Synopsys Crashed 36% in a Day. NVIDIA Answered With $2 Billion

The Worst Day Since 1992

On September 10, 2025, Synopsys - the company whose software has touched essentially every advanced chip ever made - fell 35.8% in a single session, from $604.37 to $387.78. It was the worst day in its 33 years as a public company.

Less than three months later, NVIDIA handed that same company $2 billion.

Then, this month, a financial disclosure showed a brokerage account belonging to Donald Trump had quietly bought in too.

We used Barebone AI to rebuild the whole sequence from primary documents - three quarters of earnings releases, the latest 10-Q, the $10 billion bond paperwork, the December stock purchase agreement, and the CNBC transcript where Jensen Huang explained what he was doing. The arc, in numbers: a -36% crash, a $2 billion investment at $414.79 per share, a new low of $380.47 in March, and a +40% rally into late May. Synopsys closed today at $534.56 - up 29% on NVIDIA's entry price, still 17% below its July 2025 high.

This is the story of why the company that dominates AI computing bought the dip in a stock most people have never heard of - and what the filings say about whether it was right.

The Software Where Chips Are Born

NVIDIA's Blackwell GPU packs 208 billion transistors. No human lays those out by hand. Every advanced chip on Earth - NVIDIA's, AMD's, Qualcomm's - is designed, simulated, and verified inside a category of software called EDA, electronic design automation: the environment where a chip exists as code long before a foundry ever etches it into silicon.

The economics of that software are brutal in the best way. You either find the flaw in simulation, or you find it after committing billions of dollars at the foundry. So nobody churns. Roughly 80% of Synopsys's revenue comes from long-term, time-based contracts, and the backlog - revenue already under contract but not yet recognized - stood at $11.3 billion as of January 31, 2026. That is roughly fourteen months of revenue signed before the year even starts.

One folk statistic needs correcting, though. Synopsys is often described as controlling "nearly half" the chip-design software market. The real number is closer to a third - about 31%, with Cadence at roughly 30% and Siemens EDA near 13%. The honest framing isn't a monopoly; it's a duopoly. Synopsys and Cadence together control about 60% of an industry the entire semiconductor business cannot function without.

And in 2025, Synopsys did something to break out of that duopoly box: it paid $34.9 billion for Ansys, the simulation company whose software tests how physical things behave - heat, vibration, electromagnetics, fluid dynamics - in everything from jet engines to pacemakers. The deal closed on July 17, 2025. Design the chip, simulate the system it lives in, verify the whole product: one platform.

Eight weeks later, the stock fell off a cliff.

Anatomy of a 36% Crash

On September 9, 2025, Synopsys reported its fiscal third quarter. Here is what the market expected against what it got:

Metric	Expected	Delivered
Revenue	~$1.77B	$1.74B
Non-GAAP EPS	~$3.74	$3.39
Design IP revenue	growth	-8% to $428M
FY2025 EPS guidance	$15.11 - 15.19	cut to $12.76 - 12.80

The wound was concentrated in one place: the Design IP business, where Synopsys sells pre-built circuit blocks that chipmakers license instead of designing from scratch. Roughly $140 million of expected IP revenue simply didn't show up - a casualty of U.S. export restrictions choking off chip-design starts in China, plus problems at a major foundry customer the company declined to name.

A guidance cut of that size, in a company priced for predictability, broke the spell. Bank of America downgraded the stock from Buy to Underperform overnight, slashing its target from $625 to $525. By the September 10 close, Synopsys had lost nearly 36% in a day and sat about 40% below its July high of $645.35.

The chart shows what happened next, and it's the part most retellings skip: the stock did not V-bound. It bounced, faded, and by late November was sitting in the high $300s - the market had effectively decided the safest software business in semiconductors was no longer safe.

That was the price NVIDIA walked into.

The $2 Billion Answer

On December 1, 2025, NVIDIA and Synopsys announced a strategic partnership, anchored by NVIDIA buying $2 billion of newly issued Synopsys stock at $414.79 per share - roughly the market price, about 4.8 million shares. Note the structure: a private placement. The $2 billion went onto Synopsys's balance sheet, not to a selling shareholder.

Huang's explanation, on CNBC that morning, is worth quoting precisely:

"Synopsys is pivoting across their company to transform their software and all the tools that the industry, the world's been using for some 35 years to be GPU accelerated on NVIDIA."

And the history behind it: "NVIDIA was built on a foundation of design tools from Synopsys and of course many other companies like Cadence and Siemens and Dassault."

The investment thesis has three steps, and each one compounds the next:

Step one: acceleration. Chip design and physics simulation are among the most compute-intensive workloads in existence, and they still mostly run on CPUs. Moving them to CUDA, Synopsys CEO Sassine Ghazi said in the same interview, means "taking a workload that may run for two, three weeks and reducing it to hours."

Step two: agents. The partnership explicitly integrates Synopsys's AgentEngineer technology with NVIDIA's agentic AI stack - autonomous design agents working inside the tools. Humans can't design chips fast enough for what AI demands; the bet is that AI agents, running inside EDA software, will do the designing. As Ghazi put it: "You need to re-engineer engineering."

Step three: expansion. Huang's own math on CNBC: companies like NVIDIA spend "a few hundred million dollars" on design tools but "billions of dollars in prototyping." Move prototyping into simulation - digital twins of chips, cars, factories - and, in his words, "all of a sudden, the market opportunity increases by a factor of 10 to 100."

That last number is almost certainly aspirational. But notice what NVIDIA did versus what it said: it didn't just endorse the story, it bought equity at the bottom of the range - and Synopsys popped only +4.9% on the news. As endorsements go, the market shrugged.

The President's Trust Showed Up Too

In mid-May 2026, a periodic transaction report - the OGE Form 278-T that senior officials file - disclosed that a Trump brokerage account had opened a position in Synopsys during the first quarter of 2026, valued between $1 million and $5 million.

Three caveats before anyone gets excited. First, the disclosure is a range, not a number. Second, Trump's assets sit in a trust managed by his children, and the filings don't say who actually placed the trades. Third - and most telling - the same filing shows purchases in the same band of Cadence, Broadcom, Texas Instruments, and NVIDIA. That's not a Synopsys thesis. That's an AI-supply-chain basket, and it happens to include both halves of the EDA duopoly.

The honest read: interesting color, weak signal. The $2 billion from the company that actually uses the tools is the bet worth studying.

What the Growth Is Actually Made Of

Here is where the filings get more interesting than the narrative.

Fiscal 2025 revenue came in at a record $7.05 billion, up 15% - but Ansys, consolidated from mid-July, contributed $756.6 million of it. Strip that out and core Synopsys grew about 2.8%. For fiscal 2026, guidance is $9.61 billion at the midpoint, of which $2.9 billion is expected from Ansys. The core, again, is guided to roughly $6.7 billion - mid-single-digit growth, modestly understated by about $110 million of divested businesses in the comparison.

So the "fortress" framing is half right. The revenue quality is genuinely fortress-grade: ~80% recurring, an $11.3 billion backlog of which 47% (excluding flexible-spending commitments) is expected to convert to revenue within twelve months. But the growth engine was stalling in fiscal 2025 - China restrictions, the IP stumble, customers digesting. At today's price you are not buying current hypergrowth. You are buying the Ansys integration plus an option on the agent thesis.

That distinction is the whole investment.

The Two Ledgers

Now the uncomfortable part. Synopsys currently reports two very different versions of its own profitability.

In the first quarter of fiscal 2026, GAAP net income was $65 million - $0.34 per share - against non-GAAP net income of $718.5 million, or $3.77 per share. The gap is the cost of the Ansys deal made visible: amortization of acquired intangibles, integration expenses, and interest on the $10.0 billion of bonds Synopsys issued in March 2025 to fund it - at the time, the largest U.S. investment-grade debt deal of the year outside the financial sector, with coupons running from 4.55% to 5.70% and maturities out to 2055. Total debt stood at $10.04 billion as of January 31, 2026.

Which earnings you believe determines what you think you're paying:

Lens	EPS	Implied P/E at $534.56
Trailing GAAP (last four quarters)	$6.52	~82x
FY2026 GAAP guidance (midpoint)	$2.42	~221x
FY2026 non-GAAP guidance (midpoint)	$14.42	~37x

The bulls will tell you the GAAP numbers are accounting noise - amortization isn't cash, and guided free cash flow of roughly $1.9 billion this year is real. Fair. But 37x forward earnings, on the friendliest ledger, for a business whose core is growing mid-single digits, is a multiple that assumes the agent thesis starts paying.

And the market has already demonstrated what happens when it doubts that. In late March 2026, as investors dumped software stocks on the fear that AI agents would eat seat-based software rather than feed it, Synopsys closed at $380.47 - below its crash-day close, and 8% under NVIDIA's entry price. The $2 billion endorsement was not a floor. NVIDIA is a strategic partner with a $2 billion rounding error, not a price-sensitive investor defending a level.

There's also a policy overhang the company itself flags: its fiscal 2026 targets explicitly "assume no further changes to export control restrictions." The September crash already showed what one export-policy shift does to the IP line. That risk has not gone anywhere.

What This Means

Synopsys at $534.56 is a referendum on one question: when AI agents start doing engineering, does the toolmaker capture the value - or get commoditized by it? Huang has placed $2 billion on "capture." The March lows show plenty of money betting "commoditized." A year of price action - crash, endorsement, lower low, 40% rally - is the market changing its mind in public, twice.

The signals worth watching from here, in order of importance:

1. Core growth ex-Ansys. The agent thesis is real when core Synopsys revenue re-accelerates from ~3% toward double digits. Until then it's a story.
2. The two ledgers converging. Integration costs ending and the GAAP - non-GAAP gap narrowing is the cleanest tell that the $34.9 billion deal is being digested rather than endured.
3. The China/IP line. Roughly $140 million vanished on a policy change. Any thaw - or any new restriction - moves this stock disproportionately.
4. Evidence agents expand the market. CUDA-accelerated product uptake, AgentEngineer adoption, anything that turns Huang's "factor of 10 to 100" from keynote math into bookings.

The next data point isn't far away: Synopsys reports fiscal second-quarter earnings tomorrow, May 27, with guidance calling for $2.225 - 2.275 billion in revenue. Wall Street heads into the print constructive - Citi carries a $600 target, Stifel $550, and even Bank of America, the firm that downgraded the stock the morning after the crash, is back at Buy.

The crash told you what the business was worth the day the story broke. The $2 billion told you what the story is worth to the one buyer capable of making it true. Everything since is the market arguing about the gap - and tomorrow afternoon, it marks to market again.

---

Data: Barebone | Sources: Synopsys Q3 FY2025, Q4 FY2025 and Q1 FY2026 earnings releases and Q1 FY2026 10-Q (SEC EDGAR), NVIDIA - Synopsys partnership announcement (Dec 1, 2025), CNBC Squawk on the Street transcript (Dec 1, 2025), OGE Form 278-T disclosures (May 2026) | Prices as of May 26, 2026 close