Sometimes in tech, being early isn't just an advantage—it's everything. That's the story JPMorgan analyst Harlan Sur is telling about Broadcom Inc. (AVGO) and its custom artificial intelligence chips, where an 18-month head start is translating into billions of dollars in orders and a dominant position that looks increasingly hard to challenge.

When Your Competition Is Your Customer

Sur maintained his Overweight rating on Broadcom with a $475 price target, arguing that the company's long lead in custom AI chips is accelerating its AI Application-Specific Integrated Circuit (ASIC) business into fiscal 2026 and 2027. The main catalyst? Alphabet Inc. (GOOGL)'s Google, which is both trying to design its own chips and relying heavily on Broadcom to actually make them work at scale.

Here's where it gets interesting. Broadcom has an 18-month-plus advantage over internal customer-owned programs where customers handle roughly 90% of the design work themselves. Google's 3nm Zebrafish program—one of those internal efforts—has stumbled through repeated delays and might not produce first silicon until late this quarter or early next quarter. Meanwhile, Broadcom's next-generation 3nm Tensor Processing Unit (TPU), code-named Sunfish, was already evaluated and qualified by Google earlier, giving Broadcom a substantial timing edge.

The Math That Matters

Google is finding more success deploying its TPU platform for external AI workloads, and Sur now expects the company to deploy 6 million to 7 million TPU processors in calendar 2027, up from his earlier estimate of 5 million units. Most of those 2027 shipments will support external customers, and they'll be powered entirely by Broadcom and Google's Sunfish 3nm TPU ASIC.

Even Google's internal TPU usage in 2027 should lean primarily on Sunfish, meaning more than 95% of Google's 6 million to 7 million TPU units that year will run on Broadcom silicon. That's not a partnership—that's dependency.

Orders in Hand, Supply Getting Tight

Broadcom already has multi-billion dollar purchase orders tied to the Sunfish ramp starting in the second half of this year. Sur believes Broadcom's AI backlog has grown by billions since the December earnings call, spanning both current-generation Ironwood and next-generation Sunfish shipments.

Tightening lead times for wafers, CoWoS packaging, high-bandwidth memory, and advanced substrates make early qualification critical. That dynamic will likely limit volume potential for later-starting programs like Zebrafish. And Broadcom isn't standing still—the company has already begun designing Google's next TPU ASIC on 2nm process technology, targeted for volume production in calendar 2028.

Sur's recommendation is blunt: buy aggressively at current levels. He projects first-quarter revenue of $19.14 billion and adjusted earnings per share of $2.02.

Price Action: Broadcom shares were up 1.93% at $320.10 at the time of publication on Monday.