Global semiconductor revenue passed $425 billion in the second quarter of 2026, up 31.4% from the first, according to Omdia. That’s the fastest sequential growth the firm has recorded since it started tracking the market in 2002. The previous record, 29.2%, was set one quarter earlier. First-half revenue came to $752 billion, more than any full year in the industry’s history except 2025. Omdia expects the third quarter to clear $500 billion, which would take the first nine months past $1.25 trillion, about 50% above all of 2025.
The forecast is the number to read twice. $500 billion on a $425 billion base is growth of roughly 17.6%. That’s still a double-digit quarter, the fifth in a row. It’s also about half the pace just reported, and the slowest quarter since the streak started in 3Q25. The headline says another record. The arithmetic says deceleration.
Some perspective on how unusual the run has been. Of the 97 quarters Omdia has tracked, only 10 grew more than 10% sequentially, and four of those ten are the last four. Second quarters are normally dull. Between 2002 and 2025 the typical Q2 gain for non-memory chips was a little over 3%. This year non-memory grew by more than 10%, and microprocessors rose 16% against a seasonal norm of about 1%.
Memory did most of the lifting, though. Memory ICs made up more than half of Q2 revenue, and DRAM, NAND and NOR each posted record quarterly revenue along with their strongest second-quarter growth on record. Working backwards gives a rough split. A $425 billion quarter at +31.4% puts Q1 near $323 billion. If memory was just over half of Q2, it was above $212 billion. Non-memory grew a bit more than 10%, so it went from roughly $193 billion to $212 billion. Memory is what’s left: about $130 billion rising to $212 billion, more than 60% growth in one quarter. Of the roughly $100 billion the industry added sequentially, memory supplied about four dollars in every five. The inputs are floors (“more than half”, “over 10%”), so treat the split as approximate. The shape isn’t in doubt.
And that memory growth was mostly price. Full-year 2026 estimates have DRAM average selling prices up more than 240% on bit growth of around 25%. Suppliers keep moving wafers toward HBM and server DRAM, which has pushed shortages down through the product generations. Even DDR2 contract prices were expected up 55-60% in Q2 and another 35-40% in Q3.
Here’s the problem for the third quarter. Contract price surveys still show increases, just much smaller ones. Conventional DRAM is expected up 13-18% quarter on quarter in 3Q26, NAND up 10-15%, mobile DRAM up 8-13%. Two things are doing the braking. Consumer buyers have hit their limit on what they’ll pay: notebook retail prices are rising across the board and full-year PC shipments will take the hit. And the largest US cloud buyers have signed multi-year supply agreements that stop suppliers raising prices on them. From Q3 on, fresh price gains have to come from customers without those contracts, or from volume sold outside them. On top of that, a server CPU shortage slowed rack assembly in Q2 and left DRAM piling up in cloud inventories.
None of that loosens supply. The market is still tight. What changes is the rate, and memory revenue growth tracks price growth closely.
Run the rough numbers forward and the gap shows up. Say non-memory repeats its 10% quarter and reaches about $233 billion. Say memory revenue grows in line with contract prices, call it 15%, to about $245 billion. That’s $478 billion. Omdia’s $500 billion needs roughly $22 billion more. To get there, memory revenue has to grow closer to 25%, well above contract price increases, or non-memory has to speed up again. The likely source is mix. HBM carries a much higher price per bit than standard DRAM, so every wafer shifted into it lifts average revenue per bit even where contract prices are flat. Bit shipments add the rest. So the $500 billion quarter depends less on pricing than on HBM volume and on processors and networking chips holding their Q2 pace.
The WSTS count, which runs lower than Omdia’s, suggests that’s achievable. It put Q2 sales at $403.3 billion, up 35.1%. June has since been revised up to $137.9 billion on a three-month average basis, and July came in at $146.8 billion, up 6.4% on the month and 135% on the year. That was the 17th straight monthly gain, and the first seven months of 2026 already exceed the industry’s best full year. Apply Omdia’s implied 17.6% to the revised WSTS Q2 base and the September three-month average needs to land near $162 billion. From July that’s about 5% a month, below the 6-10% monthly gains of the last several prints.
For equities this is the part that matters. Memory stocks trade on the rate of change, and they spent much of the summer well below their June highs while revenue kept breaking records. A quarter that grows 17.6% after one that grew 31.4% is a strong quarter and a slowing one at the same time. Both descriptions will be true in October. The market has already shown which one it listens to.
The next test is the August WSTS figure, due early October. It needs a three-month average around $154 billion to stay on track for $500 billion.
Anything below that and the slowdown starts in Q3.
