Synopsys promised 50% margins, then agreed to share the new revenue with OpenAI
Hours after promising operating margins approaching 50% by fiscal 2030 and $1B of buybacks at its Investor Day, Synopsys unveiled GPT-Synopsys: a frontier model trained on its tools, running on OpenAI's infrastructure, sold as one bundle with the revenue shared. The size of that share is the one number nobody disclosed.
Vincent Jiang · 2 min read
OpenAI pays to learn the tools, then takes a share of the work
GPT-Synopsys, announced 30 September, is a multi-year agreement to build a frontier model that operates Synopsys' design software like an engineer: agents run the tools, read the results, and iterate toward closure 1. The money runs both ways. OpenAI pays Synopsys a training subscription while the model learns the tools, and once customers use the product, the two split revenue by how much the model improves a design 2. The offering bundles compute, model and licenses into one bill on OpenAI-hosted infrastructure, priced by subscription, consumption and outcome 13.
The 50% promise, made hours earlier
The deal landed at the same event where Synopsys guided fiscal 2027 revenue to $11.1B to $11.2B, about 15% growth against the 11.19% analysts expected, with non-GAAP operating margin at 44% next year and approaching 50% by fiscal 2030 236. Management added $1B of buybacks and a pledge to return up to half of free cash flow 3. Amazon anchored the IP leg with a multi-year agreement worth more than $1B in license fees across generations of Graviton, Trainium and Nitro, before customization fees and royalties 34.
Guided fiscal 2027 revenue of $11.15B sits 58% above fiscal 2025
- Revenue
- Estimate
Data
| Revenue | |
|---|---|
| FY2021 | $4.2B |
| FY2022 | $4.62B |
| FY2023 | $5.32B |
| FY2024 | $6.13B |
| FY2025 | $7.05B |
| FY2026 | — |
| FY2027 (estimate) | $11.15B ($11.1–11.2B) |
Camp one: the toll widens
Ghazi told Reuters the deal "will not be cannibalizing our business," and that sign-off validation stays essential: "The model needs these guardrails in order to check the physics" 2. BofA raised its target to $600 from $500 and lifted its fiscal 2027 and 2028 estimates by 7% and 16% 6. KeyBanc went to $605 6.
The 48-hour counterargument
Camp two reads the same press release as a handover: its stated goal is frontier models becoming "experts in using EDA tools," hosted on their owner's infrastructure 1. In July, Moonshot's Kimi K3 designed a working chip in 48 hours on open-source tools alone, and Cadence and Synopsys each fell roughly 9% 7. In the same July coverage, Bloomberg Intelligence's Niraj Patel wrote that AI as the primary interface to EDA "may shift software value up the stack" 7.
The number nobody disclosed
Neither camp can price the share, because Synopsys never disclosed it. The margin promise carries a number; the leakage carries none. The stock closed 30 September up 4.78% but started Thursday down more than 9% for the year 6; by mid-morning it traded near $485, up more than 11%, still below both raised targets 68. The tell arrives with the first outcome-priced contract, the first GPT-Synopsys revenue line, and Amazon royalties, which Synopsys does not expect in fiscal 2027 3.
Both raised targets sit above the stock even after its 11% rebound
Data
| Value | |
|---|---|
| KeyBanc target | $605 |
| BofA target | $600 |
| SNPS mid-morning | $485 |
Deepdive
AI-generated from this story and its cited sources. Not investment advice.
