The Semiconductor Split

How U.S.–China Competition Is Reshaping the AI Hardware Stack

Part 4 of the Growth Solutions KC AI Infrastructure Series

Chips are no longer just chips.

For decades, semiconductors were treated as commercial technology — the components inside computers, phones, servers, vehicles, appliances, and industrial systems. That world is gone.

In the AI era, advanced semiconductors are strategic infrastructure. They shape economic growth, military capability, industrial power, energy demand, trade leverage, and national competitiveness.

The country that controls advanced compute controls part of the future. That is why the AI hardware story cannot be separated from the U.S.–China semiconductor conflict.

The first three articles in this series examined the physical stack behind AI: infrastructure, capex, and high‑bandwidth memory. But those layers now sit inside a larger geopolitical reality.

The global semiconductor ecosystem is splitting.

The United States and its allies are trying to preserve control over the most advanced chips, manufacturing tools, software platforms, packaging capabilities, and supply‑chain chokepoints. China is trying to reduce dependence on those chokepoints by building a domestic semiconductor ecosystem of its own.

This is the semiconductor split — and it may be one of the most important strategic developments in the AI infrastructure race.

The Split: Integration Is Fracturing

The semiconductor industry was built through global specialization.

• U.S. companies led in chip design, software tools, and core intellectual property.
• Taiwan became the world’s leading advanced foundry hub through TSMC.
• The Netherlands became indispensable through ASML’s lithography tools.
• Japan remained critical in materials and equipment.
• South Korea became essential in memory through Samsung and SK Hynix.
• China became the world’s largest electronics manufacturing base and a major end market for chips.

This system created efficiency — and dependence.

For years, China relied on foreign access to advanced chips, manufacturing equipment, design tools, foundry capacity, and high‑performance computing ecosystems. The United States and its allies held many of the chokepoints.

AI changed the stakes.

Advanced chips are no longer just commercial products. They are inputs into frontier AI models, autonomous systems, cyber capabilities, surveillance infrastructure, military planning, industrial automation, and national power.

That is why export controls became central to U.S. policy.

Beginning in October 2022, the U.S. Bureau of Industry and Security imposed advanced computing and semiconductor manufacturing controls aimed at restricting China’s access to advanced chips, supercomputing capabilities, and the tools needed to manufacture leading‑edge semiconductors. The stated national‑security goal was to limit China’s access to technologies that could support military modernization and advanced AI capabilities.

On paper, the strategy was straightforward:

Deny China the frontier.
Preserve America’s lead.
Protect national security.

But strategy often changes once it meets reality.

The Policy Tradeoff

This is where Part 4 picks up from The Chip Gambit and The China Chip Conundrum.

Those articles explored the central dilemma: U.S. restrictions may have slowed China temporarily, but they also accelerated China’s urgency to build around American control points. The Chip Gambit framed the risk directly — export controls may have compressed years of Chinese domestic chip development into a much shorter period by forcing national urgency.

The China Chip Conundrum sharpened the point: selling chips to China can strengthen China’s AI capabilities, but restricting those chips can accelerate China’s semiconductor independence. There is no easy answer — only tradeoffs.

The United States was not wrong to recognize that advanced AI chips are dual‑use technologies. They support commercial innovation, but they also support military and surveillance systems.

The problem is not that export controls exist.

The problem is whether the controls are targeted, coherent, durable, and strategically balanced — or whether they sacrifice commercial leverage while teaching China exactly which dependencies it must eliminate.

A policy can create short‑term friction and still produce long‑term acceleration in a competitor’s domestic ecosystem.

China’s Response: Build the Stack at Home

China’s response has been clear: build as much of the stack domestically as possible.

• Chip design.
• Manufacturing.
• Packaging Materials.
• Equipment.
• Memory.
• AI accelerators.
• Optical components.
• Software frameworks.
• Data‑center infrastructure.

China is not trying to replace one product. It is trying to reduce vulnerability across the entire semiconductor and AI infrastructure stack. That does not mean China has caught the United States and its allies at the frontier. It has not.

But the direction matters.

Reuters recently reported that Samsung and SK Hynix are testing Chinese chipmaking equipment as a hedge against tighter U.S. export‑control risk. The same reporting noted expectations that several Chinese equipment suppliers could each generate more than $1 billion in revenue in 2026, and that Chinese suppliers’ share could approach 40% in parts of the equipment market excluding lithography and metrology.

That is the signal.

Even if China remains behind in the hardest frontier layers, domestic alternatives are becoming more credible in adjacent and supporting parts of the stack.

That is how ecosystems mature:

1. Substitute where possible.
2. Improve under constraint.
3. Integrate.
4. Create captive demand.
5. Become harder to dislodge.

SMHC as a Market Signal — Not a Recommendation

The launch of VanEck’s China Semiconductor ETF, SMHC, is useful — not as a recommendation, but as a market signal.

VanEck frames the thesis around China building a full‑stack domestic semiconductor industry across design, fabrication, materials, and equipment. It excludes sanctioned companies by design, reflecting regulatory risk.

This matters because VanEck is not an obscure issuer. Its SMH fund is one of the most widely followed semiconductor ETFs in the world.

So SMHC should be read less as “buy this fund” and more as:

China’s domestic semiconductor buildout has become large and distinct enough to be packaged as its own investable category.

That is significant.

The ETF’s holdings — Cambricon, Hygon, Eoptolink, AMEC, Naura, Piotech, Montage Technology, JCET, and others — are not recommendations. They are evidence that China’s semiconductor ecosystem is broadening beyond a single company or headline chip.

The market is recognizing what geopolitics already made clear: China is no longer trying to buy into the global semiconductor stack. It is actively building its own.

NVIDIA and the China Problem

NVIDIA remains the clearest example of the semiconductor split. Globally, NVIDIA is still the dominant AI‑accelerator company. Its chips, networking, CUDA ecosystem, and full‑stack software advantage remain central to the AI infrastructure buildout.

But China is no longer a normal growth market.

NVIDIA’s Q1 fiscal 2027 materials stated that the company was not assuming any data‑center compute revenue from China in its outlook.

That sentence carries enormous strategic weight.

It means the world’s most important AI chip company is planning massive growth while largely excluding one of the world’s largest AI infrastructure markets from its assumptions.

Chinese demand did not disappear — it is being redirected.

Some moves to domestic Chinese alternatives.
Some moves through gray‑market pathways.
Some shifts to different architectures.
Some becomes policy leverage.
Some becomes pressure on China’s own semiconductor industry to improve.

The danger for U.S. companies: The market may still exist, but access to it may no longer belong to them.

Where China Still Faces Bottlenecks

China’s progress is real, but it should not be exaggerated.

The country still faces major constraints:

• Leading‑edge manufacturing
• Advanced lithography
• High‑bandwidth memory
• Advanced packaging
• EDA software ecosystems
• Yield, reliability, and scale
• Developer‑ecosystem maturity

The United States and its allies still hold major advantages across frontier chips, design software, manufacturing equipment, advanced memory, packaging, and ecosystem depth.

But the question is not whether China has already caught up. The better question is whether China is becoming good enough in enough layers to reduce U.S. leverage.

In technology competition, the gap does not have to disappear for the strategic balance to shift. Sometimes it is enough for a competitor to become less dependent, less vulnerable, and more capable of sustaining its own ecosystem.

What It Means for Investors

The semiconductor split changes how AI hardware exposure should be evaluated.

1. Global demand is not the same as accessible demand.
2. Geopolitical risk is not an abstract footnote.
3. Domestic substitution matters.
4. China exposure must be analyzed carefully.
5. ETFs and indexes are signals of market structure.

The lesson is discipline.

Do not assume global AI demand flows equally to every supplier.
Do not assume policy restrictions only hurt the target.
Do not assume China’s semiconductor ecosystem can be ignored.
Do not assume short‑term friction equals long‑term strategic success.

The AI hardware thesis is strong. But geopolitics determines who gets to participate.

What Policymakers Should Understand

For policymakers, the question is not simply:

“How do we slow China?”

The better question is:

“How does America stay ahead?”

Trying to slow a determined competitor can be part of a strategy, but it cannot be the whole strategy. Against a country with China’s scale, talent base, industrial ambition, and state‑backed capital, restrictions often create adaptation.

A stronger American strategy must combine:

• Targeted protection of the true frontier
• Domestic manufacturing strength
• Allied coordination
• Clear, durable rules
• Energy and infrastructure capacity
• Immigration and talent policy
• R&D investment
• Commercial leverage where national security permits it
• A disciplined distinction between military‑grade restrictions and broad commercial disengagement

The goal is not self‑harm disguised as toughness.

The goal is durable American advantage.

What to Watch Next

• NVIDIA’s China language — a key indicator of separation.
• Huawei, Cambricon, Hygon, SMIC, and China’s equipment suppliers — the question is not frontier parity, but “good enough” capability.
• Advanced packaging and HBM — critical for scaling AI hardware.
• U.S. policy consistency — whiplash creates openings for China.
• Allied coordination — the U.S. cannot control the stack alone.
• Market products like SMHC — signals of how capital markets categorize the split.

Compute Is Power

The semiconductor split is not just a trade dispute. It is a contest over the infrastructure of the AI age.

The United States still has enormous advantages: leading chip designers, software ecosystems, allied manufacturing networks, frontier AI companies, deep capital markets, and unmatched innovation capacity.

China has scale, urgency, state‑directed capital, manufacturing depth, and a national commitment to reducing foreign dependence.

Both realities can be true.

The AI hardware buildout is not happening in a neutral world. It is unfolding inside a geopolitical competition where chips, memory, packaging, power, software, and data centers are instruments of national strength.

Part 1 showed that AI is physical.
Part 2 showed that capex is the demand signal.
Part 3 showed that memory is a hidden choke point.
Part 4 shows that the hardware stack itself is splitting.

The future of AI will not be shaped only by who builds the best model. It will be shaped by who controls the compute.

— Matt Cucinotta  |  Growth Solutions KC  |  Inspire · Inform · Ignite

Source note: This article is based on public company materials, U.S. government export-control documents, fund issuer materials, news reporting, prior Growth Solutions KC analysis, and Growth Solutions KC interpretation. References to securities, ETFs, companies, or sectors are illustrative and are not recommendations to buy or sell any investment.

This article is for informational and educational purposes only and does not constitute personalized investment advice. Securities, companies, funds, and sectors discussed should be evaluated in light of each reader’s own financial situation, risk tolerance, time horizon, and professional guidance.