Question: Does Teradyne (TER) warrant its own sister deep-dive alongside the 2026-05-25 ATEYY (Advantest) HBM-test-equipment deep dive, or should it stay as cohort context? Verdict: stays-cohort-context — route to memory-supercycle (demand-side housing) and supply-chain-traces (alongside ATEYY); add TER to a watchlist so it sits beside ATEYY in the cohort, but do not open a separate conviction deep-dive.

Filed 2026-05-25 as a follow-up from 2026-05-25-ateyy-deep-dive-hbm-test-equipment.md. This is the ~30-min head-to-head that deep dive explicitly deferred ("No Teradyne comparative... that's a separate ~30-min comp note").

What we're asking

The ATEYY deep dive's core thesis is that KGSD (known-good-stacked-die) HBM test scales with HBM stack volume, not memory bit volume — and that ATEYY is the cohort laggard sitting on that structurally accelerating end-market. It framed TER twice: (a) competitive-positioning table — "Teradyne (TER): Smaller memory share; broader SoC + memory + analog test platform"; (b) bear case #5 — "the obvious cleaner US-listed alternative... TER is more SoC-pure with memory mix. If you want HBM-test laggard, you want ATEYY."

The decision hinges on one discriminator: is TER a genuine HBM-stack-test competitor to Advantest (in which case it deserves its own sister thesis), or is its memory-test franchise a different layer (DRAM/flash final test) that merely sits adjacent? If the former, a sister deep-dive is warranted because the HBM-test market is then contested duopoly, not Advantest near-monopoly, which materially changes ATEYY's bull case. If the latter, TER is cohort context.

What we found

Price truth (internal data, data_as_of 2026-05-29; ai-infrastructure.json / supply-chain-traces.json)

Tape note: ATEYY has rolled further since the 2026-05-25 deep dive ($169.43 → $165.13, RSI 46.6 → 43.7, now weak-down and below SMA20/SMA50). The cohort-laggard "flat tape during a cohort rally" read in that deep dive has degraded toward a "rolling, not flat" read in the four trading days since — worth flagging to whoever owns the ATEYY position thesis, though out of scope for this comp. TER is the strong cohort member; ATEYY remains the laggard. This is price-truth confirmation of the deep dive's own laggard framing — but, as below, it does not mean TER is "the safe version of the same trade."

Financials (Massive, US filer, 10-K/10-Q — pulled 2026-05-29)

TER income statements (Massive income-statements.json):

Sharp sequential ramp ($0.65B → $1.28B over four quarters), gross margin ~61%, net margin ~31% in Q1'26. Market cap ~$58.6B (info.json). TER is riding the same AI-test upcycle as Advantest. (Ratios are gated behind Massive premium — 403; not needed for this comp.)

Segment mix — the decisive comparison (event truth: TER Q1 2026 earnings call / 8-K)

Teradyne Q1 2026 Semiconductor Test = ~$1.1B (crossed $1B for the first time):

• SoC test: $882M (~80% of test) — compute-driven (compute = 75% of SoC), first merchant-GPU win
• Memory test: $203M (~19% of test) — explicitly "driven by HBM and DRAM"
• Plus Robotics $91M (+32% YoY), Product Test $80M (+8% YoY)
• AI-related demand ≈ 70% of total revenue
• Memory test "demand even stronger than three months ago, with HBM and DRAM acting as accelerators"
• TER's memory tester for DRAM and HBM is the Magnum platform (Magnum 7 = latest gen)
• TER guides to "low single-digit" memory-test share gains for 2026 — i.e. it is a share challenger ("win-back" phase vs Advantest), not the incumbent

Advantest (ATEYY/6857.T) FY2026 guide (Advantest IR):

• SoC test systems ~¥999.5B (~95% computing/communications/AI)
• Memory test systems ~¥201.0B, of which ~90% is DRAM
• → Advantest's own mix is also ~83% SoC / ~17% memory by revenue
• Advantest holds ~60–70% of the memory-test market and ~35–45% of SoC test (2023–24 share estimates)
• CY2026 market: SoC tester TAM $8.7–9.5B (up from $6.9B); memory tester TAM $2.2–2.7B (up from $2.1B)

The discriminator, resolved

The ATEYY deep dive's framing — "ATEYY dominates HBM/KGSD; TER is SoC-pure with a memory tail" — is partially wrong on the HBM-test point. Three findings:

1. TER is a genuine HBM-stack-test competitor, not a different layer. TER's memory test ($203M/qtr, growing) is explicitly HBM+DRAM via the Magnum platform, and trade press credits TER with >50% share of HBM wafer-stack test in its current win-back phase. This is the same KGSD-adjacent test gate the ATEYY thesis is built on — not a separate DRAM/flash-only final-test layer. The HBM-test market is contested, not an Advantest monopoly.

2. But both companies are ~80/20 SoC-heavy by revenue. Neither is an "HBM-test pure-play." ATEYY's memory segment is ~90% DRAM (not HBM-pure either); its HBM exposure is the fast-growing slice within a DRAM-dominated memory segment that is itself ~17% of revenue. The deep dive's implicit "ATEYY = HBM-test pure-play laggard" is an overstatement; ATEYY is a DRAM-test-heavy name with HBM optionality, sitting inside a company that is mostly an SoC tester. TER is the mirror image: slightly larger absolute memory-test dollars ($203M/qtr) than the ATEYY framing implies, also wrapped in an 80% SoC body.

3. Advantest's memory share lead is real but DRAM-weighted; HBM-stack test specifically is where TER is taking share. Advantest's ~60–70% memory-test share is anchored in DRAM. The HBM-stack-test sub-segment — the actual KGSD thesis — is precisely where TER claims >50% and is gaining. So on the narrow axis the ATEYY thesis rests on (HBM-stack test scaling with stack volume), TER is arguably the share winner, not the laggard alternative. That is a material tension with ATEYY bull-case point #2 ("ATEYY dominates this segment ~50%+").

Customer overlap

Both sell into the same three memory customers — Samsung, SK hynix, Micron — for memory/HBM test, plus the broad SoC roster (Apple, Qualcomm/Snapdragon, NVDA, networking). There is no customer differentiation that would justify two independent theses; they are competing for the same HBM-stack-test budget at the same OSAT/IDM lines. The Samsung NSEU-strike demand-pull and TrendForce contract-price mechanics in memory-supercycle drive both names' test demand identically.

Verdict + reasoning

TER stays cohort context — no sister deep-dive. Reasoning:

• Not differentiated enough to carry an independent thesis. TER and ATEYY are mirror images of the same trade: ~80% SoC / ~19% memory test, same three memory customers, same HBM-stack-test gate, same AI-test upcycle, same memory-supercycle demand driver. A second full deep-dive would restate the ATEYY thesis with the sign flipped on the tape. The right artifact is exactly this comparative, not a parallel conviction doc.

• The comparison actually sharpens — and partially corrects — the ATEYY thesis rather than spawning a new one. The single most important output here is the correction to ATEYY bull-case #2: HBM-stack test is a contested market where TER is the share-gainer (>50%, "win-back"), not an Advantest ~50% monopoly. That tightens the ATEYY bear case (bear #1c — "Teradyne taking memory-test share at the margin" — is now confirmed by TER's own guidance, not speculative). This belongs as a cross-reference / log note on the ATEYY thesis and the memory-supercycle perspective, not a new deep-dive.

• If anything deserved a sister deep-dive it would be TER on its own merits, not as "the ATEYY alternative" — TER is a strong-up $58.6B name (+377% 1y, +60% vs SMA200, RSI 54.5, golden cross) with a genuine merchant-GPU SoC-test win-back story and Robotics optionality. But that is a different thesis (SoC-test + GPU-test + robotics), and TER is already a large-cap, well-covered name — it does not pass the "genuine coverage gap" bar the ATEYY deep dive cleared. If a user wants that, it's a fresh /deep-dive TER on the SoC/GPU-test angle, not a memory-test sister to ATEYY.

• Housing correction: the ATEYY deep dive asserted "TER is in ai-power-delivery key_tickers." That is inaccurate — TER does not appear in the ai-power-delivery README (which covers power-delivery semis: MPWR, BESI, Vertiv, Eaton-class). TER is currently only in the ai-infrastructure watchlist. Its correct cohort home is memory-supercycle (demand) + supply-chain-traces (sits beside ATEYY as the strong-side mirror of the HBM-test gate).

Mutations / follow-ups (described, not applied — producer-only)

1. Watchlist add (recommended): add TER to supply-chain-traces so it sits beside ATEYY as the HBM-test cohort's strong-side mirror, and/or memory for the demand-side housing. Command for the lead/reviewer: cd stonks && deno task watchlist:add supply-chain-traces TER. (TER already has fresh OHLC + is in ai-infrastructure; this is purely cohort-grouping.)
2. Cross-reference on ATEYY thesis: log a note on 2026-05-25-ateyy-deep-dive-hbm-test-equipment / memory-supercycle that HBM-stack test is contested (TER >50% claimed share, "win-back," confirmed memory-share-gain guidance) — this hardens ATEYY bear-case #1c and bounds the bull case. Suggested via beta:emit log-entry --perspective memory-supercycle.
3. Housing fix (low priority): correct the "TER in ai-power-delivery key_tickers" assertion in the ATEYY deep dive (it is not) if/when that artifact is next touched.
4. Optional future: a standalone /deep-dive TER on the SoC-test + merchant-GPU-test + Robotics angle is defensible on its own merits, but is out of scope for this memory-test comparative and not blocking.

Sources

• research/deep-dives/2026-05-25-ateyy-deep-dive-hbm-test-equipment.md — the parent deep dive being compared against; competitive-positioning table + bear-case #1c/#5.
• stonks/data/summaries/ai-infrastructure.json (TER) + supply-chain-traces.json (ATEYY), data_as_of 2026-05-29 — price/RSI/trend truth.
• stonks/data/stocks/TER/info.json — segment structure (Semiconductor Test / Robotics / Other), market cap $58.6B, Magnum memory-test platform description.
• Massive financials (income-statements.json, US filer 10-K/10-Q) — TER quarterly revenue/margin ramp, Q1 2026 $1.28B / EPS $2.53.
• TER Q1 2026 earnings call / 8-K (web): SoC $882M vs Memory $203M of ~$1.1B test; memory "driven by HBM and DRAM"; Magnum 7; "low single-digit" memory-share-gain guide; AI ~70% of revenue. (Nasdaq/Zacks, BigGo/Q1 transcript summary, Motley Fool transcript)
• TER >50% HBM wafer-stack test share, Advantest ~58–70% memory / 35–45% SoC share (trade press / Zacks via finviz).
• Advantest FY2026 guide: SoC ~¥999.5B / Memory ~¥201B (~90% DRAM); CY26 SoC TAM $8.7–9.5B, memory TAM $2.2–2.7B (Advantest IR FY2025 2Q results).
• research/perspectives/2026-05-19-memory-supercycle/README.md — shared demand-side driver (Samsung NSEU strike, TrendForce contract prices) for both names' test demand.
• research/perspectives/2026-05-03-ai-power-delivery/README.md — checked; TER is not a key_ticker there (housing correction).