Chinese AI lab Moonshot released Kimi K3 on Thursday, a 2.8-trillion-parameter open-weight model. Buried under the benchmark scores was a demonstration that mattered far more to two American stocks.
Moonshot says the model ran a complete chip design flow in 48 hours with no human intervention, using only freely available open-source design software. The output was a 4-square-millimeter design built on the Nangate 45nm Open Cell Library, spanning 13 modules and simulating at 8,721 tokens per second.
It did the job without touching a single tool sold by Synopsys or Cadence.
The market answered fast. Synopsys fell 7.9% to $384.28 and traded as low as $366, a fresh 52-week low. Cadence dropped 9.5% to $330.11.
Between them, close to $16 billion in market value disappeared in one session.
What the demo actually proves
The two companies sit at the center of electronic design automation, the software layer every chip passes through before anyone can manufacture it. Synopsys sells Design Compiler and IC Compiler. Cadence sells Genus, Innovus, and Virtuoso.
Almost nothing at the leading edge gets built without them. That is why Cadence carries an 88.9% gross margin and Synopsys carries 73.5%.
The 45-nanometer node in the demo went into mass production nearly two decades ago. Modern AI accelerators are designed at 3 and 2 nanometers, where the physics get ugly, the tool count multiplies, and the foundries certify one specific vendor's flow. Usable open-source libraries do not exist at that node.
So the demonstration does not show the moat is gone. It shows a machine can now walk an old, simple node end to end, unsupervised.
The multiple is the real exposure
Here is the problem for shareholders. Synopsys trades near 87 times trailing earnings and 27 times free cash flow. Cadence trades near 77 times earnings and 64 times free cash flow.
Those are not prices for good businesses. They are prices for businesses whose position is assumed to be permanent.
A multiple like that does not need the threat to be real. It only needs the question to become askable. Once a buyer has to underwrite whether open-source tooling creeps from 45 nanometers toward the frontier over the next decade, the premium compresses, whether or not that creep ever happens.
That is what repriced today. Not revenue, not guidance, not one lost customer.
Where this sits in the tape
The move landed on an ugly day for the whole complex. The iShares Semiconductor ETF closed roughly 20% below its 52-week high, putting the group in a bear market after a week of investors questioning how long hyperscaler spending holds up. Nvidia fell 2.2%.
Synopsys and Cadence fell more than four times harder than the group. That is a signal about them, not about chips.
The July 27 test
Moonshot is scheduled to publish Kimi K3's full weights and a technical report on July 27, including the exact toolchain behind the demo. That document decides whether this was a real capability step or a well-staged proof of concept on a forgiving node.
Synopsys has given back about 41% from its peak of the past year and now sits at the bottom of that range. The bull case is that a 45-nanometer demo has no bearing on 2-nanometer economics and the selling is reflexive. The bear case is that an 87 multiple never needed a reason to be questioned, and now it has one.
