Invest Today

Investment Research | November 2026

Introduction: The Difference Between a Great Company and a Great Investment

Nvidia has arguably been the greatest beneficiary of the artificial intelligence revolution. Under Jensen Huang’s leadership, the company has transformed itself from a graphics processing specialist into the dominant supplier of AI computing infrastructure.

Its graphics processing units (GPUs) power some of the world’s most advanced artificial intelligence models, while its CUDA software ecosystem has become an industry standard. Nvidia has established an extraordinary competitive position, delivering remarkable revenue growth, profitability and shareholder returns.

AMD, meanwhile, has spent much of the past decade rebuilding its competitive position in semiconductors. Under CEO Lisa Su, the company has transformed itself into a formidable competitor in CPUs and is now making significant investments in AI accelerators.

Both businesses are exceptionally well positioned to benefit from the enormous capital expenditure taking place across artificial intelligence infrastructure. However, I believe AMD has the potential to outperform Nvidia as an investment over the next five years.

My reasoning is relatively straightforward. Nvidia does not need to lose its leadership position for AMD to become an enormously successful business. AMD simply needs to capture a meaningful portion of the incremental demand in a rapidly expanding market.

The distinction between the best company and the best investment is particularly important here. Nvidia may remain the dominant player, but AMD’s smaller revenue base offers considerable potential for growth.

1. Understanding Nvidia’s Dominance

To appreciate AMD’s opportunity, we must first understand the scale of Nvidia’s dominance.

In its second quarter reported in 2026, Nvidia generated approximately $96.2 billion in revenue, an increase of 106% year over year. Data Center revenue reached $89 billion, highlighting the extraordinary demand for accelerated computing infrastructure.

AMD generated $11.5 billion in total revenue during its second quarter of 2026, representing annual growth of approximately 50%. Its Data Center segment generated $6.7 billion, more than doubling from the previous year.

The financial comparison is striking.

Financial metric	Nvidia	AMD
Latest quarterly revenue	$96.2B	$11.5B
Annual revenue growth	106%	50%
Quarterly data centre revenue	$89.0B	$6.7B
GAAP gross margin	75%	54%
Market capitalisation	~$4.94T	~$824B
Forward P/E ratio	~16.9×	~45.1×

Sources: Nvidia and AMD Q2 2026 results and Stock Analysis market data as of 30 October 2026. Nvidia’s quarter ended in July 2026 and AMD’s in June 2026.

Nvidia’s financial performance is extraordinary, and its profitability reflects a strong competitive moat. Nevertheless, I believe AMD’s considerably smaller revenue base creates a compelling opportunity.

An additional $10 billion in annual revenue would represent a relatively modest increase for Nvidia. For AMD, the same amount would be transformational.

This asymmetry sits at the centre of my investment thesis.

2. AMD Doesn’t Need to Beat Nvidia

One of the mistakes investors sometimes make is assuming that an industry must have a single winner.

History suggests otherwise.

Intel dominated the CPU market for decades, yet AMD successfully challenged that position by developing increasingly competitive processors and capturing market share.

I believe something similar could happen in AI infrastructure.

Nvidia enjoys significant technological advantages, particularly through CUDA, its networking technologies and its integrated computing systems. However, the largest technology companies have strong incentives to develop alternatives.

Microsoft, Amazon, Google, Meta and other hyperscalers are collectively spending enormous sums on AI infrastructure.

These companies are sophisticated buyers. They understand the risks associated with relying too heavily on a single supplier.

A credible second supplier provides greater negotiating leverage, additional manufacturing capacity and potentially lower computing costs.

AMD does not necessarily need to outperform Nvidia technologically across every application. Instead, it needs to offer an attractive combination of performance, pricing, availability and software compatibility.

As AI infrastructure becomes increasingly competitive, I believe AMD could benefit considerably from the demand for supplier diversification.

Importantly, this does not mean Nvidia must lose revenue.

In an expanding market, AMD can gain market share while Nvidia continues to grow.

3. The Market-Share Opportunity

This is where I believe the investment thesis becomes particularly interesting.

Rather than attempting to predict whether AMD will overtake Nvidia, I prefer to consider what happens if AMD captures a relatively modest percentage of the future AI accelerator market.

For illustration, suppose the annual global market for AI accelerators eventually reaches $300 billion.

This is a hypothetical market-size assumption, not a verified forecast. The figure excludes CPUs, conventional servers and other categories of data centre infrastructure.

Under this assumption, AMD’s annual accelerator revenue could develop as follows:

AMD Illustrative AI Accelerator Revenue Scenarios

Scenario	AMD market share	Annual accelerator revenue
Conservative	10%	$30B
Base case	15%	$45B
Optimistic	20%	$60B

Illustrative scenarios using a hypothetical $300 billion annual market. These are not company guidance or revenue forecasts.

The implications are significant.

At a 15% market share, AMD would generate approximately $45 billion annually from AI accelerators alone.

For context, AMD’s entire business generated approximately $34.6 billion in revenue during 2025.

Such an outcome would represent a substantial expansion of the company.

Furthermore, these figures exclude AMD’s server CPU business, consumer processors, gaming products and embedded technologies.

Naturally, market-share gains would depend on successful execution, manufacturing capacity, product competitiveness and customer adoption.

The size of the addressable market also remains uncertain. A smaller market, lower selling prices or slower infrastructure investment would materially affect these outcomes.

Nevertheless, the exercise illustrates why I find AMD attractive.

The company does not need to control the market to achieve substantial revenue growth.

It simply needs to establish itself as a meaningful second supplier.

4. The Economics of Competing Against Nvidia

Nvidia’s extraordinary profitability is one of its greatest strengths.

Its gross margins of approximately 75% demonstrate the considerable pricing power it currently enjoys.

AMD operates with lower gross margins, reporting a GAAP gross margin of 54% and a non-GAAP gross margin of 56% during Q2 2026.

Nevertheless, I believe AMD’s existing profitability provides a strong foundation for expansion.

As a growing company, AMD has the opportunity to increase its revenue substantially while continuing to improve its manufacturing economics, product mix and operating efficiency.

Consider the perspective of a hyperscaler purchasing tens of billions of dollars’ worth of computing infrastructure. Even relatively small reductions in total computing costs could translate into enormous savings.

AMD may be able to compete by offering customers attractive performance at a lower price while maintaining healthy gross margins.

This would not necessarily be detrimental to AMD shareholders. A smaller company expanding its revenue base rapidly can generate substantial additional gross profit, even if its margins remain below those of the market leader.

Importantly, I believe AMD does not need to approach Nvidia’s 75% gross margin to establish a highly attractive business.

If AMD can maintain gross margins in the mid-50% range while substantially expanding its AI accelerator revenue, the resulting increase in gross profit could be significant.

To illustrate this opportunity, I have developed three scenarios using assumptions broadly consistent with AMD’s existing profitability.

Illustrative AMD AI Accelerator Gross Profit Scenarios

Scenario	Accelerator revenue	Assumed gross margin	Annual gross profit
Conservative	$30B	50%	$15.0B
Base case	$45B	55%	$24.75B
Optimistic	$60B	60%	$36.0B

These are illustrative AI accelerator scenarios, not company forecasts. AMD’s Q2 2026 company-wide GAAP gross margin was 54%. Actual accelerator margins are not separately disclosed and could differ significantly. Gross profit excludes research and development, administration and other operating expenses.

The figures demonstrate the scale of AMD’s potential opportunity.

Under my base case, AMD would generate approximately $45 billion in annual AI accelerator revenue, producing nearly $25 billion in gross profit.

This would be a substantial contribution from a single product category.

The optimistic scenario assumes that AMD can improve its gross margins to approximately 60% as it achieves greater scale and a more favourable product mix.

Such an outcome would depend on improvements in production efficiency, competitive positioning and the company’s ability to maintain attractive pricing.

However, even the conservative scenario demonstrates that AMD could generate significant additional gross profit while operating below its current company-wide margin.

There are several reasons why I believe maintaining or improving gross margins is plausible.

Firstly, AMD is already generating gross margins above 50% across its business.

Secondly, the expansion of high-value data centre products could support a more profitable revenue mix, provided those products earn attractive margins.

Thirdly, increasing production volumes could improve manufacturing economics, although advanced packaging and memory costs remain important constraints.

Finally, AMD’s growing relationships with major technology companies may provide greater visibility into future demand, allowing the company to plan production more efficiently.

Of course, these advantages must be weighed against intensifying competition and potential pricing pressure.

Gross profit also should not be confused with net income. AMD will continue investing substantial sums in research, software development, administration and customer support.

For shareholders, the ultimate question is how much additional gross profit AMD can convert into operating income and free cash flow.

Nevertheless, I believe AMD’s existing gross margins, combined with the potential for substantial revenue growth, provide a compelling foundation for long-term earnings expansion.

5. OpenAI, Anthropic and the Importance of Commercial Validation

One of the most encouraging developments for AMD has been its growing commercial relationships with leading artificial intelligence companies.

In October 2025, AMD announced a multiyear agreement with OpenAI covering six gigawatts of GPU infrastructure.

The agreement outlined an initial one-gigawatt deployment of AMD Instinct MI450 GPUs beginning in the second half of 2026.

In July 2026, AMD announced a separate strategic partnership with Anthropic involving up to two gigawatts of MI450-based infrastructure.

The Anthropic partnership includes collaboration on optimising Claude workloads for AMD GPUs and improving AMD’s ROCm software ecosystem.

These developments are significant for several reasons.

Firstly, they demonstrate that AMD is increasingly being considered for some of the world’s most demanding AI workloads.

Secondly, relationships with major AI developers could accelerate improvements in AMD’s software ecosystem.

Finally, these partnerships provide greater commercial credibility when AMD approaches other large infrastructure customers.

There is, however, an important distinction between announced infrastructure agreements and realised revenue.

Deployments of this scale require considerable electrical capacity, manufacturing resources, physical infrastructure and financing.

The timing and profitability of these agreements will therefore be important indicators of AMD’s future success.

Nevertheless, I believe these partnerships represent meaningful steps towards establishing AMD as a credible alternative to Nvidia.

6. CUDA, ROCm and the Software Moat

Nvidia’s greatest competitive advantage may not be its hardware.

It may be the software environment surrounding it.

CUDA allows developers to build and optimise applications specifically for Nvidia GPUs. Over many years, Nvidia has established an extensive ecosystem of development tools, software libraries and optimised applications.

This creates substantial switching costs.

Even when competing GPUs offer attractive hardware specifications, migrating existing workloads may require considerable engineering resources.

AMD’s response is ROCm, its open software platform for accelerated computing.

ROCm aims to make it easier for developers to build and deploy applications on AMD hardware.

The challenge for AMD remains substantial. Nvidia’s software advantage extends beyond basic compatibility into optimisation, reliability, developer familiarity and integration with broader computing systems.

Nevertheless, I believe the continued development of open software ecosystems could gradually reduce Nvidia’s competitive advantage.

The partnership between AMD and Anthropic is particularly interesting in this regard.

By collaborating directly with major AI developers, AMD can improve the performance and reliability of its software stack on real-world workloads.

Over time, this could lower the cost and complexity of adopting AMD accelerators.

I do not expect CUDA’s competitive advantage to disappear. However, AMD does not need to eliminate this advantage entirely.

It only needs to narrow the gap sufficiently for more customers to consider its hardware economically viable.

7. AI Inference Could Change the Competitive Landscape

Perhaps the most important technological development underpinning my thesis is the transition from AI training towards an increasingly inference-intensive market.

During the early stages of the AI revolution, much of the industry’s attention was directed towards training increasingly sophisticated models.

Training requires enormous computing resources and has been a major driver of Nvidia’s success.

However, the widespread deployment of AI systems introduces another significant source of computing demand: inference.

Inference is the process through which trained models generate responses, make predictions and perform useful work.

Consider a future in which millions of AI agents continuously perform tasks across corporate software systems.

These agents could manage administrative processes, analyse financial information, coordinate workflows and execute transactions.

Every interaction requires computing capacity.

As the number of AI-powered applications expands, the demand for inference could become enormous.

Importantly, inference economics differ from those of training.

Training often places a premium on obtaining the greatest available computational performance to complete large workloads efficiently.

Inference introduces additional considerations, including latency, memory capacity, power consumption, utilisation rates and cost per token.

For companies operating large-scale AI services, these economics are particularly important.

If AMD can offer competitive inference performance at an attractive total cost of ownership, it may be able to capture a meaningful portion of this expanding market.

I believe this creates an opportunity for alternative accelerator architectures.

However, Nvidia is also investing aggressively in inference performance and software optimisation. It would therefore be incorrect to assume that growing inference demand will automatically benefit AMD disproportionately.

AMD must demonstrate competitive performance in real-world deployments.

Ultimately, the companies that can provide reliable computing at the lowest economically sustainable cost will be well positioned to capture demand.

8. AMD’s CPU Business Provides Another Growth Engine

One aspect of AMD’s business that I find particularly attractive is its exposure to both CPUs and GPUs.

While GPUs dominate attention in AI infrastructure, CPUs remain essential components of modern computing systems.

They coordinate workloads, manage operating systems and support the broader infrastructure surrounding AI accelerators.

AMD’s EPYC processors have established a meaningful position in the server market.

During the second quarter of 2026, AMD’s Data Center segment generated approximately $6.7 billion in revenue, increasing 107% year over year.

This growth was supported by strong demand for both EPYC processors and Instinct GPUs.

AMD is therefore positioned to benefit from two important trends: continued demand for conventional computing and the expansion of accelerated AI workloads.

The company is also developing integrated infrastructure solutions through its Helios rack-scale platform.

This matters because major customers increasingly purchase complete computing systems rather than individual chips.

By combining CPUs, GPUs and networking technologies, AMD has an opportunity to compete for a greater portion of infrastructure expenditure.

I believe this diversification provides AMD with additional opportunities beyond AI accelerators alone.

9. Valuation: The Biggest Challenge to My Thesis

A great business does not necessarily represent a great investment at any price.

This is particularly important when comparing Nvidia and AMD.

As of 30 October 2026, Nvidia traded at approximately 16.9 times forward earnings, compared with roughly 45.1 times for AMD.

AMD therefore commanded a substantial forward earnings premium despite Nvidia’s stronger profitability.

This is arguably the strongest counterargument to my investment thesis.

Investors are already paying a significant premium for AMD’s expected earnings growth.

For AMD to outperform Nvidia, its future earnings must grow sufficiently quickly to justify that premium.

Simply gaining market share may not be enough.

A useful way to understand the relationship between growth and valuation is to consider the following principle:

Shareholder returns depend not only on how quickly a company’s earnings grow, but also on the valuation investors are willing to assign to those earnings.

For example, a company that doubles its earnings while its P/E ratio halves would experience little change in its share price, assuming no changes in its share count.

Conversely, rapid earnings growth accompanied by a stable valuation multiple can produce substantial shareholder returns.

AMD therefore needs to deliver strong earnings growth while avoiding excessive valuation compression.

The market-share and profitability scenarios presented earlier demonstrate AMD’s potential, but they do not establish that its current share price is inexpensive.

A complete valuation would also need to account for AMD’s existing earnings, capital requirements, dilution, future margins and the possibility that Nvidia grows faster than anticipated.

I believe AMD offers considerable growth potential, but the price paid for that growth is crucial.

10. The Risks to My Investment Thesis

There are several risks that investors should consider.

Nvidia’s competitive moat may prove more durable than expected. Its software ecosystem, networking capabilities and integrated computing systems create formidable barriers to entry.

AMD may struggle to convert commercial agreements into profitable revenue. Large infrastructure commitments are complex, and deployment schedules can change.

AI infrastructure spending may slow. If hyperscalers fail to generate sufficient returns on their capital expenditure, accelerator demand could weaken.

Competition could intensify beyond Nvidia and AMD. Google, Amazon and other major technology companies are developing proprietary AI chips, potentially reducing the addressable market available to independent suppliers.

Pricing competition could reduce profitability. AMD may gain market share without generating the margins necessary to justify its valuation.

AMD’s valuation already incorporates substantial growth expectations. Even strong operational performance may not produce attractive shareholder returns if investors reassess the company’s future earnings potential.

These risks are particularly important given the semiconductor industry’s historical cyclicality and the extraordinary expectations surrounding AI.

11. Conclusion: Investing in the Challenger

I believe the artificial intelligence infrastructure market is large enough to support multiple successful semiconductor companies.

Nvidia has established an extraordinary competitive position and continues to demonstrate remarkable financial performance. Its leadership is unlikely to disappear simply because alternative suppliers become more competitive.

However, dominance does not necessarily translate into superior future investment returns.

AMD has the opportunity to benefit from supplier diversification, increasing inference demand, improving software compatibility and continued growth in server CPUs.

Its partnerships with OpenAI and Anthropic demonstrate growing commercial interest in its accelerator technology, while its existing CPU business provides an additional growth engine.

My central argument is that AMD does not need to replace Nvidia to become a substantially larger and more profitable company.

Even a relatively modest share of a growing AI accelerator market could generate significant additional revenue.

The most attractive scenario would be one in which AMD captures incremental market share, improves its software ecosystem and converts growing revenue into strong earnings and free cash flow.

Of course, this outcome is not guaranteed. AMD’s elevated valuation makes successful execution particularly important.

Nevertheless, I believe the company’s opportunity deserves serious consideration.

Nvidia may remain the dominant supplier of AI infrastructure for years to come.

But if AMD can successfully establish itself as a meaningful second supplier, the resulting expansion of its business could be considerable.

For these reasons, I believe AMD has the potential to outperform Nvidia over the next five years—not because Nvidia is destined to fail, but because AMD has substantial room to grow.