Applied Materials builds for 2030: plan a flat cost curve

Chip earnings tell you what happened. Chip equipment earnings tell you what management believes about the next three years, because a tool ordered today produces wafers in 2028. Applied Materials reported a record quarter on August 13 and said it is investing in manufacturing capacity to serve demand through the end of the decade. That is the useful sentence for anyone building an automation business case right now.

What actually happened

Applied Materials — the largest maker of the equipment that fabs use to build chips — posted Q3 FY2026 revenue of $9.12 billion, up 25% year over year. GAAP EPS came in at $3.17 (up 43%); non-GAAP EPS at $3.50 (up 41%). Non-GAAP operating margin reached 34.0%, with gross margin above 50%.

Segment detail:

• Semiconductor Systems: $7.04 billion, 37.7% operating margin
• Applied Global Services: $1.78 billion, 30.1% operating margin

Guidance for Q4 is $10.25 billion ± $0.5 billion in revenue and non-GAAP EPS of $4.02 ± $0.20 — a sequential step up from a record quarter, not a plateau.

CEO Gary Dickerson tied it to AI adoption driving demand for the company's materials engineering tools. The forward-looking line came from CFO Brice Hill, who said the company is making additional manufacturing capacity investments to support projected demand through the end of the decade, per the prepared remarks and earnings release.

Why it matters for your business

We have written before that you should stop pricing in cheaper tokens. This is the version of that argument with a date attached.

The supply chain runs in one direction: equipment maker → fab → chip → cloud → your API invoice. Each step adds lag. Tools shipping in 2026 become fab capacity in 2027 and volume wafers around 2028. So Applied Materials guiding up and expanding its own capacity for demand "through the end of the decade" is not a statement about this quarter. It is the toolmaker saying it does not expect the constraint to clear inside four years.

For a small business running automation, that changes one specific thing: the shape of the cost curve in your model.

Underwrite flat, not declining. If your automation ROI only works because you assumed per-token or per-hour compute drops 40% by 2028, the project is not approved — it is a bet on a supply glut that the people building the supply are not planning for. Run the same model at today's prices. If it still clears, build it. If it only clears on the discount, you have a financing problem disguised as a technology decision.

Buy optionality instead of waiting. The right response to a flat cost curve is not to delay. It is to build so that cost improvements come from your side of the wire — caching, routing cheap models to easy work, trimming context, batching. Those are engineering wins you own, and they compound whether or not vendor prices ever move. We have covered how much the harness alone is worth.

Keep contracts short. In a market where the constraint is expected to persist, multi-year committed-spend deals at today's list price are the vendor's hedge, not yours. Prefer annual terms with the ability to move. Portability is worth more than a 15% commit discount when you cannot forecast the price three years out.

One caveat worth stating: equipment demand is a leading indicator, not a promise. Capex cycles overshoot. If AI demand cools, tool orders correct before wafer prices do. But planning for the boom to break is a worse default than planning for it to hold.

Sources: Applied Materials Q3 FY2026 prepared remarks, Applied Materials Q3 FY2026 earnings release.