Broadcom Has Contracts Through 2031, But the Market Still Doesn't Believe It
Morgan Stanley published on Tuesday, July 14, a defense note on Broadcom that deserves to be read carefully — not for what it says about the stock, but for what it reveals about the value architecture underpinning the semiconductor manufacturer and why that architecture has still failed to convince investors.
The starting point is the concern that took hold in the market following a report by The Information in March: MediaTek, the Taiwanese chip manufacturer, would be collaborating with Alphabet to develop the next generation of Tensor Processing Units (TPUs) used by Google in its data centers. The result was a correction that pushed Broadcom's shares to nearly 22% below their all-time closing high of $481.57 recorded on June 2. The year-to-date return settled at approximately 14%, while the iShares Semiconductor ETF advanced roughly 90% over the same period.
The market's reaction is, in itself, an analytical signal. It does not express irrational panic. It expresses a hypothesis about the design of the relationship between Broadcom and its large clients — and about how much of that relationship is genuinely difficult to replace versus how much rests on the temporary absence of alternatives.
The Mechanics Morgan Stanley Is Defending
Morgan Stanley's argument does not rest on Google's loyalty. It rests on something more concrete: transition costs, the technical depth of the relationship, and the contractual evidence that already exists.
In April of this year, Broadcom formalized in a regulatory filing a long-term agreement with Google to design and supply future generations of TPUs through at least 2031. The same document includes a Supply Assurance Agreement for networking components used in next-generation artificial intelligence racks, also with potential validity through 2031. These are not verbal commitments or strategic intentions. They are contractual structures that imply joint engineering planning, shared investment in design, and bidirectional technical dependency.
What Morgan Stanley explicitly points out is that MediaTek's participation is real, but it does not jeopardize Broadcom's position. The logic supporting that reading lies in Google's incentives: Alphabet has a vested interest in reducing dependence on any single supplier in order to maintain cost flexibility and negotiating leverage. But diversifying suppliers is not the same as displacing the dominant supplier. It means introducing a second player that pressures prices and validates the technology. Within that framework, Morgan Stanley projects that Broadcom will maintain approximately 80% of the TPU business over time, and characterizes the bearish scenarios of a drop to 50% or total displacement as premature.
The technical reason analysts cite for that structural stability is not trivial. High-bandwidth memory, execution in advanced chip packaging, and production scale are capabilities that cannot be replicated quickly. Broadcom does not sell just a chip: it sells an integration stack that Google has incorporated into its computing architecture cumulatively over several years. Replacing that, even if desired, requires time, investment, and tolerance for technical risk that no hyperscaler can take lightly at the moment when the race for artificial intelligence infrastructure is at its most intense.
The Contracts with Apple, Meta, and Anthropic as a Signal of the Model
The Broadcom case cannot be fully understood if it is reduced to its relationship with Google. In the week prior to the Morgan Stanley note, Apple announced an extension of its chip manufacturing agreement with Broadcom valued at more than $30 billion. In April, Meta extended its own custom silicon contract. And the same agreement disclosed with Google includes an additional component: beginning in 2027, Anthropic will access approximately 3.5 gigawatts of TPU-based computing capacity through Broadcom and Google — a significant leap from the roughly 1 gigawatt available in 2026.
What emerges from this chain of agreements is not a company with one large customer. It is a manufacturer that has built deep contractual positions with five of the most consequential players in the expansion of artificial intelligence infrastructure: Google, Anthropic, Meta, Apple, and potentially others that have not yet made public announcements. Each of those relationships involves joint design, which creates technical reciprocity. It is not simply that Broadcom manufactures and they buy. It is that the design cycle of each chip generation incorporates engineering from both sides, which makes the exit cost genuinely high for the customer.
This model has a particular distributive logic. Broadcom concentrates the technical risk of design and production, but in exchange it obtains long-term revenue visibility and exit barriers that go beyond price. Its customers sacrifice some independence and short-term bargaining power in exchange for access to guaranteed production capacity and a technological curve they cannot replicate internally within the timeframe they require. The supply assurance agreement through 2031 with Google is the clearest example: it is not merely about buying chips — it is about reserving production capacity in a market where that capacity is the scarcest resource.
Why the Market Is Still Discounting What the Contracts Already Guarantee
The paradox of Broadcom's stock performance in 2026 is that the company reported solid results in its fiscal second quarter, signed long-term agreements with customers of the highest relevance, received public defense from one of the most influential investment banks in the world — and yet its shares fell more than 12% following the results and have accumulated a severe lag relative to the sector.
The most honest explanation does not lie in technical analysis of the stock. It lies in the structure of the problem the market perceives — one that the contracts have still not resolved.
The first element is concentration. Broadcom has deep ties with a small number of enormously large customers. This creates enviable revenue visibility, but it also exposes the company to a structural vulnerability: if any one of those customers makes a strategic decision to internalize design or switch suppliers, the impact would not be marginal. The MediaTek news activated precisely that fear, regardless of how well-founded it actually is.
The second element is the conservative guidance that the company itself offered regarding artificial intelligence revenues for the following quarters. When a company operating in the hottest segment of the technology market issues cautious guidance after a strong result, investors interpret that as internal friction in business visibility. Not necessarily that the business is deteriorating, but that the company itself does not know with certainty how much of the current cycle will be sustained.
The third element is comparative. The semiconductor ETF rose roughly 90% over the year because several components of the index are directly exposed to the GPU demand cycle for training artificial intelligence models — Nvidia in particular. Broadcom plays at a different layer: custom chips for inference and networking, not GPUs for massive-scale training. That makes its growth profile more predictable but less explosive, which generates a relative discount in a market that rewards the narrative of vertical growth over contractual stability.
The Tension That Contracts Alone Cannot Resolve
Broadcom has one of the strongest long-term commitment portfolios in the semiconductor sector for artificial intelligence. Its agreements are not mere intentions: they are structures with defined timelines, committed capacities, and joint technical architectures. That differentiates it from competitors who sell at spot volumes or who depend on annual renewal cycles.
But the market is asking a different question from the one Morgan Stanley answers. The bank's analysts defend the view that Broadcom will retain 80% of the TPU business. The market's implicit question is something else entirely: given that the remaining 20% could go to MediaTek or to an in-house Google design, and given that overall growth in the total artificial intelligence chip market could concentrate in the segment where Broadcom does not participate directly, how much is that 80% position actually worth in practice?
Morgan Stanley's price target of $502 per share implies a considerable appreciation runway from current levels. That target assumes the market will eventually recognize the value of Broadcom's contractual commitments and its technical position. It may well do so. But the timing of that recognition depends on variables that the contracts do not control: the speed at which MediaTek demonstrates real capability in high-end TPUs, the depth with which Anthropic executes the 3.5-gigawatt ramp in 2027, and the credibility with which Broadcom translates those commitments into growth guidance that the market can model.
What this case illustrates with clarity is that having solid contracts and having market recognition are two things that can remain out of sync for extended periods. Broadcom is in that interval right now. The distributive structure of the business favors the retention of its most important customers. What remains unresolved is whether that retention translates into revenue growth at a pace sufficient to justify the valuations that the sector demands at this particular moment in the artificial intelligence investment cycle.
