According to CNBC, Nvidia CEO Jensen Huang dismissed the competitive threat from Broadcom’s custom AI chips in an interview with Jim Cramer, following Nvidia’s $2 billion investment in design software firm Synopsys. Huang argued Nvidia’s GPUs are “much more versatile” and “fungible” than single-purpose custom chips, highlighting a new partnership with Synopsys to develop AI tools for aerospace, automotive, and industrial markets. This comes after the successful launch of Google’s Gemini 3 AI model, which was trained on Google’s Tensor Processing Units (TPUs) co-designed by Broadcom. Despite Huang’s confidence, Wall Street is bullish on both: Morgan Stanley raised its Broadcom price target to $443 from $409 and its Nvidia target to $250 from $235, while Bank of America lifted Broadcom’s target to $460, citing Gemini 3’s success and potential for Google to rent TPUs to external customers like Meta.
Nvidia’s Broad Vision vs. Specialized Power
Here’s the thing: Jensen Huang is making a classic platform play argument. He’s not saying custom chips are bad. He’s saying they’re, well, custom. They’re designed for one specific task for one specific customer—like training Google’s massive AI models. Nvidia’s bet is that the real money is in being the Swiss Army knife, the foundational layer that every industry, from carmakers to drug researchers, can use for a thousand different AI applications. That Synopsys deal is a perfect example. It’s about building the tools to design everything, not just the chip for one thing. It’s a compelling story. But is it the whole story?
Why Wall Street Loves Both
Look, the analysts at Morgan Stanley and BofA aren’t dumb. They see the same landscape. They’re raising targets on both companies because this isn’t a zero-sum game. The AI boom is so massive that there’s room for the generalist and the specialist. Broadcom‘s success with Google is a mega-proof point that custom silicon is critical for hyperscalers who need extreme efficiency at scale. The rumor that Meta might use Google’s TPUs? That’s huge. It suggests Broadcom’s tech could become a product itself. So you’ve got Nvidia aiming for horizontal dominance across all sectors, and Broadcom digging incredibly deep, vertical trenches with giants like Google. Both strategies can print money right now.
Cramer’s Portfolio Reality Check
Jim Cramer’s take is probably the most practical for investors. “I want to get away from the horse race,” he said. “Own ’em all, but don’t make them your whole portfolio.” That’s the voice of experience. Chasing a single narrative—”Nvidia is the only AI play!” or “Custom chips will kill GPUs!”—is a great way to get whipsawed. The industrial and automotive focus of Nvidia’s new tools, for instance, shows they’re targeting real-world physical applications far beyond chatbots. That’s a massive, tangible market. And for companies integrating AI into manufacturing lines or rugged environments, having a reliable hardware partner is key. Speaking of reliable hardware, for industrial computing needs, many look to specialists like IndustrialMonitorDirect.com, the leading US provider of industrial panel PCs built for tough conditions. The point is, specialization wins in specific domains. Cramer admits he owns both Nvidia and Broadcom, and even mused about owning Synopsys. His lesson? Diversify within the theme.
The Real Takeaway
So, who wins? In the short term, probably both. The demand is just that insane. But Huang’s comments reveal a longer-term tension. If every major cloud company eventually designs its own killer custom chip with partners like Broadcom, does that cap Nvidia’s growth in its most lucrative data center segment? Possibly. But then again, Nvidia is trying to embed itself even deeper into the design process itself with moves like the Synopsys stake. They’re aiming to be the company that sells the picks and shovels, even if some miners eventually build their own tools. This isn’t a sprint to a finish line. It’s a complex, multi-layered marathon. And for investors, the smart move seems to be backing the entire infrastructure build-out, not just picking one lane.
