The market is making the same mistake it always does. It's pricing the gold, but the pick-and-shovel supplier is trading at a discount. Cadence CEO Anirudh Devgan recently argued his company is undervalued amid the AI boom. The semiconductor design tool giant—whose software is essential to every modern chip from Nvidia's H100 to Apple's A17—is telling investors they are missing the forest for the trees. The market sees a software company. The CEO sees a leveraged tax on the entire AI infrastructure buildout. This is not a semiconductor story. It is a narrative architecture story. And in crypto, we have the exact same blind spot.
Consider the parallel. Every cycle produces a new set of 'critical enablers'—the infrastructure layer that everything else depends on but whose value is systematically underestimated by speculators focused on the flashy end-product. In 2017, it was Ethereum's smart contract platform vs. ICO tokens. In 2021, it was Layer 1s vs. DeFi protocols. Now, in 2025, the same dynamic is playing out with the 'pick and shovel' of the AI-agent economy: the blockchain middleware, oracle networks, and zero-knowledge proving systems that enable the machine-to-machine economies we keep hearing about. The market is discounting the enablers again.
=== Hook: The Cadence Signal ===
On March 12, 2025, Cadence's CEO sat down with a crypto-focused outlet—Crypto Briefing—to make a bold claim: his company's stock is undervalued relative to the AI boom. 'We are not just a tool provider,' he said. 'We are the backbone of the AI chip design itself.' The interview was not a typical earnings call. It was a narrative intervention. The CEO was signaling that the market's mental model of Cadence—as a mature software vendor with steady growth—was fundamentally wrong. The truth, he argued, is that Cadence is a 'critical enabler' whose revenue leverage to AI infrastructure spending is far higher than what the current valuation multiple implies.
Here is the data point that caught my attention: Cadence's EDA tools support the design of every major AI chip—Nvidia, AMD, Google TPU, Amazon Trainium, Microsoft Maia. The global AI infrastructure capex from the top four cloud providers is projected to reach $300 billion annually by 2027. Every dollar of that capex flows into chip design, and every chip design flow pays a tax to Cadence in the form of EDA licenses and IP royalties. The CEO's math is simple: if AI capex grows 3x, EDA spending should grow disproportionately because each new chip generation requires more design iterations, more verification, and more IP. But the market is still pricing Cadence at a 30-35x PE, while AI chip designers like Nvidia trade at 50-60x. The gap is the arbitrage.
=== Context: The Crypto Parallel ===
This is not new. In crypto, we have seen this pattern repeat. In 2020, Chainlink was the 'pick and shovel' of DeFi—every protocol needed price feeds, but LINK's valuation lagged behind the explosive growth of the composable money legos. In 2023, Ethereum's Layer 2 scaling solutions were the enabler for the next wave of dApps, but the market was fixated on the dApps themselves, not the infrastructure. The same narrative blindness is happening now with the AI-agent economy. The protocols that provide the compute, storage, and verification layers for autonomous agents—think of platforms like Bittensor, Akash, or even the underlying ZK-proof networks—are being priced as if they are speculative bets, not critical infrastructure.
But the Cadence case adds a layer of depth. The CEO's argument hinges on a structural shift in the value chain: from 'tool licensing' to 'platform ecosystem.' Cadence is not just selling software; it is embedding itself into the entire chip design workflow, from architectural exploration to sign-off. The transition is from a one-time license to a recurring subscription that captures a share of the value created by the chip itself. This is the same move that crypto infrastructure projects are making—from simple oracle services to full-stack data availability layers, from basic compute rental to autonomous agent operating systems. The market is still thinking in terms of 'software' when the reality is 'platform.'
=== Core: The Narrative Mechanism of the Undervalued Enabler ===
Why does the market systematically undervalue the pick-and-shovel? The answer lies in the narrative architecture of the speculative cycle. When a new technology narrative emerges—AI, DeFi, AI-agent economy—the hype naturally gravitates to the most visible, highest-return assets. For AI, it is Nvidia and the magic chips. For DeFi, it was UNI and the protocols that promised yield. For the AI-agent economy, it is the AI tokens that promise to disrupt everything. The market's attention is captured by the 'story' of the end product—the autonomous agent that trades, the AI that writes code, the decentralized compute that powers the metaverse. These stories are easy to tell, easy to FOMO into, and easy to price in terms of potential future revenue.
The enabler, on the other hand, is boring. Its story is about process, not product. 'We design the tools that design the chips that run the AI' is a harder sell than 'We are building the AI that will replace all jobs.' The former requires technical understanding; the latter requires emotional resonance. And that is the narrative arbitrage. The market overpays for the story and underpays for the structure. Code talks, but stories sell. The enabler's code is solid, but its story is weak. So the market discounts it.
But here is the technical insight: the enabler's value is not linear; it is exponential. Cadence's CEO pointed out that the design cost for a 2nm chip is projected to be $5-7 billion, and EDA/IP spending accounts for 25-30% of that. That means each new chip generation delivers a 2-3x increase in EDA revenue per chip, far outpacing the growth in chip count. Similarly, in crypto, as the AI-agent economy scales, the demand for middleware—oracle networks, ZK-proof systems, data availability layers—grows super-linearly because each agent interaction requires multiple verifications, data points, and governance checks. The infrastructure becomes a 'tax' on every transaction, and as the transaction volume grows, the tax revenue compounds.
=== Contrarian: But What If the Market Is Right? ===
Here is the contrarian angle that the narrative hunter must consider. The CEO's underestimation claim assumes that the market will eventually see the structural value. But what if the market is right to discount the enabler? The counter-argument is that enablers face a fundamental risk: commoditization. If the tool is essential but not proprietary, the market will eventually price it as a utility, not a growth asset. Cadence's moat—its integration with fabs, its IP library, its decades of design data—is real, but it is not impenetrable. Synopsys is a direct competitor, and Ansys's acquisition by Synopsys (though blocked by regulators) showed that the competitive landscape is shifting. The market might be pricing in the risk that the EDA duopoly becomes a race to the bottom on pricing, especially as cloud-based EDA-as-a-service lowers barriers to entry.
In crypto, the same risk exists. The 'pick and shovel' of the AI-agent economy—oracle networks, compute layers, ZK-proof systems—are all open-source or easily replicable. The differentiation is not in the technology but in the network effects and integrations. Chainlink's competitive advantage is its adoption, not its code. But if a new oracle network with better latency or cheaper fees emerges, the market can switch. The enabler's value is only as strong as its ability to maintain its position as the 'default' layer. Hype decays; utility endures. But utility can be commoditized.
The Cadence CEO's argument also ignores the geopolitical risk. EDA tools are now subject to export controls, and Cadence's China revenue (14-17% of total) is at risk. The market might be applying a 'geopolitical discount' that the CEO sees as irrational but the market sees as prudent. In crypto, the analogous risk is regulatory uncertainty: infrastructure protocols that rely on US-based servers or compliance-facing operations face similar headwinds. The market might be pricing in the possibility that the enabler's 'essential' status could become a liability if regulators decide to clamp down on the middle layer.
=== Takeaway: The Next Narrative ===
So where does this leave us? The Cadence valuation debate is a microcosm of a larger narrative pattern. The market is currently in the 'hype the end-product' phase of the AI-agent economy cycle. The attention is on the agents themselves, the tokens with AI branding, the protocols that promise to replace human labor. But the structural value is in the enabling layer—the middleware that ensures these agents can communicate, pay, and verify without trust. The narrative is about to shift. When the hype cycle matures and the market realizes that the AI-agent economy is not a single product but a stack, the infrastructure layer will reprice. Narrative is the new liquidity. Those who understand the mechanism will be positioned to front-run the repricing.
I am not saying Cadence is a buy. I am not saying Chainlink is a buy either. What I am saying is that the narrative architecture of the current cycle—the story of the 'enabler' vs. the 'end-product'—is creating a mispricing that the market will eventually correct. The question is: will you be the one who sees the structural value before the story catches up?
Code talks, but stories sell. The enabler's code is already proven. The story is only now being written.
