Skip to content
All theses

chips-compute

ChipsJune 16, 2026medium conviction6 min read
View Chips & Compute names in screener

The structurally important development this cycle is that the chips-compute value chain has bifurcated clearly enough to act on: a durable-rent tier sitting upstream of NVIDIA, and a contestable tier that includes NVIDIA itself. That bifurcation is sharpened by the single most novel signal of this scan — a US-government directive on 12 June 2026 suspending all access to Anthropic's Fable 5 and Mythos 5 model variants (https://www.anthropic.com/news) — which, if it establishes a domestic-model-variant export-control category, converts what has been a market-pricing question about platform-layer scarcity into a regulatory-permission question.

The revenue gap has not closed

The Cahn arithmetic underlying the AI capex ceiling thesis remains unchanged: roughly $840 billion per year implied against less than $50 billion observed, a 16× gap. NVIDIA's fiscal Q3 FY26 revenue of $57 billion — approximately 90% data-centre, with four hyperscaler customers (MSFT, META, AMZN, GOOGL) each exceeding 10% of FY26 revenue and two of them accounting for 36% combined — is impressive in isolation and nowhere near closing that arithmetic. Anthropic's Series H at $65 billion, valuing the company at $965 billion post-money and followed by a confidential S-1 submission to the SEC on 1 June 2026 (https://www.anthropic.com/news), is being priced against a forward revenue trajectory no historical software cohort has demonstrated at this scale. When the S-1 becomes public — expected Q3 to Q4 2026 — it is the cleanest single falsification test for the $200 billion AI-attributable-revenue threshold.

The commoditisation leg of the thesis has simultaneously strengthened. Open-source enterprise deployment extended from 23% to 67% across 2025. DeepSeek-R1 was trained at a verified cost below $6 million, open-source, runnable on Huawei Ascend hardware [arXiv:2501.12948]; Chinese AI's global use share moved from under 1% to roughly 15% over the same window. More structurally, Anthropic's distribution configuration now routes Claude through three competing hyperscaler channels — GCP Vertex AI, Microsoft Azure Foundry, and AWS — plus new enterprise-SI partnerships with TCS (12 June 2026) and DXC (11 June 2026) (https://www.anthropic.com/news). The model lab has settled into the role of content layer inside someone else's distribution stack. The early-2000s cable-vs-content analogy is exact: HBO earned the high gross margin, Comcast earned the volume and the relationship. The 60%-plus AI-service gross margin at two or more hyperscalers for three consecutive years that would refute commoditisation moves further out of reach under this arrangement.

Where rent durably sits

The investable implication is to migrate exposure away from the contestable merchant-GPU sub-tier toward the upstream-monopoly and picks-and-shovels layers where rent compresses least.

At the apex sits ASML (ASML), the sole supplier of EUV lithography globally — its Twinscan NXE:3800E and NXE:5000 High-NA units gate every 2-nm-class node at TSMC, Samsung, and Intel Foundry. Substituting ASML requires a multi-decade national-industrial-policy commitment; Japan's national EUV consortium and China's SMEE programme are both at engineering-development stage, not production. One layer inside that monopoly, Lasertec (6920.T) holds the only EUV-actinic mask-blank inspection franchise — a second-order monopoly inside an already irreplaceable stack. The wafer-fab-equipment oligopolists — Tokyo Electron (8035.T), Applied Materials (AMAT), Lam Research (LRCX), KLA Corp (KLAC) — carry 3-to-5-year requalification cycles that insulate them from displacement.

The binding sub-bottleneck for 2026 to 2028 is not transistor density at N3/N3P — wafer-out is sufficient for Blackwell and Rubin demand — it is TSMC's (TSM) CoWoS-S/L advanced-packaging throughput. CoWoS interposer area, redistribution-layer yield, and hybrid-bond pitch constrain finished-accelerator exits regardless of die availability; the constraint relaxes only after 18 to 36 months of capex and qualification. The picks-and-shovels into that bottleneck are BESI (BESI.AS) on hybrid-bonding equipment and ASM Pacific (0522.HK) on bonding-and-handling — neither features in the typical NVIDIA-centric AI conversation. In HBM memory, SK Hynix (000660.KS), Micron (MU), and Samsung (005930.KS) form a three-supplier oligopoly in tight allocation against demand for HBM4 — 12-Hi stacks, 1024-bit interface, approximately 1.5 TB/s bandwidth — the qualifying memory for Blackwell Ultra and Rubin. Micron's HBM revenue is tracking toward roughly $8 to $10 billion annualised by end of FY2026.

At the accelerator silicon layer the relevant distinction is between NVIDIA (NVDA) — an avoid at cap-weighted peak consistent with both the capex-ceiling and value-migration theses — and the custom-ASIC design houses: Broadcom (AVGO), designing Google's TPU v5p/v6e, Meta's MTIA, and ByteDance silicon; Marvell (MRVL), designing AWS Trainium and Inferentia. Hyperscaler custom-ASIC decisions are 18-to-36-month commitments that NVIDIA cannot displace inside one product cycle; AVGO's AI revenue FY2026 run-rate is tracking toward $25 to $30 billion. AMD (AMD) provides merchant-GPU second-source optionality. The commissioned design-house sub-layer — Alchip Technologies (3661.TW) and Global Unichip (3443.TW) — sits one tier below AVGO and MRVL. System integrators (Super Micro SMCI, Dell DELL, HPE HPE) and the neocloud cluster (CoreWeave CRWV, Nebius NBIS, Applied Digital APLD, Iris Energy IREN) occupy the most commoditised downstream layer and are the transmission channel through which GPU-collateralised debt — funded partly by hyperscaler issuance of $79 billion from MSFT and $25 billion from META in 2025 alone — routes into private-credit and insurance balance-sheet exposure.


What would change my mind

Three concrete observations would falsify the positioning. First: TSMC CoWoS capacity additions accelerate materially beyond the current +60% year-on-year 2026 expansion plan, signalling supply catching demand and compressing advanced-packaging rent durability. Second: NVIDIA prints a Q3 FY27 gross margin below 70%, ahead of the commodity-compression timeline the thesis assumes. Third: Anthropic's S-1 reveals an annualised revenue run-rate below $30 billion — breaking the $965 billion private mark and triggering a re-rate across the frontier-lab tier that pulls hyperscaler capex guidance lower and closes the Cahn gap via commitment cuts rather than revenue compounding.

What I'm watching next

The 12 June 2026 US-government directive against Anthropic's Fable 5 and Mythos 5 (https://www.anthropic.com/news) is the load-bearing new datapoint of this cycle: a domestic-model-variant restriction is categorically distinct from the BIS export-control regime applied to China since 2022. I am monitoring Commerce Department BIS docket filings for July through September 2026; if this resolves as a Section 1758 Export Control Reform Act action against a domestic commercial product, the regulatory-moat dynamic enters operational reality and the platform layer's revenue ceiling caps lower. Alongside that: NVDA Q1 FY27 10-Q (SEC EDGAR returned 403 this cycle), TSMC May and June 2026 monthly revenue, SK Hynix HBM4 mass-production milestone, and ASML Q1 2026 High-NA EUV bookings are all deferred primaries for the next cycle. The Anthropic S-1 remains the cleanest single falsification observable in the pipeline.

Surfaced Names

research_scan theme='chips-compute'; value_chain_position='upstream-enabler'; Monopoly supplier of EUV-actinic mask-blank and patterned-mask inspection systems — the second-order monopoly inside the ASML stack. Every EUV layer at every leading-edge fab depends on Lasertec inspection. Tokyo-listed; small float; durably scarce. Verified via standard semicap literature; primary IR fetch not re-sourced this cycle.

research_scan theme='agi-infrastructure'; value_chain_position='picks-and-shovels'; Ethernet / UEC back-end AI-network spine — the multi-vendor alternative to NVDA NVLink captive fabric. Beneficiary of the same hyperscaler multi-sourcing dynamic that promotes AVGO/MRVL custom silicon; sells into all hyperscalers regardless of which accelerator wins each slot.

research_scan theme='chips-compute'; value_chain_position='upstream-enabler'; Single-supplier monopoly on EUV and High-NA EUV lithography — gates every leading-edge logic node at TSMC, Samsung, and Intel. Substitution is structurally infeasible inside a decade; durable rent extraction independent of which merchant-GPU or custom-ASIC vendor wins the next architectural cycle. Picks-and-shovels analogue at the digital-upstream-monopoly layer to AA/GEV at the physical layer that C-26 names.

research_scan theme='space-economy'; value_chain_position='pure-play'; Bluebird Block 2 launch cadence remains slipped into 2027 on Blue Origin New Glenn constraint (1–3 flights/yr). HALO Europa dual-use direct-to-cell ASIC funding (Feb 2026) established. Watch: FCC SCS docket resolution and MNO partnership deployment count. Structurally novel direct-to-cell competitor to Starlink Direct-to-Cell.

research_scan theme='defense-modernization'; value_chain_position='pure-play'; Switchblade loitering-munition franchise and the post-BlueHalo combined portfolio give AVAV exposure to the attritable-autonomy layer the CSIS June 8 'wartime footing' report explicitly names ('low-cost unmanned systems'); allied procurement of US autonomy primes (e.g. Poland Shield AI V-Bat June 2026) confirms FMS-equivalent direct-sale demand expansion.

research_scan theme='agi-infrastructure'; value_chain_position='picks-and-shovels'; Custom AI ASIC design partner for Google TPU, Meta MTIA, and multiple additional hyperscaler captive-silicon programs — 18–36 month qualification cycle protects rent against merchant-GPU price competition. As Sonnet 5 promotional pricing signals platform-layer margin compression, custom-ASIC rent captured by AVGO is one layer out and structurally more durable than NVDA merchant-GPU margin under C-04/C-26 rent-migration frame.

research_scan theme='chips-compute'; value_chain_position='picks-and-shovels'; Hybrid-bonding equipment vendor — the binding equipment-side enabler of HBM4 stack assembly and the 2027+ chip-to-chip 3D integration roadmap. Single-supplier-effective for sub-10-micron hybrid bonding at production scale; ASM Pacific competes at the older bonding nodes. Picks-and-shovels of the HBM ramp picks-and-shovels.

research_scan theme='quantum-computing'; value_chain_position='upstream-enabler'; Laser and photonic components across ion-trap, neutral-atom, and photonic quantum stacks; sits alongside Toptica (private) in the modality-agnostic upstream. Also carries optical-interconnect exposure to AI-datacentre buildout — a two-theme picks-and-shovels position.

research_scan theme='chips-compute'; value_chain_position='downstream-beneficiary'; AVOID per C-08 posture. Neocloud in the Vera Rubin four-cloud production ramp (2026-07-21, alongside GCP/Azure/OCI); preferential merchant-GPU allocation tightens coupling to NVDA product-cycle success ahead of the Q3-Q4 2026 material refinancing window against the Blackstone $11.5B facility (March 2025). C-08 transmission channel to private-credit / insurance-balance-sheet exposure into C-21 default chain.

research_scan theme='agi-infrastructure'; value_chain_position='picks-and-shovels'; LLM observability extended off the installed observability footprint — moat is the pre-existing agent footprint, not model-specific IP. Structurally advantaged versus standalone LLM-eval (LangSmith, Arize, Braintrust, W&B) which now face the Anthropic-AWS-MSFT-GOOG-Glasswing cross-lab consortium eval-primitive control per 2026-06-30 Fable 5 industry framework.

research_scan theme='quantum-computing'; value_chain_position='upstream-enabler'; US-domiciled Trusted-Foundry mature-node + RF/photonic capacity at Malta NY; runs the PsiQuantum photonic-chip wafer process. The IonQ-SkyWater $1.8B (announced 2026-01-26, closing Q2-Q3 2026) revealed the strategic value of the Trusted-Foundry mature-node tier; GFS is the largest-float listed exposure in that tier. Also gallium/germanium-exposed via the 2024-2025 China MOFCOM control regime referenced in materials-supply/2026-06-13.

research_scan theme='quantum-computing'; value_chain_position='downstream-beneficiary'; Retained economic interest in Quantinuum post-IPO (residual stake percentage not disclosed at press-release grain — S-1/424B5 verification queued for next scan). Public-equity proxy for pure-play quantum hardware capital-gain optionality; Honeywell industrial cash-flow base underwrites the exposure without single-name volatility on Quantinuum.

research_scan theme='robotics-humanoids'; value_chain_position='pure-play'; Hexagon Robotics AEON is BMW's chosen humanoid platform at Leipzig (test Dec 2025, further test Apr 2026, pilot summer 2026). Distinct from Figure at BMW Spartanburg — flagship customer running multi-vendor architecture signals OEM race not yet won; bullish for platform-agnostic picks-and-shovels (TEL, BESI). Summer 2026 pilot announcement has not yet landed as of 2026-06-20; tracking for Q3 disclosure.

research_scan theme='quantum-computing'; value_chain_position='picks-and-shovels'; Only mature-cap incumbent running a publicly-committed logical-qubit roadmap (Loon 2025 -> Kookaburra 2026 -> Cockatoo 2027 -> Starling 2029 -> Blue Jay 2033) using bivariate-bicycle qLDPC codes; hyperscaler quantum cloud channel (IBM Quantum) alongside AWS/Azure/GCP. Quantum sits inside IBM's aggregate revenue rather than dominating it — a lower-risk exposure to the theme than the pure-plays. IBM Quantum blog primary source returned 403 this scan; roadmap milestones inferring on prior published cadence.

research_scan theme='quantum-computing'; value_chain_position='pure-play'; Trapped-ion pure-play with $260-270M FY26 revenue guide (raised at Q1 2026 print +755% YoY), $3.1B cash, $470M RPO; pending SkyWater Trusted-Foundry acquisition close Q2-Q3 2026 vertically integrates control-electronics supply. IonQ Clavis XG Multiplex (2026-06-17) opens a PQC-migration/QKD revenue bridge on regulated federal spend. Named-equity C-02 valuation-regime test (relevant-but-not-injected) at ~37x forward revenue multiple.

research_scan theme='defense-modernization'; value_chain_position='pure-play'; XQ-58 Valkyrie CCA-track program and MACH-TB hypersonic-target franchise are the public-market autonomy + hypersonic-test exposure; the Anduril NGC2 designation reduces KTOS's representativeness of the autonomy frontier but the hypersonic-target line (MACH-TB) is differentiated and bound by reconciliation flow.

research_scan theme='ai-capex-ceiling'; value_chain_position='pure-play'; Hyperscaler capex commitment layer with $25B 2025 debt issuance and ~$100M senior-AI-researcher comp packages signalling the C-04 capex/revenue arithmetic strain. Explicitly named in Druckenmiller Q4 2025 13F exits per C-25 spine and on the C-09 ad-revenue-on-3-5yr-lag avoid list.

research_scan theme='defense-modernization'; value_chain_position='picks-and-shovels'; Rugged compute / radar / EW module supplier at the intersection of chips-compute and defense-modernization. Beneficiary of the classified programs $12.1B line and the airborne-targeting/space-networks $4B line in the July 2026 supplemental. Sells into every prime radar and ISR system; the durability read is on Golden Dome MEO tracking-sat AMDT3 build-out.

research_scan theme='agi-infrastructure'; value_chain_position='picks-and-shovels'; Custom silicon partner for AWS Trainium/Inferentia — the second-order beneficiary of hyperscaler multi-sourcing away from NVDA. AWS Bedrock margin captured by Amazon rather than by Anthropic under the Sonnet 5 promotional-pricing configuration; MRVL sells the silicon that makes that capture possible.

research_scan theme='ai-capex-ceiling'; value_chain_position='pure-play'; Q4 FY2026 earnings scheduled 2026-07-29 (verified) — first hyperscaler capex-guide read on whether the ~16x capex/revenue gap begins to close in CY2026; direct falsification-observable candidate for criterion #26. MSFT sits on supports_avoid per C-04 posture ahead of the print.

research_scan theme='chips-compute'; value_chain_position='upstream-enabler'; US-listed proxy for the three-supplier HBM oligopoly (SK Hynix and Samsung are KS-listed). HBM3e ramping; HBM4 (12-Hi, 1024-bit) qualifying for Blackwell Ultra / Rubin in tight allocation against demand. Memory-bandwidth wall is the binding constraint for inference workloads at agent and reasoning model scale.

research_scan theme='quantum-computing'; value_chain_position='downstream-beneficiary'; Deepest disclosed PQC deployment by share of TLS handshakes — sits on the federally-mandated 7-year PQC migration cycle (NIST FIPS 203/204/205 + NSA CNSA 2.0 + OMB M-23-02 + NIST IR 8547 2035 deadline). The IonQ Clavis XG Multiplex announcement (2026-06-17) operationalises the hybrid QKD-plus-PQC architecture the federal guidance describes abstractly — extends NET's addressable surface from pure software PQC into hybrid network-layer security.

research_scan theme='agi-infrastructure'; value_chain_position='pure-play'; Now Assist rides the CMDB installed-base moat — enterprise workflow orchestration where the customer's process graph is not substitutable within one refresh cycle. More insulated than lightweight orchestrators against Anthropic Cowork forward integration, but ~60× fwd P/E prices durability that requires enterprise IT budget conditions Essay 04 §II questions.

research_scan theme='autonomous-mobility'; value_chain_position='Layer 2 (AV inference SoC — DRIVE Thor)'; DRIVE Thor Blackwell-derived is OEM-agnostic AV silicon platform. Same die family and distribution strategy as Jetson Thor for humanoids — 'Android of physical AI' extends across both themes. Durably profitable at compute-provider tier. BPA has NVDA on supports_avoid side of C-04 at cap-weighted peak (posture 41), but on AV-specific chessboard NVDA is least-exposed compute-side incumbent. Held as calibration exposure.

research_scan theme='quantum-computing'; value_chain_position='upstream-enabler'; Public-market cryogenics oligopolist (dilution refrigerators, vacuum systems, cryogenic optics) selling into every quantum modality regardless of architecture — the durable picks-and-shovels tier. Sales cycle 12-24 months, capacity constrained by specialist manufacturing; benefits from D-Wave gate-model build-out on top of existing trapped-ion, superconducting, and photonic customer footprint.

research_scan theme='defense-modernization'; value_chain_position='downstream-beneficiary'; Marine Corps Maven Smart System mandate (June 23, 2026) is the operational-deployment inflection at the tactical-app layer. Updated framing this scan: the July 9 Accenture $821M War Data Platform award confirms PLTR is not the sole DoD enterprise-data-platform winner — durable at tactical-app, capped at enterprise-integration. Wallet-share ceiling is lower than the June 25 reading implied.

research_scan theme='quantum-computing'; value_chain_position='pure-play'; Pivoted 2026-06-01 from annealing-only to dual-architecture annealing-plus-gate-model with a published 6-year fault-tolerant roadmap (17 physical qubits 2026, 49 in 2027, 181 in 2028, 10 logical 2030, 100 logical 2032). DOC CHIPS LoI signed 2026-05-21 ($100M disclosed); Microelectronics Commons Year-2 funding 2026-05-26. The gate-model pivot concedes annealing does not reach the fault-tolerant prize alone; structurally crowds the gate-model competitive set with a sixth entrant.

research_scan theme='quantum-computing'; value_chain_position='pure-play'; Trapped-ion hardware-leading pure-play; IPO priced 2026-06-03 and closed 2026-06-05 at $60/share, 28M shares sold on Nasdaq for $1.68B gross proceeds; 2026-06-21 HPE strategic collaboration extends the HPC-hybrid enterprise channel beyond BMW/BP/Synopsys/Mitsubishi Electric. Sits inside the C-02 valuation-regime test (relevant-but-not-injected) at named-equity level. Post-IPO revenue disclosure needed to establish comparable-set multiple vs IONQ FY26 $260-270M guide.

research_scan theme='robotics-humanoids'; value_chain_position='picks-and-shovels'; Sensor + connector supplier across humanoid BOMs (force-torque, tactile, position). Trades ~16x forward with sensor-content-per-unit rising as OEMs move to 25-DoF hands (1X 2026-07-09) and multi-vertical deployments (Catalyst apparel DC extends beyond auto). Under-priced picks-and-shovels edge — captures rent regardless of which OEM/model wins.

research_scan theme='chips-compute'; value_chain_position='upstream-enabler'; Foundry monopoly at N3/N3P and CoWoS-S/L advanced-packaging — the binding 2026-2028 sub-bottleneck on Blackwell/Rubin shipment cadence. Capacity at Arizona Fab 21 phase 2 (N3, 2026-2028 ramp) is the friend-shoring leg. Earns rent regardless of which accelerator-silicon vendor wins. Primary IR fetch blocked this cycle; baseline anchored in BPA Essay 04 chip-stack discussion.