Broadcom Stock Is Up 3% After Hours and It’s Only the Beginning of the AI Chip Story

It was announced sometime in the late afternoon of April 14, 2026, that Meta Platforms had committed to an initial deployment of more than one gigawatt of computing capacity as part of an extension and expansion of its custom AI chip partnership with Broadcom through 2029. After hours, Broadcom’s stock was trading at $392.97, up over 3% from the close of the regular session at $380.78. Pre-market trading had the stock moving back toward $392 by early the following morning. It was the kind of move that occurs when a particular catalyst strikes a stock that was already primed by a bigger story, and Broadcom’s bigger story at the moment is truly among the most captivating in the semiconductor sector.

It’s helpful to know what Broadcom CEO Hock Tan said during the company’s most recent earnings call in order to comprehend why investors responded to the Meta announcement in this manner. He informed analysts that Broadcom currently has six significant AI clients, which are hyperscale businesses constructing the infrastructure needed to run massive AI workloads and language models for the ensuing ten years. According to him, the company has a clear path to earning more than $100 billion from AI chips by 2027. That’s not a nebulous goal.

It’s a figure based on committed supply arrangements and signed contracts, the kind of revenue visibility that semiconductor companies hardly ever have this far ahead. The Meta agreement, which is based on Meta’s unique silicon program called MTIA—the Training and Inference Accelerator that Mark Zuckerberg has been developing to give Meta control over its own AI hardware destiny—fits neatly into that picture. It is a 1-gigawatt commitment that will grow to multiple gigawatts by 2027 and beyond.

Here, the larger context is important. Meta, Google, Amazon, Microsoft, and a few other hyperscale tech companies have been gradually attempting to lessen their reliance on Nvidia’s costly but dominant GPU architecture. They’re not giving up on Nvidia. Many AI training workflows still rely heavily on Nvidia’s H100s and later chips.

However, these businesses are astute enough to understand that creating custom silicon for their most specialized, high-volume workloads results in efficiency gains that increase with scale. A custom chip tailored to your specific workload can be significantly more efficient than a general-purpose GPU when you’re processing billions of inference requests every day. This efficiency directly translates into lower operating costs and a competitive edge. For precisely this custom silicon build-out, Broadcom has established itself as the go-to design and supply partner.

This positioning is already producing revenue at a rate that surprised even optimistic analysts, according to the recently released Q1 FY2026 results. With $19.31 billion in total revenue, it was up 29.47% from the previous year. In particular, AI semiconductor revenue reached $8.4 billion, up 106% from the same period last year. That number is noteworthy because it shows that Broadcom’s AI business is not just theoretical or experimental. It is accelerating and operating at scale. The stock’s trajectory following the earnings beat on both EPS and revenue sent it from a brief decline below $300 in late March toward the $380 level before the news of the Meta deal drove it higher.

Alongside the Meta agreement, the Google partnership merits equal consideration. For Google’s next-generation AI infrastructure, Broadcom has agreed to design and supply custom tensor processing units, or TPUs, through 2031. This five-year supply contract covers not only chip design but also networking components, PCIe switches, optical connectivity, and the larger infrastructure ecosystem that enables massive AI deployments at the data center level. It is with one of the most picky technology buyers in the world. With 3.5 gigawatts of computing capacity beginning in 2027, the Anthropic agreement adds yet another significant AI player to a roster that is starting to stand out for its concentration of AI infrastructure spending. In just a few weeks, three deals have been announced, and during the earnings call, Hock Tan confirmed six major AI clients. The business does not depend on a single clientele.

Over his nearly two decades as CEO of Broadcom, Hock Tan has become something of a quiet legend in the semiconductor industry. He is renowned for his disciplined capital allocation, strategic acquisitions that truly integrate, and willingness to make long-term bets that the market initially underestimates. When the $69 billion acquisition of VMware was announced in 2023, it was greeted with a great deal of skepticism. Since then, it has significantly increased recurring software revenue for a company that was already making good cash flows from semiconductors. Tan’s transition from Meta’s board to a chip strategy-specific advisory position, which was disclosed as part of the Meta acquisition, demonstrates how closely the two businesses are now linked.

The P/E ratio of 74x trailing earnings is the figure that makes anyone approaching this with a valuation-first mindset pause. It’s still unclear how much further Broadcom stock can go from here. However, the forward P/E ratio presents a different picture, and the analyst price target consensus of $473, which is based on Tan’s revenue trajectory, indicates that the market may not have fully priced in what 2027 looks like if the $100 billion AI chip revenue forecast comes to pass. It’s difficult to ignore the fact that the businesses entering into these long-term contracts are not doing so carelessly. Meta, Google, and Anthropic are indicating where they anticipate demand when they lock in supply through 2029 and 2031. And the business at the heart of all three of those commitments is Broadcom.