Broadcom (Nasdaq: AVGO) fell 5.9% Friday. Advanced Micro Devices (Nasdaq: AMD) rose 6.5%. A SemiAnalysis chip-research excerpt circulated during the session, and the fastest read was simple: AMD is taking Google's chip business from Broadcom.

Keep the causality honest. Broadcom also faced a separate Bank of America financing-risk note. AMD priced a $4.75 billion bond offering of its own. Broadcom reports earnings September 2. So the SemiAnalysis excerpt cannot explain the entire stock spread by itself. But it did create an important reading test, and that test is today's real story.

The excerpt contained three separate items. A 2026 production cut tied to advanced chip packaging. A 2027 allocation change involving Meta and OpenAI. And unconfirmed market chatter about AMD and a future Google TPU. Those are three separate signals. They are not one causal chain.

Google, a unit of Alphabet Inc. (Nasdaq: GOOGL), uses both NVIDIA chips and its own custom AI accelerators, called TPUs — tensor processing units built to train and run AI models fast.

Google controls the TPU product roadmap. Broadcom has agreed to develop and supply future TPU generations, and separately to supply networking and other components for Google's next-generation AI racks, through as late as 2031. The exact division of architecture, physical design, packaging, and manufacturing responsibility between the two companies is not fully public.

We reviewed a circulating excerpt, not the full report. That limits how far we can go. The conclusions below are SemiAnalysis estimates — not Broadcom guidance or company commitments.

The excerpt lowers its 2026 estimate for Broadcom-related CoWoS output because of a difficult CoWoS-S ramp. CoWoS is Taiwan Semiconductor Manufacturing Co. Ltd.'s (NYSE: TSM) advanced-packaging platform. It connects large logic chips with high-bandwidth memory inside one package, wired to talk to each other at very high speed. The excerpt attributes that 2026 reduction to the packaging ramp — not to AMD.

For 2027, the model lowers expected TPU output again and shifts more modeled packaging capacity toward Meta Platforms (Nasdaq: META) and OpenAI, both already Broadcom customers. Meta's part of that estimate is described as fluid. OpenAI's Jalapeño partnership with Broadcom is public; the specific capacity SemiAnalysis models for it is not a disclosed number.

Separate from both changes, the excerpt cites unconfirmed market chatter: Google may be working with AMD on a future TPU generation, called v10. AMD's angle would be packaging skill and CPU know-how, useful if Google wants a TPU with a built-in general-purpose processor core.

Read closely, and nothing in the excerpt we reviewed says AMD caused either estimate change. The 2026 cut is a packaging bottleneck. In SemiAnalysis's model, the 2027 change comes from Broadcom-related reserved capacity shifting toward Meta and OpenAI. The AMD chatter is a third, unrelated item about a possible future design win.

Nothing in the excerpt links AMD to Broadcom losing volume.

But the excerpt stacks all three points back to back. A busy portfolio manager skims forty reports before lunch. Read that fast, and it looks like one story: AMD is taking Broadcom's business. That's not what the data says. It's what the layout implies.

SemiAnalysis may be right about all three items individually. Broadcom may have a packaging problem. Meta and OpenAI may be taking more of its capacity. AMD may be working with Google on v10. The mistake isn't believing any single claim. The mistake is treating proximity on a page as proof of causality.

Broadcom's April agreement with Google is real. It's strong evidence that Broadcom stays central to Google's TPU roadmap. Public filings describe two parts: one deal to develop and supply future TPU chips, and a separate deal to supply networking and rack parts. Together they run through as late as 2031.

That argues against Broadcom being replaced. It does not guarantee Broadcom keeps every piece of every future TPU, or the same share of Google's spending each year.

Google can keep Broadcom as a main partner, add AMD for a role in a later design, and still reduce Broadcom's share of total spending. All three things can happen at once. Adding a new vendor isn't the same as firing the old one.

The AMD side deserves the same care. We don't know yet what AMD would actually supply: CPU chips, packaging know-how, joint design work, or some mix of the three. We don't know if that work would sit alongside Broadcom's or eventually replace some of it. We don't know the volume or the dollars involved either.

One more wrinkle: Google already has its own Axion CPU family, and uses CPU services alongside its TPU and GPU infrastructure for orchestration and surrounding workloads. That makes the exact purpose of a possible AMD CPU role more important to pin down — not less. Until Google, AMD, or Broadcom confirms the role, the price, and the volume, this is upside for AMD to prove — not a transfer of Broadcom's chip business.

A bigger idea is moving through markets right now, and it applies here. Some AI companies are pure bets on having the smartest model — call them frontier labs. Others make money no matter which model wins, because they collect a toll on the infrastructure underneath all of them.

Broadcom's chip business increasingly looks like the second kind. It designs custom chips built for one customer's job, instead of general-purpose chips anyone can buy — that's what "ASIC" means. Broadcom doesn't need to be Google's only chip partner. Google, Meta, and OpenAI all show up in this same excerpt as Broadcom customers. More customers means less risk from any one of them walking away.

But Broadcom is not a passive toll collector. Every lane on this road has to be re-won. It has to win each new chip design, build it, secure enough foundry and packaging capacity, and ship it on time. Then it has to do that again for the next generation. More customers lower one kind of risk. They don't remove the risk of losing a design, or the risk of running short on capacity.

Broadcom and Samsung Electronics Co., Ltd. (Korea Exchange: 005930) also signed an MOU estimated at more than $200 billion through 2030. The planned collaboration covers high-bandwidth memory, Samsung's 2-nanometer-and-below foundry processes, and advanced chip packaging. That creates a credible potential second manufacturing ecosystem for future Broadcom products. It is not a disclosed Google-TPU order, a final product-specific supply contract, or a demonstrated fix for the current CoWoS-S constraint.

The real risk to Broadcom's custom-silicon business is not whatever role AMD may eventually win. It's whether Broadcom keeps winning valuable design content and securing enough capacity to ship what it wins. The competition for new custom-silicon programs won't disappear even if this particular AMD rumor proves wrong.

Broadcom Inc. (Nasdaq: AVGO). The excerpt we reviewed doesn't show AMD causing either the 2026 or 2027 estimate changes. Broadcom also remains central to Google's roadmap through as late as 2031 and has separate public, multi-generation partnerships with Meta and OpenAI. Meta's initial commitment above one gigawatt and OpenAI's multi-generation Jalapeño platform are both publicly confirmed; the specific production allocations in the SemiAnalysis excerpt are not. Risk: a packaging problem can still turn into a revenue problem. Fewer packaged chips mean delayed shipments, and a long-term agreement doesn't guarantee the same share in every future generation.

Alphabet Inc. (Nasdaq: GOOGL). Google already mixes its own TPU chips, its own Axion CPU chips, and outside partners across design and manufacturing. If Google adds AMD or another qualified partner, it gains more leverage and greater supply resilience. Risk: more vendors also means more moving parts to manage. More partners doesn't automatically mean lower total cost.

Advanced Micro Devices, Inc. (Nasdaq: AMD). A confirmed TPU-v10 role would represent genuine upside and an important custom-silicon foothold. Today, it remains market chatter, with no disclosed scope, price, or production volume.

Meta Platforms, Inc. (Nasdaq: META) and OpenAI. Meta's multi-generation Broadcom partnership is public and initially covers more than one gigawatt. Its specific modeled capacity allocation in the SemiAnalysis excerpt is described as fluid. OpenAI's Jalapeño platform is publicly confirmed, but scaled production and program economics remain to be proved.

We aren't chasing the "AMD long, Broadcom short" trade the market read into Friday's move. The core thesis still looks intact: Broadcom remains part of Google's roadmap, it has public programs with Meta and OpenAI, and Samsung could broaden its future manufacturing base. So we're evaluating the pullback as a possible entry point and would prefer defined-risk options to an unrestricted equity position. We're watching for a disclosed Meta capacity commitment, since the excerpt's estimate remains fluid. And we aren't adding to AMD on TPU chatter alone.

I get more cautious on Broadcom if Google materially amends or reduces its agreement, if subsequent disclosures show further packaging reductions beyond those modeled here, or if AMD confirms a TPU-v10 role with material economics and production volume.

I get more constructive if the CoWoS-S ramp improves, if Broadcom converts the Samsung MOU into product-specific production orders, or if Meta's modeled capacity turns into a disclosed shipment commitment.

SemiAnalysis cut its Broadcom TPU estimate. That matters. But the reason matters more.

In the model, the 2026 reduction came from a packaging constraint. The 2027 reduction came from reserved capacity shifting toward Meta and OpenAI. AMD appeared in a separate, unconfirmed item about a possible future project.

That doesn't make AMD irrelevant. It doesn't make Broadcom untouchable. It means the market skipped a step in the causal chain.

Broadcom's real test is whether it keeps winning valuable design content — and secures enough packaging capacity to turn those wins into shipments. Three separate signals became one trade. The missing sentence is the one that separates them.

The HEAT (Hedge, Edge, Asymmetry and Theme) Formula is designed to empower investors to spot opportunities, think independently, make smarter (often contrarian) moves, and build real wealth.

The views and opinions expressed herein are those of the Chief Executive Officer and Portfolio Manager for Tuttle Capital Management (TCM) and are subject to change without notice. The data and information provided is derived from sources deemed to be reliable but we cannot guarantee its accuracy. Investing in securities is subject to risk including the possible loss of principal. Trade notifications are for informational purposes only. TCM offers fully transparent ETFs and provides trade information for all actively managed ETFs. TCM's statements are not an endorsement of any company or a recommendation to buy, sell or hold any security. Trade notification files are not provided until full trade execution at the end of a trading day. The time stamp of the email is the time of file upload and not necessarily the exact time of the trades. TCM is not a commodity trading advisor and content provided regarding commodity interests is for informational purposes only and should not be construed as a recommendation. Investment recommendations for any securities or product may be made only after a comprehensive suitability review of the investor’s financial situation.© 2026 Tuttle Capital Management, LLC (TCM). TCM is a SEC-Registered Investment Adviser. All rights reserved.