The semiconductor cycle

The semiconductor cycle is expanding on several measures, but the commercial opportunity depends on where a business participates. SIA’s latest available sales release, WSTS’s annual forecast and SEMI’s equipment reports answer different questions. Read together, they support an investment cycle with pronounced differences across product categories, rather than a single growth rate that every supplier can use.

Three decisions follow. Procurement teams should separate component demand from pricing and inventory exposure. Equipment suppliers should evaluate committed projects and qualification schedules before treating a market forecast as their order book. Strategy teams should test how much of their addressable market actually depends on advanced memory and computing.

The strongest claim supported here is direction and dispersion: reported sales and equipment billings are rising, while the product outlook is exceptionally uneven. These public releases do not establish supplier margins, end-user consumption, inventory quality or the share of revenue attributable specifically to AI.

The September 4 SIA release puts the July-ending sales average at $146.8 billion ¹, with June at $137.9 billion ² and the prior July at $62.5 billion ³. It reports year-over-year growth of 135.1% ⁴. All three sales levels use the same release vintage.

These are three-month moving averages of dollar sales. They are unsuitable as raw monthly observations in a model of shipments, production or inventory drawdown. Multiplying July by twelve would produce an illustrative run rate, not a measured annual market; summing overlapping averages would also obscure the underlying timing. This report does neither.

For operational planning, maintain two separate files: the observation period and the date the figure was published. Preserve an earlier release when updating a historical comparator. Otherwise an apparent change in momentum may partly reflect a revision. The appropriate follow-up is a consistent vintage history with price, product-mix and volume evidence, rather than a causal claim based on revenue alone.

July year-over-year growth was positive in all five SIA regional groupings shown below. The Americas lead at 171.3% ⁵, followed by Asia Pacific / All Other at 134.9% ⁶, China at 123.6% ⁷, Europe at 85.2% ⁸ and Japan at 50.8% ⁹.

This is evidence of breadth across sales markets. It does not locate the fabs producing the chips, identify the headquarters of their suppliers or measure regional self-sufficiency. The residual Asia Pacific / All Other category must also remain separate from China and Japan; replacing its label with “Asia” would change the meaning.

For market entry, treat this chart as a screening signal. The next analysis should connect customer location to device type, qualification requirements and actual purchasing entities. A faster-growing region can still be difficult for a particular supplier to serve. Regional percentage changes also say nothing by themselves about the incremental dollars available in a narrowly defined product niche.

WSTS’s Spring 2026 forecast projects annual chip sales of approximately $1.51 trillion ¹⁰, or about 90% growth ¹¹. Its selected category forecasts range from roughly 250% for memory ¹² to 37% for logic ¹³, 10% for analog ¹⁴ and 8% for discrete semiconductors ¹⁵. These are projections, not completed-year results.

The analytical implication is concentration of forecast risk. A planning model built from the headline rate could badly misstate the outlook for a business selling analog or discrete products. Memory needs separate assumptions for price, bit demand, product mix and customer commitments; the public category total does not isolate those contributions or provide an HBM-only market size.

A useful commercial model therefore starts at the product line. Identify the revenue exposed to each category, document the relevant forecast vintage, and test a lower-growth case without assigning unsupported probabilities. The chart compares selected categories on a common annual revenue-growth basis; it deliberately avoids suggesting that these four categories exhaust the market.

SEMI reports worldwide equipment billings of $36.55 billion in Q1 ²¹ and $40.53 billion in Q2 2026 ²². These historical quarterly results are evidence of spending reaching equipment suppliers. They do not show how quickly purchased tools become productive capacity, or how much profitable chip output that capacity will eventually deliver.

In its July forecast, SEMI projects $165.9 billion of total equipment sales in 2026 ¹⁶ and $229.5 billion in 2028 ¹⁷. The 2026 breakdown is $143.9 billion for wafer fab equipment, $15.3 billion for test and $6.7 billion for assembly and packaging ^{18 19 20}.

The allocation is useful for supplier positioning, while scale alone is a poor measure of scarcity or bargaining power. A smaller process segment may still be essential to a customer's production plan. Validate that proposition with customer schedules, qualification status and alternative suppliers. Do not convert the equipment forecast into a chip-revenue forecast: installation, process development, utilization and yield sit between the two.

SEMI’s silicon-wafer series records 3,573 million square inches shipped in Q2 2026 ²³, up 7.4% year over year ²⁴. It adds a physical measure to the revenue and equipment evidence.

The growth rate is much smaller than the chip-sales increases discussed earlier, but the periods and measurement bases differ. That contrast cannot quantify a semiconductor price increase. Wafer area moves through a manufacturing process before chips are sold; different device designs, die sizes, yields and product values also prevent a fixed conversion from substrate area to revenue.

Use the series to ask better questions. Is a supplier's demand driven by more material, more processing per wafer, or a different product mix? Is shipment timing aligned with production and customer acceptance? A purchasing plan should identify which of those mechanisms matters and seek corresponding evidence. The public area total is a useful diagnostic input, not sufficient proof of a broad consumer-electronics recovery.

Refresh the sales series when the next release arrives, retaining its revised comparators. Track whether momentum remains broad across regions rather than relying on the aggregate alone. For product exposure, update memory and logic assumptions separately from mature product categories and record any change in the forecast's definition or vintage.

For capacity exposure, connect equipment orders and billings to project milestones: delivery, installation, qualification and commercial utilization. Demand evidence should include customer commitments and inventory conditions where available. Neither is established by the public aggregate series in this report.

Three conditional scenarios help discipline decisions. If revenue growth moderates while equipment commitments continue, test utilization and cash-conversion risk. If physical shipments strengthen alongside diversified end-market orders, investigate whether the expansion is broadening. If demand remains concentrated in a few advanced products, test customer concentration and qualification dependencies. These are monitoring scenarios, not predictions or investment recommendations.

Research is frozen at September 22, 2026. This report uses six public primary-source releases: SIA for monthly sales, WSTS for the Spring 2026 outlook, and SEMI for quarterly equipment billings, the July equipment forecast and silicon-wafer shipments. SIA republishes WSTS sales data; those organizations do not provide two independent confirmations of the same series.

Historical observations and forecasts are labeled separately. Dollar figures are nominal US dollars. Monthly labels identify the ending month of the sales moving average; quarterly equipment figures are billings totals. Annual forecasts retain their publication vintage and source rounding. The WSTS release date is confirmed by the linked forecast PDF.

Each numerical claim has a source locator and a minimal verbatim evidence fragment, with fuller measurement context in its definition. The report does not reproduce subscription datasets or infer undisclosed HBM revenue, unit prices, inventories or supplier profitability. Recommendations and scenarios are StatOrigin editorial analysis, informed by the cited evidence and prepared with AI-assisted research.