Broadcom, the semiconductor company that rode the first wave of artificial intelligence infrastructure spending, now faces a more demanding test as the industry's center of gravity shifts from training to inference. BMO Capital Markets has turned cautious following third-quarter earnings, not because the company is failing, but because the conditions that made its success so effortless are quietly dissolving. In the longer arc of technological cycles, this is a familiar inflection point — the moment when a rising tide recedes and individual seamanship begins to matter.
BMO flags headwinds for Broadcom as AI chip demand shifts to inference
The easy growth phase appears to be ending.
Mark
Why does it matter that capital is shifting from training to inference? Aren't both just AI work?
Mimi
Training is the heavy lifting—building the models from scratch. Inference is running them. Training requires massive, expensive chips optimized for parallel processing. Inference is more distributed, more varied, and the economics are completely different. Broadcom made a fortune on training infrastructure. Inference is a different game.
Mark
So Broadcom is losing market share?
Mimi
Not necessarily losing it, but the margin on that share is compressing. When training was the only game in town, Broadcom could command premium prices. Now there are more players, more options, and customers have leverage. The same chip might sell for less.
Mark
Is this temporary or permanent?
Mimi
It's structural. The training phase was a narrow window—a specific problem requiring specific solutions. Inference is broader, messier, and will be served by many different architectures. Broadcom will remain important, but it won't have the same pricing power.
Mark
What does BMO think will happen to the stock?
Mimi
They're signaling caution. Not a sell, but a wait-and-see. The next investor update from Broadcom will be crucial. If management can articulate a credible strategy for inference, the stock might stabilize. If they can't, there's more downside.
Mark
Why mention Micron specifically?
Mimi
Micron is positioning itself well for the inference era. It's a signal that the competitive landscape is shifting, and Broadcom's old advantages might not hold.
Le Pouls
- BMO Capital Markets has issued a cautious signal on Broadcom stock after Q3 earnings revealed margin compression even as revenues continued to grow.
- The AI infrastructure market is undergoing a structural shift — capital that once flooded into model training is now redirecting toward inference, a lighter and more competitive computational workload.
- Broadcom's traditional strengths in networking and training-era chips are less decisive in the inference market, where rivals like Micron are aggressively staking new ground.
- A major investor update is approaching, and the stakes are high — management must convince markets it has a credible roadmap for the inference era before confidence erodes further.
- The broader semiconductor AI supercycle is not over, but its easy-growth phase appears to be closing, leaving margin pressure and intensifying competition as the new normal.
Broadcom, the semiconductor company that rode the first wave of artificial intelligence infrastructure spending, now faces a more demanding test as the industry's center of gravity shifts from training to inference. BMO Capital Markets has turned cautious following third-quarter earnings, not because the company is failing, but because the conditions that made its success so effortless are quietly dissolving. In the longer arc of technological cycles, this is a familiar inflection point — the moment when a rising tide recedes and individual seamanship begins to matter.
Bank of Montreal's capital markets team has grown cautious on Broadcom after the semiconductor giant's third-quarter earnings revealed a stubborn tension: revenue is growing, but profit margins are shrinking. The concern is structural, not incidental.
For the past eighteen months, the AI boom was defined by training — the massive computational effort required to build large language models. Broadcom thrived in this environment, supplying the networking and infrastructure chips that held vast server farms together. But the market has begun to turn. Capital is now flowing toward inference, the lighter work of running trained models in the real world. Inference demands different chips, different architectures, and different suppliers — and Broadcom's advantages in the training era do not translate cleanly into this new phase.
BMO's warning is measured but pointed. The gap between what it costs Broadcom to produce chips and what customers will pay is narrowing, squeezed by intensifying competition, shifting product mix, and customers who now have more leverage. Rivals like Micron are moving aggressively into the inference landscape, and the competitive dynamics are far less settled than they were a year ago.
The moment of reckoning is near. Broadcom is preparing a major investor presentation where management will need to articulate a credible strategy for the inference era — one that preserves pricing power and market share in a more fragmented market. The company remains deeply profitable and broadly well-positioned, but the effortless growth phase appears to be ending. BMO's caution signals that analysts are not yet persuaded the transition will be smooth.
Bank of Montreal's capital markets team has turned cautious on Broadcom, the semiconductor giant that has ridden the artificial intelligence boom for the past two years. The concern, laid out after the company's third-quarter earnings, centers on a familiar but stubborn problem: the company's profit margins are shrinking even as revenue grows.
The shift is real and structural. The semiconductor industry has been bifurcated by the AI revolution. For the past eighteen months, the money flowed toward training—the massive computational work required to build and refine large language models. Broadcom benefited enormously from this phase, supplying the networking and infrastructure chips that connected the vast server farms doing this work. But the market is now moving. Capital is flowing toward inference, the lighter computational work of actually running trained models in production. This is a different problem, requiring different chips, different architectures, and different suppliers.
BMO's warning is not that Broadcom will collapse. It is that the company faces a period of margin compression—the gap between what it costs to make a chip and what customers will pay for it is narrowing. This happens when competition intensifies, when customers gain leverage, or when the product mix shifts toward lower-margin applications. In Broadcom's case, all three appear to be happening simultaneously. The training boom that lifted all boats is giving way to a more fragmented inference market where Broadcom's traditional strengths matter less.
The timing matters. Broadcom is preparing a major investor update, the kind of presentation where management lays out its vision for the next several years. Investors will be listening intently for signs that the company has a credible strategy for the inference era. The question is whether Broadcom can maintain its pricing power and market share as the industry's center of gravity shifts. Competitors like Micron are positioning themselves aggressively in this new landscape, and the competitive dynamics are far less settled than they were in the training phase.
What makes this moment significant is not that Broadcom is in trouble—the company remains enormously profitable and well-positioned in many markets. Rather, it is that the easy growth phase appears to be ending. The semiconductor industry's AI supercycle is not over, but it is entering a new chapter, one where margin pressure and competitive intensity will be the defining features. For Broadcom shareholders, the question is whether management can navigate this transition without a significant hit to earnings. BMO's caution suggests the bank's analysts are not yet convinced they can.
Citations marquantes
BMO sees negative signals for Broadcom's near-term stock performance based on Q3 earnings metrics and forward guidance— BMO Capital Markets analysis
Envie de l'histoire complète ?
Lire l'original sur Google News ↗
Economy & Finance