To The Moon

AI chip stocks are moving higher ahead of Nvidia CEO Jensen Huang’s October 21 keynote at GTC Berlin.

Nvidia gained 1.38%, Marvell climbed 4.06%, AMD rose 1.76%, and Broadcom advanced 1.78%.

The market is looking for fresh signals about accelerated computing, AI infrastructure, AI agents and physical AI. But rather than focusing only on which company launches the most powerful chip, I think another question deserves attention:

What happens when AI becomes cheaper to run?

Lower inference costs could make AI applications more economical for businesses, potentially expanding demand beyond today’s biggest technology companies.

That could create opportunities across the semiconductor industry, although each company faces different challenges.

🟢 $NVIDIA(NVDA)$ : More Efficient AI, More Demand?

Nvidia has built a powerful position in AI computing through its GPUs, networking products and software ecosystem.

As businesses move from training AI models to using them in everyday applications, inference is becoming an increasingly important part of the market.

If computing becomes more efficient and affordable, companies may deploy AI across more products and workflows.

That could support additional demand for computing infrastructure.

However, investors must also consider competition, customer spending concentration and whether future growth can justify the valuation.

The key indicator: Can Nvidia sustain demand as AI workloads expand beyond model training?

🔵 $Marvell Technology(MRVL)$ : The Infrastructure Behind the Chips

AI systems require efficient connections between processors, memory and servers.

Marvell operates in custom silicon and data-centre connectivity, giving it exposure to the infrastructure needed to support increasingly complex AI workloads.

As major cloud providers develop specialised computing systems, custom chips and high-speed connectivity could become increasingly important.

But anticipated projects do not always translate immediately into revenue. Customer commitments, production schedules and actual orders will matter.

The key indicator: Is demand for custom AI silicon translating into sustained revenue growth?

🔴 $Advanced Micro Devices(AMD)$ : Can Competition Expand the Market?

AMD offers investors another way to participate in AI computing through its Instinct accelerators and broader processor business.

The opportunity is not necessarily to replace Nvidia entirely. As AI adoption expands, some customers may seek alternatives that suit their performance, cost and infrastructure requirements.

AMD’s ability to compete will depend on hardware performance, software compatibility and customer adoption.

Even a growing AI market does not guarantee that every competitor will gain market share.

The key indicator: Can AMD convert its AI product development into meaningful, recurring business?

🟣 $Broadcom(AVGO)$ : The Case for Custom AI Chips

Not every AI workload requires the same hardware.

Some large technology companies are investing in custom accelerators designed around their specific computing needs. These systems can offer advantages in efficiency and cost for suitable workloads.

Broadcom’s custom silicon and networking businesses give it exposure to this trend.

The opportunity depends on customers continuing to invest in specialised infrastructure and Broadcom converting those projects into profitable growth.

The key indicator: Can custom AI chips and networking sustain growth as customers optimise their infrastructure spending?

📊 The Bigger Picture: Follow the Economics

The next stage of AI investment may be shaped by three developments:

🔹 Lower costs: More affordable inference could encourage wider AI adoption.

🔹 Different architectures: GPUs, custom accelerators and networking solutions may all play important roles.

🔹 Financial returns: Businesses will increasingly need to demonstrate that AI spending delivers measurable productivity gains or additional revenue.

This creates an important distinction between a growing market and an attractive investment.

Even if AI infrastructure demand continues to expand, individual stocks may perform differently depending on competition, margins, execution and valuation.

GTC Berlin could provide useful clues about the industry’s direction. The more important test will be whether future announcements translate into orders, revenue and cash flow.

💬 The Question After GTC

If AI computing costs fall significantly, do you think the biggest opportunity will come from selling more powerful chips, developing cheaper custom accelerators, or connecting increasingly complex AI systems?

And which company is best positioned to benefit from that shift: Nvidia, Marvell, AMD or Broadcom?

The next winner may not simply be the company powering today’s AI boom, but the one best positioned for how AI is used tomorrow.

This post is for information and discussion only, not financial advice.