ASML and TSM both report this week — why that's not a coincidence for AI chips

Two of the most structurally important companies in the AI chip supply chain report earnings back-to-back this week: ASML on Wednesday, October 14 (before market open), and TSM (Taiwan Semiconductor) on Thursday, October 15 (before market open). That sequencing is not a scheduling coincidence worth skipping past — in AI007's own supply chain map, ASML sits directly upstream of TSM, and the order they report in roughly mirrors the order their businesses actually feed into each other.

Why these two are linked, not just adjacent

ASML makes the lithography systems — including the EUV machines it has an effective monopoly on — that chipmakers use to print the smallest, most advanced transistor patterns onto silicon. TSM is one of the handful of companies in the world that actually buys and runs those machines at scale, using them to fabricate the 3nm and 2nm chips that power NVIDIA's GPUs, AMD's accelerators, and Apple's silicon. In AI007's AI Supply Chain Universe, that relationship is not descriptive flavor text — it is a direct, mapped connection: TSM is listed as one of ASML's downstream customers, and ASML is listed as one of TSM's upstream suppliers.

Downstream of TSM sits essentially the rest of the AI compute stack: NVIDIA, AMD, Broadcom, Apple, Qualcomm, Marvell, Arm, Intel, Astera Labs, Monolithic Power. Whatever happens at the fabrication layer — capacity, yields, pricing, lead times — ripples into all of them before it ever shows up in their own earnings.

The numbers behind why this week carries weight

AI007 tracks ASML at a 9.8/10 criticality score with a market cap of $307.6B, and TSM at a 9.9/10 criticality score with a market cap of $960.0B — combined, roughly $1.27 trillion in market value reporting across two days, concentrated in a part of the chain almost nobody outside semiconductors thinks about day-to-day. The one meaningful asymmetry: AI007 flags ASML's geopolitical risk as "Low" (Netherlands-based) against TSM's as "Extreme" — a reflection of Taiwan's position as both the center of advanced chip fabrication and a live geopolitical flashpoint.

Both companies have had a strong run of beating estimates. Per Twelve Data's earnings history, ASML has beaten consensus EPS in four of its last five quarters — most recently by 9.1% in July — with its one miss (-2.8%) back in January. TSM's streak is cleaner: five straight beats, ranging from 4.8% to 26.8%, most recently 10.8% in July.

What to actually watch, not predict

Rather than guessing at a beat or miss, the more useful read on both reports is what they say about capacity, not just revenue:

• ASML's order backlog and bookings mix. Lithography orders today are fab capacity 12-24 months from now — one of the cleanest leading indicators in the entire AI chip chain for how much new fabrication capacity is actually being built, versus just talked about.
• ASML's commentary on export restrictions and regional demand. Where orders are coming from matters as much as how many there are, given how exposed the EUV supply chain is to trade policy.
• TSM's capacity utilization and capex guidance for 3nm/2nm. This is the number that tells you whether AI GPU supply constraints are easing or tightening.
• TSM's CoWoS advanced packaging commentary. Advanced packaging, not raw wafer fabrication, has been the actual bottleneck behind NVIDIA's GPU supply — TSM's own capacity updates here are arguably more market-moving than its headline EPS.

Because ASML reports a day earlier, its bookings commentary is effectively a preview of what TSM's own capacity story might look like — the same upstream-before-downstream logic AI007's Supply Chain Universe is built around, just compressed into a 24-hour window this particular week.

Both companies — and the rest of the chain on either side of them — are mapped live in AI007's AI Supply Chain Universe, under Layer 3 (Semiconductor Equipment) and Layer 4 (Chip Fabrication & Advanced Packaging), with their real upstream and downstream connections, criticality scores, and ownership data.