Morgan Stanley pushes back on bearish takes about Google TPU, as Broadcom shares rebound

Broadcom shares rose on Tuesday, July 14, 2026, after Morgan Stanley issued what the report called one of its most forceful defenses yet of Broadcom’s position in the artificial intelligence chip supply chain. The trigger for the move was not a product announcement from either company, but a dispute over who benefits most from the next wave of AI infrastructure spending.

At the center of the debate is Google’s Tensor Processing Unit, or TPU, a custom chip designed to accelerate machine learning workloads for Google’s own AI systems and for customers that deploy Google’s cloud services. In recent market commentary, some investors have argued that Google’s TPU strategy could reduce the need for certain third-party AI accelerators that Broadcom relies on through its portfolio of networking and semiconductor assets.

According to the Yahoo Finance report, Morgan Stanley challenged that bearish framing directly, arguing that critics have overstated the impact of Google’s TPU approach on Broadcom’s opportunities. While the post summarized the tone of the note as particularly strong, it did not provide a detailed list of the specific assumptions Morgan Stanley used, or the exact quantitative changes analysts made to their models.

The immediate stock reaction suggested that at least some investors viewed Morgan Stanley’s view as a credible counterweight to the broader concerns. Still, the article described the “relief” as fragile, indicating that the market remains sensitive to any new evidence about demand for AI chips and the competitive balance between custom silicon and merchant AI accelerators.

The episode underscores how quickly AI chip narratives can shift, even without new company disclosures. In this cycle, investors often weigh (1) how much compute is being built for AI training and inference, (2) whether hyperscalers increasingly prefer custom chips such as TPUs, and (3) how networking and data-center interconnects factor into total AI infrastructure spending.

For Broadcom, the key question is whether its products and customer relationships capture a meaningful share of AI datacenter buildouts, even if large buyers like Google deploy custom accelerators. Morgan Stanley’s stance, as characterized in the report, implies that the firm believes Broadcom can still benefit from AI demand despite the presence of alternative chip architectures.

For Alphabet, the implication is that TPU is not simply a winner-take-all substitute for the rest of the industry. But the report did not cite any new Alphabet TPU performance metrics, roadmaps, or customer commitments. It also did not tie the dispute to any new earnings commentary or guidance from either Alphabet or Broadcom during the same timeframe.

What is not clear from the available material is the depth of Morgan Stanley’s evidence, including any estimates for how TPU adoption translates into the share of AI infrastructure spending that accrues to Broadcom’s relevant businesses. Investors will likely watch for next steps, such as follow-up analyst notes with updated numbers, or disclosures in upcoming earnings reports that either firm provides about AI-related demand and customer spending patterns.