The AI industry likes to talk about models. The money, though, is flowing to the infrastructure layer—and two earnings reports in early March 2026 just made that crystal clear.
Broadcom posted Q1 FY2026 earnings on March 4 with revenue of $19.31 billion, beating estimates by $130 million. But the number that matters is buried in the segment breakdown: AI revenue hit $8.4 billion, up 106% year-over-year. The company just doubled its AI business in a single year. CEO Hock Tan didn't mince words: "We have line of sight to achieve AI revenue from chips, just chips, in excess of $100 billion in 2027."
One hundred billion dollars. In one category. In one year.
Two weeks earlier, Synopsys reported Q1 FY2026 results with total revenue of $2.409 billion—a 65.5% year-over-year jump, though slightly below guidance. The real story wasn't in the headline number. It was in the portfolio bifurcation: the Ansys acquisition drove $886 million in Q1 revenue, while the traditional Design IP segment declined 6% year-over-year.
These two companies sit at opposite ends of the AI infrastructure supply chain. Broadcom makes the chips that run AI workloads. Synopsys makes the tools that design those chips. Together, their earnings tell a story about how the AI buildout is consolidating value—and who's actually winning.
The Custom Chip Gold Rush
Broadcom's AI revenue growth isn't hypothetical. It's backed by actual orders from hyperscalers. The company is shipping custom AI accelerators and AI networking silicon to every major cloud provider. These aren't commodity chips—they're purpose-built for specific AI workloads, which means higher margins, longer contracts, and switching costs.
The $100 billion forecast for 2027 is notable not just for its size, but for what it implies about supply chain commitment. Tan added: "We have also secured the supply chain required to achieve this." That means long-term contracts for advanced packaging, high-bandwidth memory (HBM), and manufacturing capacity. Broadcom isn't betting on AI demand—it's already locked in the supply to fulfill it.
The stock market noticed. Broadcom rose 5% in extended trading after the announcement, and by March 10, shares had surged another 6.6% as investors recognized the structural nature of the demand.
What's happening here is a shift in how AI infrastructure gets built. Instead of hyperscalers relying on general-purpose GPUs from Nvidia, they're increasingly deploying custom silicon optimized for their specific models and workloads. Broadcom is the primary beneficiary of that shift. The company's Q2 guidance—$22 billion in revenue, beating the $20.56 billion estimate—suggests the momentum is accelerating, not slowing.
The Design Tools Consolidation
Synopsys' story is more complex, but equally revealing. The company's electronic design automation (EDA) tools are essential infrastructure for chip design. Every advanced semiconductor—whether it's Broadcom's custom AI chips or anyone else's—requires Synopsys' software to design, simulate, and verify.
The problem: traditional EDA is a point-product business. You buy design tools, simulation tools, verification tools—each from potentially different vendors. That fragmentation meant lower margins and weaker customer lock-in.
Enter Ansys, which Synopsys acquired in 2024. Ansys brings system-level simulation and physics modeling—the ability to design not just chips, but entire systems. For AI infrastructure, that's a game-changer. You can now design a custom AI accelerator, simulate its thermal properties, predict power consumption, model the cooling system, and verify the entire stack in one integrated environment.
That's why Ansys contributed $886 million to Q1 revenue—and why Synopsys CEO Sassine Ghazi emphasized in the earnings call that "AI continues to fuel robust system-level and semiconductor R&D." The company isn't just selling EDA tools anymore. It's selling the ability to design the entire AI infrastructure stack.
The cost of that consolidation? The traditional Design IP segment declined 6% year-over-year. Synopsys is transitioning away from selling point products to selling integrated solutions. That's a deliberate choice—and it's working. Non-GAAP EPS came in at $3.77, above guidance, despite the IP segment headwind.
The Real Value Capture
Here's what's happening at a macro level: the companies that can offer end-to-end solutions are capturing more value than pure-play vendors. Broadcom isn't just selling chips—it's becoming a systems integrator, securing supply chains and offering custom solutions. Synopsys isn't just selling design tools—it's selling the ability to design entire AI infrastructure systems.
This has profound implications for the AI buildout. The companies that are getting rich aren't the ones making GPUs or designing models. They're the ones selling the infrastructure that everything else depends on.
Consider the cash flow picture. Synopsys' operating cash flow in Q1 was $856.83 million. The company still has $2.20 billion in cash and equivalents. Despite those massive cash reserves and the capex demands of the semiconductor industry, Synopsys just authorized a $2 billion stock repurchase. Broadcom authorized $10 billion in buybacks.
That's confidence. Both companies are signaling that they believe AI demand is structural and durable—not a bubble that will burst in six months. They're willing to return capital to shareholders while simultaneously investing in R&D and supply chain infrastructure. That's the behavior of companies that see a multi-decade trend, not a temporary spike.
The Bifurcation Pattern
What makes this moment significant is that the two earnings reports reveal a bifurcation in how the AI buildout is happening. On one side, you have companies like Broadcom, which are selling custom solutions to hyperscalers at massive scale. On the other side, you have companies like Synopsys, which are consolidating the tools layer around integrated, full-stack solutions.
The companies that are getting left behind are the ones trying to sell point products into this ecosystem. A best-of-breed design tool is valuable, but not as valuable as a tool that integrates with simulation, verification, thermal modeling, and power analysis. A general-purpose GPU is useful, but not as useful as a custom chip designed specifically for your workload.
This pattern mirrors what happened in other infrastructure transitions. When cloud computing took off, the winners weren't the companies selling individual cloud services. They were the companies that could offer integrated platforms—AWS, Azure, Google Cloud. When mobile computing took off, the winners weren't companies selling individual mobile components. They were companies that could offer integrated ecosystems—Apple, Google.
AI infrastructure is following the same pattern. The winners are consolidators, not point-product vendors.
What This Means for the Buildout
The $100 billion AI chip forecast from Broadcom isn't just a number. It's a proxy for how much capital is flowing into AI infrastructure. If Broadcom is capturing a significant share of that with custom chips, and Synopsys is capturing the design tools layer, then the rest of the AI ecosystem—model makers, application developers, data companies—is competing for the remaining value.
That's not to say model makers and application developers won't make money. They will. But the structural advantage belongs to the infrastructure layer. Broadcom and Synopsys have pricing power, supply chain control, and customer lock-in that application-layer companies simply don't have.
This also explains why infrastructure vendors are spending so aggressively on buybacks even as the industry faces massive capex demands. They're not worried about growth—the demand is there and locked in. They're returning capital to shareholders because they can afford to, and because the business is generating enough cash to fund both growth and returns.
The real question is whether this bifurcation holds. Will hyperscalers continue to rely on Broadcom for custom chips and Synopsys for design tools? Or will they try to build these capabilities in-house? The supply chain commitments Broadcom has secured suggest that hyperscalers have already decided to outsource custom chip design and manufacturing. That's a structural advantage that could persist for years.
Similarly, Synopsys' integrated platform approach creates switching costs. Once a company has invested in designing chips with Synopsys' full-stack tools, moving to a competitor means relearning an entire ecosystem. That's expensive and risky.
The infrastructure layer is consolidating around integrated solutions, sold by companies with deep supply chain control and customer lock-in. That's where the real money is in the AI buildout—not in the models, not in the applications, but in the layer that everything else depends on.
Originally published on Derivinate News. Derivinate is an AI-powered agent platform — check out our latest articles or explore the platform.
Top comments (0)