarters, beginning in 3Q25, have all posted double-digit sequential growth, and Omdia expects 3Q26 to follow that pattern [1].

That matters because it points to a broader industrial reset rather than a single-quarter price spike. The AI build-out is pulling on multiple layers of the supply chain at once: high-bandwidth memory for accelerators, conventional DRAM for servers, NAND for storage-rich workloads, microprocessors for infrastructure control and compute, and advanced packaging capacity to bind those parts together [1]. In practical terms, the value of a chip is increasingly tied to whether it can unlock scarce bandwidth.

The processor market is benefiting too. Omdia said microprocessor revenue rose 16% quarter over quarter in 2Q26, far above the firm’s historical second-quarter seasonal benchmark of about 1% [1]. That suggests AI infrastructure is lifting more than just memory suppliers: CPUs, custom accelerators, networking silicon and platform controllers are all being pulled into the same capital-spending wave [1].

Still, the mix is uneven. Omdia noted that memory is now so influential that it can distort the top-line view of the industry; excluding memory, the wider semiconductor market still grew more than 10% quarter over quarter, which is well above the historical second-quarter pattern of just over 3% [1]. That detail is important because it shows the recovery is broader than a pure memory squeeze, even if memory is setting the tempo [1].

The loot-box economics of silicon

The phrase “silicon has entered its loot-box era” captures the strange economics of this moment. Buyers are not simply purchasing chips; they are buying access to constrained performance pools. A hyperscaler ordering AI servers is effectively buying a bundle of compute, bandwidth, power delivery, networking and future model capacity. The most valuable part of that bundle may change from one quarter to the next, depending on where shortages appear.

In 2Q26, Omdia’s data show that memory was the prize inside the box [1]. AI accelerators require vast amounts of high-bandwidth memory, and the industry’s ability to produce and package that memory has become a gating factor for deployment. When memory suppliers shift production priorities toward AI-grade products, supply-demand balances change, average selling prices rise, and revenue expands even faster than unit growth [1].

That dynamic creates powerful operating leverage for suppliers, but it also raises risk for customers. If memory captures a larger share of system cost, then server makers, cloud providers and device brands must decide whether to absorb the increase, pass it on, delay procurement or redesign products around lower memory intensity. The same price mechanism that boosts semiconductor revenue can become a tax on downstream innovation.

The risk is already visible in market behavior. A September 12 Bloomberg report carried by the Taipei Times said shares of SK hynix and Samsung Electronics fell after DeepSeek’s latest AI model raised doubts about the strength of high-bandwidth-memory demand, with SK hynix down 2.2% and Samsung down 3.5% in that session [3]. The report also said investors broadly viewed the development as a short-term setback, while some fund managers remained positive on South Korean memory makers because of earnings, valuations and continuing Big Tech spending [3].

That episode illustrates the new sensitivity of the cycle. A model architecture update, a hyperscaler capex comment or a hint that inference can be done with less memory can move expectations for an entire supplier group. The revenue base is larger, but the narrative risk is sharper.

A broader recovery, but not a risk-free one

Omdia’s 2Q26 release makes clear that the semiconductor recovery has become broader and more lucrative [1]. Non-memory chips grew above seasonal patterns, microprocessors outperformed their usual second-quarter trajectory, and total industry revenue for the first half of 2026 reached $752 billion [1]. Omdia also forecasts semiconductor revenue will surpass $500 billion in 3Q26, taking revenue for the first three quarters of 2026 above $1.25 trillion [1].

That forecast would be extraordinary: Omdia says the first three quarters of 2026 would be 50% higher than total semiconductor revenue for all of 2025 [1]. If achieved, it would confirm that AI infrastructure has pulled the industry into a scale of quarterly revenue that would have seemed implausible only a few cycles ago.

But the same concentration that makes the numbers spectacular also makes them fragile. The industry is increasingly exposed to AI-capex slowdowns, memory pricing swings and investor skepticism about whether the demand curve can keep steepening. The Taipei Times report on Korean memory shares shows how quickly enthusiasm can cool when the market sees evidence that future AI models may require less HBM per task [3].

The better interpretation is not that the boom is over, but that the market has become more option-like. Upside comes from continued AI deployment, rising inference workloads and persistent supply constraints. Downside comes from efficiency breakthroughs, delayed data-center construction, procurement pauses, or a sharper-than-expected normalization in DRAM and NAND pricing.

What to watch next

The most important near-term marker is Omdia’s 3Q26 forecast. If semiconductor revenue really does move above $500 billion in the third quarter, the industry will have crossed from recovery into a new revenue regime [1]. Investors and customers should watch whether memory’s share keeps rising, whether non-memory semiconductors continue to grow above seasonal patterns, and whether microprocessor revenue remains tied to AI infrastructure demand [1].

The second marker is volatility in memory expectations. The Korean memory-stock reaction to DeepSeek shows that the market is already debating whether AI demand is structurally unlimited or merely exceptionally strong for now [3]. In this environment, supplier discipline matters as much as demand. Too much capacity could recreate the old memory-cycle bust; too little could push costs through the rest of the technology stack.

For now, Omdia’s message is clear: the semiconductor industry has reached a record $425 billion quarter, and memory has become the dominant engine of the expansion [1]. AI has not just revived the chip cycle. It has repriced it, reordered it and made bandwidth the industry’s most valuable scarce resource.

Sources from the last 72 hours

1. [1]Omdia: Global Semiconductor Revenue Surges 31% to Record $425bn in 2Q26 | WEEK/HOIABCSep 14, 2026, 8:00 AM UTC
2. [2]DeepSeek tests South Korean memory makers’ stocks with demand doubtsSep 11, 2026, 4:00 PM UTC
3. [3]Omdia: Global Semiconductor Revenue Surges 31% to Record $425bn in 2Q26Sep 14, 2026, 8:00 AM UTC

AI-generated article based on recent web research, then preserved as a dated editorial snapshot.