A record quarter, not a normal cycle
Global semiconductor revenue surpassed $425 billion in 2Q26, setting another all-time quarterly high as the market grew 31.4% from the previous quarter, according to newly released Omdia research distributed on September 14, 2026 . The headline number matters not only because it is large, but because it confirms that the current semiconductor expansion is no longer just a recovery from the 2023–2024 inventory cycle. It is a structural acceleration driven by AI data centers, memory shortages, custom silicon, networking, storage and leading-edge foundry capacity.
Omdia says the 31.4% quarter-over-quarter rise exceeded the previous sequential growth record of 29.2% set in 1Q26, meaning the industry has now broken its own growth record in back-to-back quarters . Since Omdia began quarterly semiconductor tracking in 1Q02, only 10 of 97 quarters had posted sequential revenue growth above 10%, and the last four quarters starting in 3Q25 have all delivered double-digit growth . That sequence is a useful warning against treating the $425 billion quarter as an isolated spike. It looks more like a market reset.
The first half of 2026 reached $752 billion in semiconductor revenue, already exceeding every full-year industry total except 2025, according to Omdia . Omdia also expects revenue to surpass $500 billion in 3Q26, which would lift the first three quarters of 2026 above $1.25 trillion and put the industry roughly 50% ahead of the entire 2025 total . If that forecast is met, the record $425 billion quarter will be remembered less as a ceiling than as the moment the industry’s revenue base visibly moved upward.
Memory is the engine of the surge
The most important driver is memory. Omdia says memory ICs alone accounted for more than half of semiconductor revenue in 2Q26, reflecting sharp revenue growth in DRAM, NAND and NOR chips . DRAM, NAND and NOR each recorded their highest second-quarter sequential growth since Omdia began tracking the market, and all three reached their highest recorded quarterly revenue in 2Q26 . That is a remarkable concentration of growth in one product family, but it is not surprising in the context of AI infrastructure.
AI systems consume memory in multiple ways. High-bandwidth memory is central to accelerators, DDR5 supports server platforms, and NAND is required for fast storage and data pipelines. The industry’s production mix has shifted toward higher-value AI-related memory, and that has affected supply-demand balances and average selling prices . Omdia’s framing is clear: the market is not simply shipping more bits; it is selling memory into a tighter and more expensive environment.
That memory-led surge also explains why the top-line semiconductor number can rise so violently while some downstream device categories remain pressured by component costs. A shortage that boosts chip revenue can simultaneously squeeze PC makers, smartphone vendors and server buyers. The $425 billion quarter is therefore bullish for semiconductor suppliers, but it is not automatically benign for every company that depends on chips.
Non-memory chips are also outperforming
The record is not only a memory story. Omdia says non-memory semiconductor revenue grew by more than 10% quarter over quarter in 2Q26, far above the historical second-quarter seasonal pattern of just over 3% based on Omdia data from 2002 through 2025 . That matters because it shows the AI buildout is spilling across the semiconductor stack rather than being confined to DRAM and NAND.
Microprocessor revenue rose 16% quarter over quarter in 2Q26, compared with a seasonal growth pattern of around 1%, as processors became more important across AI infrastructure . In practice, AI data centers require not only GPUs or custom accelerators, but also CPUs, networking chips, power management ICs, storage controllers, optical components and a long list of supporting semiconductors. The “AI chip” story is therefore broader than a single category.
This helps explain why Omdia describes memory as the primary driver while also emphasizing above-seasonal growth in other semiconductor components . The industry’s revenue mix is being pulled upward by a cluster of requirements: compute density, memory bandwidth, storage speed, network throughput and power efficiency. A data center rack is becoming a semiconductor-dense capital asset.
Foundry momentum confirms the demand signal
Fresh company-level signals point in the same direction. Taiwan Semiconductor Manufacturing posted record August revenue of NT$514.8 billion, up 53.3% year over year, as AI chip demand pushed its foundry business to new highs, according to a September 12 analysis by Simply Wall St . The same report noted that TSMC’s stock price had already reflected heavy investor interest, with strong year-to-date and one-year total shareholder returns tied to AI upside and higher expectations .
That matters because TSMC is a central manufacturing partner for many leading AI processors and custom silicon designs. When the foundry layer reports record sales, it reinforces the idea that the Omdia market-wide number is supported by real production demand rather than only by accounting effects or one-off pricing. The $425 billion quarter is visible at the foundry level, at the memory level and at the system level.
TSMC’s record also shows the capacity side of the story. The market is rewarding advanced manufacturing, but the same strength raises questions about capital intensity, geopolitical exposure and whether demand can keep absorbing higher wafer, packaging and memory costs. The record revenue number is therefore both an achievement and a stress test.
Custom AI silicon is becoming a second growth pillar
Broadcom’s latest AI semiconductor figures add another layer to the record. A September 12 report said Broadcom’s AI semiconductor revenue reached $16.7 billion in its fiscal third quarter, up 221% from a year earlier and equal to roughly 56% of total revenue . The same report said Broadcom’s broader semiconductor solutions revenue reached $20.8 billion, up 127% year over year, driven largely by custom AI accelerators and networking products .
That is important because the AI infrastructure market is diversifying beyond merchant GPUs. Hyperscalers and frontier AI companies increasingly want custom accelerators, specialized networking and optimized rack-scale systems. Broadcom’s remaining performance obligations reportedly rose to about $179.2 billion as of August 2, suggesting new contracts were being signed faster than revenue was being recognized .
Insider Monkey’s September 13 analysis similarly framed Broadcom’s custom AI hardware as a major part of the current cycle, noting that management expected accelerated Ironwood shipments to Anthropic in the fourth quarter and that the company had reported $16.7 billion of AI semiconductor revenue in the fiscal third quarter . The same analysis warned that investors should distinguish shipped hardware, consumed capacity and the profit retained by each partner . That caution is useful: the semiconductor boom is real, but the economics may differ significantly between chip suppliers, cloud platforms and AI model companies.
System demand is strong, but shortages are slowing conversion
The $425 billion revenue record also reflects a market where demand is outrunning component availability. A September 12 report on Hewlett Packard Enterprise said the company ended fiscal Q3 2026 with a $7.6 billion AI backlog, up 14% from the previous quarter, but memory shortages were slowing how quickly orders could become revenue . The report said DDR5 DRAM and NAND flash remained in short supply and that constraints were expected to persist into 2027 .
That backlog dynamic is critical. It shows that chip revenue is being supported not only by speculative buying, but by system-level orders from enterprise and government customers. HPE’s quarterly revenue reportedly reached $12.2 billion, up 34% year over year, while AI systems orders reached $2.4 billion and cumulative AI systems bookings crossed $16 billion by mid-2026 . More than 60% of cumulative AI orders came from enterprises and sovereign clients rather than hyperscale cloud giants, suggesting that AI infrastructure demand is broadening .
This broadening helps support the Omdia view that 3Q26 could remain strong. If hyperscalers, enterprises and governments all compete for memory, accelerators, networking and storage, supply tightness can last longer than a single procurement cycle.
What the $425 billion record means
The semiconductor industry has entered a phase where quarterly revenue numbers resemble full-year totals from earlier eras. The $425 billion figure is not just a milestone; it is a signal that AI infrastructure has changed the scale, urgency and pricing power of the chip supply chain. Memory suppliers are benefiting from scarcity, foundries are benefiting from advanced-node demand, custom silicon vendors are converting hyperscaler designs into large revenue streams, and system companies are building backlogs faster than they can ship.
The risk is that such rapid growth can create overconfidence. Omdia’s numbers show extraordinary momentum, but they also point to a market heavily influenced by memory pricing and AI infrastructure spending . If AI deployment slows, financing tightens, or new capacity arrives faster than expected, today’s record conditions could become tomorrow’s comparison problem. For now, however, the current evidence points in the opposite direction: 2Q26 set a $425 billion record, 1H26 reached $752 billion, and Omdia expects 3Q26 to move above $500 billion .
The simplest reading is that semiconductors have become the financial center of the AI economy. The more complex reading is that this boom is uneven, capacity-constrained and increasingly strategic. Both readings are true.