Debrief: Mtng W Andrzej Strojwas Pdf And Bliss Perry And Dustin Ross — 2026-05-22
Summary
Andrzej Strojwas, co-founder and long-tenured CMU professor at PDF Solutions, walked Bliss and Dustin through PDF’s 900-person semiconductor manufacturing intelligence platform — covering Exensio, FDC systems, the Symmetrics and Securewise acquisitions, and the SAP Sapiens Manufacturing Hub partnership. The conversation surfaced deep structural context on data sharing limitations across the semiconductor supply chain, PDF’s proprietary data model, and the emerging complexity of advanced packaging traceability. Andrzej offered to introduce the founders to PDF CEO John Kibarian and product owner Kimon, and flagged upcoming Stanford-area conferences as in-person meeting opportunities.
Key Themes
1. PDF Solutions Platform & Capabilities Andrzej framed PDF as a full-stack manufacturing intelligence company, not merely a software vendor. The Exensio platform provides ML/AI-based manufacturing analytics and serves as a dashboard for leading fabless companies monitoring product performance across foundries. FDC (fault detection and classification) is deployed across all TSMC fabs — ‘every fab, whether it’s legacy or whether this is the newest, greatest being built uses our system.’ The Symmetrics acquisition brought equipment connectivity to 300+ clients; Securewise added remote engineer access to equipment, with ASML as the flagship client. This positions PDF as infrastructure-layer software embedded deeply into the operational workflows of the most critical nodes in the semiconductor supply chain — not a compliance or visibility overlay, but a manufacturing process control system.
2. Data Sharing Challenges Across the Supply Chain Andrzej was unambiguous: ‘Very little data sharing. And I think that’s a big problem in our industry.’ The exchange between TSMC and fabless companies is largely limited to PDKs and wafer acceptance tests. Equipment makers won’t share equipment secrets; fabs won’t share process recipes. ‘The equipment guys don’t want to reveal all the secrets in the equipment and the manufacturing guys don’t even give the equipment guys the process recipes.’ This structural opacity has intensified as TSMC’s dominance concentrates crown-jewel risk. PDF itself is constrained — despite owning characterization vehicle data aggregated across customers, a single data leak ‘would probably mean the end of PDF.’ The data sharing problem is not a technical gap but a trust and competitive incentive gap, and it predates the current geopolitical context.
3. PDF’s Proprietary Data Assets and Business Model PDF’s characterization vehicles — test structures placed on silicon across leading-edge process ramps — generate thousands to tens of thousands of high-density data points per wafer. Critically, PDF owns this data. They aggregate it across customers to build models and benchmarking services, but never sell data on the open market: ‘Everything we do is really part of engagement.’ The business model has historically included a yield-based fee structure (‘geisha model’) where PDF was paid on a per-wafer basis when yield targets were met. With IDM consolidation, that model has contracted, and the platform/subscription model is now primary. The LLM-based tools being developed on top of this proprietary data — without revealing source customers — represent PDF’s emerging AI data moat.
4. SAP Partnership and Sapiens Manufacturing Hub SAP dominates ~95% of semiconductor financial transactions but has been ‘really blind in terms of manufacturing data.’ PDF’s Sapiens Manufacturing Hub bridges the top-floor/shop-floor gap. This is relevant because SAP’s transaction dominance means any compliance, traceability, or financial instrument layer built on top of semiconductor supply chain data would need to integrate with or route through SAP infrastructure. Andrzej recommended the Stanford team study SAP HANA and the Sapiens Manufacturing Hub specifically, and offered to connect them with SAP contacts.
5. Advanced Packaging and Emerging Data Integration Gaps The Intel Ponte Vecchio example — 47 dice from different manufacturers and nodes assembled into a single package — illustrated the new traceability complexity in advanced packaging. ‘Deciding which dice to package vs. reject’ is a novel optimization problem with no clean data infrastructure solution today. PDF is addressing this with blockchain-based traceability and DAX (data exchange format) software enabling secure data transfer from OSATs back to foundries. This is a growth frontier for PDF and a signal that the compliance and traceability problem is migrating upstack into the packaging layer.
6. Stanford Research Framing: Visibility and Financial Instruments Bliss and Dustin framed their research around two structural gaps: (1) supply chain visibility tools inadequate for the industry’s strategic importance, and (2) the absence of financial hedging instruments — no price indices, derivatives, or parametric insurance products analogous to oil markets. Andrzej did not push back on either framing, though he did not validate them from a buyer-urgency perspective. The founders are one month into the research, and this conversation was more orientation than validation.
Notable Quotations
“Very little data sharing. And I think that’s a big problem in our industry. Right. So even between TSMC and the leading fablets, there is not a lot of data sharing.” — Andrzej Strojwas. Context: Direct characterization of the structural opacity that defines semiconductor supply chain data flows; validates the visibility gap thesis but reframes it as a trust/incentive problem, not a technical one.
“A single leakage would probably mean the end of PDF.” — Andrzej Strojwas. Context: Explains why even a data-rich intermediary like PDF cannot share or sell its proprietary manufacturing data — the competitive sensitivity is existential, not merely legal or contractual.
“SAP is totally dominating all the transactions happening in semiconductor industry to the tune of like probably 95% — except that they have been really focusing much more on what’s known as the top floor.” — Andrzej Strojwas. Context: Frames SAP as the dominant financial transaction layer with a structural blind spot on manufacturing data — the gap PDF’s Sapiens Manufacturing Hub is designed to fill, and a constraint any compliance or financial product in this space must navigate.
Themes & Contradictions
This conversation is largely additive to the corpus rather than contradictory, but it surfaces several important tensions with prior context.
The most significant confirmation is of the data sharing problem. The Gemini and Claude venture selection memos both treat semiconductor supply chain opacity as a given structural condition enabling a compliance wedge. Andrzej’s account — ‘very little data sharing… equipment guys don’t want to reveal all the secrets… manufacturing guys don’t even give the equipment guys the process recipes’ — provides the most granular, operator-level confirmation of that opacity yet seen in the corpus. This is not a regulatory gap; it is a competitive trust deficit that has persisted for decades and predates UFLPA or BIS enforcement.
However, there is a soft contradiction with the compliance wedge thesis as currently framed. The Claude and Gemini memos identify export compliance (ECCN classification, BIS enforcement) as the highest-urgency entry point, with a clear buyer archetype (VP Export Compliance). Andrzej’s framing of data problems centers on manufacturing process control and yield management — not regulatory compliance. PDF’s platform is used to manage manufacturing quality, not to manage export or forced labor risk. This suggests the ‘shop floor’ data problem and the ‘compliance’ data problem may be different buyer problems requiring different solutions, even if they share structural causes.
The SAP finding is notable in context of the P0004 internal session scrap, which flagged that no VP Export Compliance has been interviewed. SAP’s 95% transaction dominance means any compliance workflow touching financial transactions in semiconductors likely runs through SAP infrastructure. If a compliance product needs to integrate with how money and orders flow, that is SAP territory — and PDF’s partnership with SAP is a potential channel or constraint worth understanding before building.
The Richard Dasher meeting (November 2025) covered US competitiveness and critical materials dependency at a macro level but did not produce operator-level insight into supply chain data infrastructure. This Andrzej conversation fills that gap substantially. The geopolitical framing (semiconductors as new oil) was present in both conversations, but Andrzej grounds it in operational reality — TSMC dominance, Intel/Samsung struggles — rather than policy abstraction.
Business Problems & Painpoints
Andrzej articulated pain on behalf of the industry rather than as a personal buyer, but the pains named are structurally real and operationally grounded.
The deepest pain is data opacity across supply chain boundaries. The exchange between foundry and fabless is limited to PDKs and wafer acceptance tests — a decades-old, low-bandwidth interface for a relationship that determines yield, quality, and ultimately product revenue. Equipment makers and fabs cannot share process recipes or equipment parameters without risking exposure of crown jewels. This creates debugging and optimization problems that neither party can solve alone, and it creates yield risk that is currently absorbed through conservative design margins rather than shared data.
Advanced packaging introduces a new tier of this pain. When 47 dice from different nodes and manufacturers are assembled into a single package, the question of which dice to accept or reject requires data that crosses multiple organizational boundaries — foundries, OSATs, fabless customers. No clean data infrastructure exists for this. PDF is building toward it (DAX, blockchain traceability), but the market is ahead of the tooling.
The SAP blind spot is a structural workflow pain at scale. 95% of semiconductor transactions flow through SAP, but SAP has no manufacturing data. Finance teams cannot see shop-floor reality; shop-floor teams cannot connect yield and process data to financial outcomes. PDF’s Sapiens Manufacturing Hub is the current answer, but Andrzej acknowledged ‘a long distance to go in terms of penetrating the market.’
Andrzej did not express pain around export compliance, forced labor, or regulatory burden — the pain categories most central to the founders’ current thesis. This is a signal worth noting: a deeply embedded operator in the manufacturing data layer did not spontaneously name compliance as a burning problem, which may reflect his functional vantage point (manufacturing intelligence vs. legal/regulatory) rather than the absence of that pain in the market.
Emotional Signals
Andrzej came across as generous, intellectually engaged, and genuinely curious about what the founders were building. He interrupted his own monologue multiple times to redirect attention to Bliss and Dustin, framing his questions about their backgrounds as necessary for calibrating how he communicated — a sign of a practiced educator and communicator. His energy was highest when explaining technical architecture (FDC systems, characterization vehicles, the multi-tenant data question) — these topics triggered the longest and most detailed responses. He was notably careful and measured when discussing data sharing risks, pausing to emphasize that even a single data leakage ‘would probably mean the end of PDF.’ That statement carried weight and was not delivered casually. There was warmth in the personal exchanges — the Polish connection with Dustin, curiosity about backgrounds — which suggests genuine rapport has been established. He offered introductions freely, which signals confidence in the Stanford team despite their early-stage knowledge.
For Founders
-
Andrzej did not name export compliance, forced labor, or regulatory burden as pain points — even though he operates at the center of the data infrastructure problem the compliance thesis depends on. Does that silence reflect his functional vantage point (manufacturing intelligence vs. legal/regulatory), or does it suggest the compliance pain lives in a different organizational layer that PDF simply doesn’t touch?
-
The SAP finding — 95% of semiconductor transactions flowing through a platform that is ‘blind’ to manufacturing data — is a structural fact that hasn’t appeared in prior interviews. If any compliance, traceability, or financial product needs to integrate with how money and orders flow in semiconductors, does that make SAP a required partner, a distribution channel, or a competitive ceiling?
-
When Kibarian and Kimon introductions happen, what is the specific question Bliss and Dustin want answered — and does it test the compliance wedge thesis, the financial instruments thesis, or something this conversation has surfaced that wasn’t in the prior framing?