US AI Semiconductor

US AI Semiconductor is a thematic investing concept covering the U.S.-listed companies that supply the compute behind the AI build-out: NVIDIA’s GPUs, Broadcom’s custom accelerators, the memory of Micron and Sandisk, and the foundry capacity of Intel and TSMC. Hyperscalers plan around $700 billion of 2026 capital spending, and most of it lands on these chipmakers.

CategoryThematic / Semiconductors — AI compute (US)
Representative companies10
Related ETFKODEX US AI Semiconductor TOP3 Plus ETF (0151S0)
Last updated2026-06-10
Key takeaways
  • US AI Semiconductor groups the U.S.-listed companies that supply the compute behind the AI build-out: GPUs, custom accelerators, memory, networking silicon, and the foundries that fabricate them.
  • The demand side is an AI capex supercycle. For 2026, Alphabet guided capital spending to $180-190 billion, Meta to $125-145 billion, Amazon to about $200 billion, and Microsoft's March-2026 quarter alone ran $31.9 billion, roughly $700 billion of combined hyperscaler investment.
  • NVIDIA remains the anchor: Q1 FY2027 (the February-April 2026 quarter) revenue was a record $81.6 billion, up 85%, with data-center revenue of $75.2 billion, up 92%.
  • Custom silicon and memory are the fastest-growing layers. Broadcom's AI semiconductor revenue grew 143% to $10.8 billion in fiscal Q2 2026, and Micron's fiscal Q2 2026 revenue nearly tripled year over year to $23.86 billion on AI memory demand.
  • Since January 13, 2026 the concept is investable in Korea through the KODEX US AI Semiconductor TOP3 Plus ETF (0151S0), which concentrates over 60% in NVIDIA, TSMC, and Broadcom and tracks the Akros index behind this concept.

What is the US AI semiconductor concept?

It is the supply side of the AI boom, expressed as ten U.S.-listed stocks. Every AI model is trained and served on a stack of silicon: an accelerator such as a GPU or a custom ASIC, high-bandwidth memory stacked beside it, networking chips that lash thousands of accelerators into one cluster, and a leading-edge foundry that fabricates all of it. The concept holds the leaders of each layer. NVIDIA supplies the GPUs and earned a record $75.2 billion of data-center revenue in its February-April 2026 quarter, up 92% year over year (NVIDIA, May 20, 2026). Broadcom and Marvell build the custom accelerators and networking, AMD is the second GPU source, Micron and Sandisk supply the memory, Arm licenses the CPU architecture, Cerebras sells wafer-scale inference systems, and Intel and TSMC’s ADR provide the fabs.

Why now? What is the AI capex supercycle?

Because the four biggest cloud builders have turned AI from a research budget into the largest capital program in corporate history. For 2026, Alphabet guided capital expenditures to $180-190 billion and said 2027 will rise significantly (Q1 2026 earnings call), Meta raised its range to $125-145 billion (Meta, Apr 29, 2026), Amazon said it expects to invest about $200 billion (CNBC, Apr 29, 2026), and Microsoft spent $31.9 billion in its March 2026 quarter alone, with AI revenue running at a $37 billion annualized pace (CNBC, Apr 29, 2026).

The spending is showing up in chipmaker income statements with unusual speed. NVIDIA’s total revenue reached $81.6 billion in a single quarter, up 85%, and it guided the next quarter to about $91.0 billion (NVIDIA, May 20, 2026). NVIDIA founder and CEO Jensen Huang framed the demand bluntly:

“The buildout of AI factories, the largest infrastructure expansion in human history, is accelerating at extraordinary speed.”

— Jensen Huang, founder and CEO, NVIDIA (NVIDIA, May 20, 2026)

Memory is the loudest echo. Micron’s fiscal Q2 2026 revenue nearly tripled year over year to $23.86 billion, and the company guided the following quarter to $33.5 billion with a gross margin around 81% (Micron 8-K, Mar 18, 2026). Sandisk, the NAND maker spun off from Western Digital in February 2025, grew revenue 97% sequentially to $5.95 billion and signed $42 billion of minimum multiyear supply agreements (Sandisk 8-K, Apr 30, 2026).

Why is custom silicon rising alongside the GPU?

Because hyperscalers want a second lever on cost and supply, and that race feeds different companies in the same basket. Broadcom, which co-designs Google’s TPU and Meta’s MTIA accelerators, grew AI semiconductor revenue 143% year over year to $10.8 billion in fiscal Q2 2026 and told investors to expect $16.0 billion in fiscal Q3, more than tripling year over year (Broadcom 8-K, Jun 3, 2026). Amazon said its in-house chips business topped a $20 billion annualized revenue run rate, growing triple digits (Amazon, Apr 29, 2026).

The accelerator market is also widening beyond the incumbents. OpenAI committed to deploy 6 gigawatts of AMD Instinct GPUs across multiple generations, starting with a 1-gigawatt MI450 deployment in the second half of 2026 (AMD, Oct 6, 2025), and AMD’s data-center revenue grew 57% to $5.8 billion in Q1 2026 (AMD 8-K, May 5, 2026). Cerebras, which went public on Nasdaq on May 14, 2026 in a $5.55 billion IPO, sells a wafer-scale processor 58 times larger than NVIDIA’s B200 and signed a deal worth more than $20 billion with OpenAI for 750 megawatts of inference compute (Cerebras S-1; CNBC, May 14, 2026).

Why do Intel and the US government matter to this concept?

Because the United States has made domestic leading-edge manufacturing a policy goal, and Intel is the instrument. In August 2025 the US government invested $8.9 billion in Intel common stock, 433.3 million shares at $20.47, a 9.9% stake that made it Intel’s largest shareholder (Intel 8-K, Aug 22, 2025). A month later NVIDIA invested $5 billion in Intel and the two agreed to co-develop x86 CPUs for NVIDIA’s AI platforms (NVIDIA, Sep 18, 2025). The operating turn is visible: Intel’s Q1 2026 revenue of $13.6 billion beat its own guidance for a sixth straight quarter, with data-center revenue up 22% to $5.1 billion and foundry revenue up 16% to $5.4 billion (Intel 8-K, Apr 23, 2026). Policy cuts the other way too: export rules toward China have shifted repeatedly since the April 2025 H20 license requirement, including a 15% revenue-share arrangement announced in August 2025 (Council on Foreign Relations).

Who should consider US AI semiconductor exposure?

It suits an investor who wants the AI build-out at the layer where the money is actually being spent, and who can size for single-theme concentration. The revenue is current, not promised: NVIDIA booked $81.6 billion in one quarter (NVIDIA), Broadcom guided AI revenue to $16.0 billion for the quarter ending in August 2026 (Broadcom 8-K), and Micron guided a single quarter to $33.5 billion (Micron 8-K).

The caveat is that the whole basket is one macro trade. Four companies, Alphabet, Meta, Amazon, and Microsoft, fund the bulk of demand with roughly $700 billion of planned 2026 capex, so a deceleration in any of those budgets would hit every layer at once (Alphabet Q1 2026 call). For an individual investor that argues for a satellite position beside a diversified core; for an institution it works as a liquid, high-beta AI infrastructure sleeve whose factor overlap with growth and momentum books has to be budgeted explicitly.

Which companies represent US AI Semiconductor?

CompanySectorWhat it does
NVIDIA (NVDA) Information Technology · Semiconductors The anchor of AI compute. Its data-center GPUs and networking generated a record $75.2 billion in Q1 FY2027 (the February-April 2026 quarter), up 92% year over year, which was 92% of company revenue of $81.6 billion. CUDA software and the Blackwell platform are the de facto standard the rest of the industry builds around.
Broadcom (AVGO) Information Technology · Semiconductors The custom-silicon counterweight to the GPU. Broadcom co-designs hyperscaler ASIC accelerators (Google's TPU, Meta's MTIA) and supplies the Ethernet switch silicon of AI fabrics. AI semiconductor revenue grew 143% to $10.8 billion in fiscal Q2 2026, with $16.0 billion guided for fiscal Q3.
Intel (INTC) Information Technology · Semiconductors The only US-headquartered leading-edge logic foundry and the x86 CPU incumbent. Q1 2026 revenue rose 7% to $13.6 billion with foundry revenue up 16% to $5.4 billion, and the turnaround is state-backed: the US government bought a 9.9% stake for $8.9 billion in August 2025 and NVIDIA invested $5 billion a month later.
AMD (AMD) Information Technology · Semiconductors The credible second source for AI GPUs and server CPUs. Data-center revenue grew 57% to $5.8 billion in Q1 2026, and OpenAI has committed to deploy 6 gigawatts of AMD Instinct GPUs starting with the MI450 in late 2026, with Meta planning up to 6 gigawatts more.
Micron Technology (MU) Information Technology · Semiconductors (Memory) The only US-based maker of HBM, the high-bandwidth memory stacked next to every AI accelerator. Fiscal Q2 2026 revenue nearly tripled year over year to $23.86 billion with a 74.4% gross margin, and Micron guided fiscal Q3 revenue to $33.5 billion as AI memory demand outruns supply.
Taiwan Semiconductor Manufacturing (ADR) (TSM) Information Technology · Semiconductor Foundry The NYSE-listed ADR of the foundry that fabricates nearly every leading AI accelerator, from NVIDIA's Blackwell to Google's TPU. Q1 2026 revenue rose 40.6% in US dollars to $35.9 billion with a 66.2% gross margin, on leading-edge nodes and CoWoS advanced packaging.
Cerebras Systems (CBRS) Information Technology · Semiconductors (AI Systems) The wafer-scale challenger. Its Wafer-Scale Engine is 58 times larger than NVIDIA's B200 chip and is sold as a fast-inference system. Revenue grew 76% to $510 million in 2025, OpenAI signed a deal worth more than $20 billion for 750 megawatts of its compute, and its May 2026 Nasdaq IPO raised $5.55 billion.
Sandisk (SNDK) Information Technology · Semiconductors (NAND Flash) The NAND flash maker spun off from Western Digital in February 2025. AI data centers are rewriting its economics: fiscal Q3 2026 revenue of $5.95 billion nearly doubled quarter over quarter, data-center revenue more than tripled sequentially to $1.47 billion, and it has signed $42 billion of minimum multiyear supply agreements.
Arm Holdings (ADR) (ARM) Information Technology · Semiconductors (IP) The CPU-architecture licensor whose designs sit inside NVIDIA's Grace CPUs and the custom server chips of AWS, Google, and Microsoft. Fiscal Q4 2026 revenue rose 20% to a record $1.49 billion, and data-center royalties more than doubled year over year as hyperscalers standardize on Arm-based hosts.
Marvell Technology (MRVL) Information Technology · Semiconductors The data-infrastructure specialist: custom AI accelerators for hyperscalers plus the 800G and 1.6T optical DSPs that move data inside AI clusters. Data center was about 76% of its record $2.418 billion fiscal Q1 2027 revenue, which grew 28% year over year.

What are the risks of US AI Semiconductor?

The cash flows are real, but the concentration is too.

  • Capex dependence. A handful of hyperscalers fund most AI chip demand; Meta already flagged higher component costs while raising its capex range to $125-145 billion (Meta, Apr 29, 2026). If AI returns disappoint, those budgets are the first thing cut.
  • Policy and export controls. China rules have flipped repeatedly: an H20 license requirement in April 2025, a 15% revenue-share deal in August 2025, and an H200 approval debate in late 2025 (CFR). Washington’s 9.9% Intel stake also injects political risk into a commercial turnaround (Intel 8-K).
  • Memory cyclicality. Micron’s 74.4% gross margin and Sandisk’s 78.4% are supercycle margins; memory has historically swung from shortage to glut faster than logic (Micron 8-K; Sandisk 8-K).
  • Single-name risk in new issues. Cerebras derived 86% of its 2025 revenue from two UAE-linked customers, and its IPO valuation embeds the OpenAI ramp executing on schedule (Cerebras S-1).

Related concepts, securities & terms

Related indices & ETFs

These references describe index-tracking relationships as a matter of fact and are not a recommendation to buy any product. Akros, as the index provider, may receive licensing fees from product sponsors. Review the product's prospectus before investing.

Frequently asked questions

What is the US AI semiconductor investment concept?

It is the set of U.S.-listed companies that supply the compute behind the AI build-out: NVIDIA's GPUs, Broadcom's and Marvell's custom accelerators and networking silicon, AMD's GPUs and CPUs, Micron's and Sandisk's memory, Arm's CPU architecture, Cerebras's wafer-scale systems, and the foundry capacity of Intel and TSMC's US-listed ADR. The demand side is hyperscaler capital spending of roughly $700 billion planned for 2026 across Alphabet, Meta, Amazon, and Microsoft (Alphabet Q1 2026 call; Meta, Apr 29, 2026; CNBC, Apr 29, 2026).

Which companies represent the US AI semiconductor concept?

Ten US-listed names: NVIDIA (NVDA), Broadcom (AVGO), Intel (INTC), AMD (AMD), Micron (MU), TSMC's ADR (TSM), Cerebras Systems (CBRS), Sandisk (SNDK), Arm Holdings (ARM), and Marvell Technology (MRVL). They span the full stack: GPUs, custom ASICs, memory, CPU architecture, networking silicon, and leading-edge foundry. NVIDIA alone reported $75.2 billion of data-center revenue in Q1 FY2027, up 92% (NVIDIA, May 20, 2026).

How big is hyperscaler AI capital spending in 2026?

Around $700 billion across the big four. Alphabet guided 2026 capital expenditures to $180-190 billion (Q1 2026 earnings call), Meta raised its range to $125-145 billion (Meta, Apr 29, 2026), Amazon said it expects to invest about $200 billion (CNBC, Apr 29, 2026), and Microsoft spent $31.9 billion in its March 2026 quarter alone (CNBC, Apr 29, 2026). Most of that capex buys chips, memory, and the systems around them.

Is custom silicon a threat to NVIDIA's dominance?

It is the fastest-growing alternative, and this concept holds both sides of the race. Broadcom, which co-designs Google's TPU and Meta's MTIA accelerators, grew AI semiconductor revenue 143% to $10.8 billion in fiscal Q2 2026 and guided fiscal Q3 to $16.0 billion (Broadcom 8-K, Jun 3, 2026). Amazon said its in-house chips business topped a $20 billion revenue run rate, growing triple digits (Amazon, Apr 29, 2026). NVIDIA still grew data-center revenue 92% over the same period, so to date custom silicon has added to demand rather than replaced it.

Why are Intel and Micron in an AI semiconductor basket?

Because AI compute is more than the accelerator. Every accelerator ships with stacked high-bandwidth memory, and Micron, the only US-based HBM maker, saw fiscal Q2 2026 revenue nearly triple to $23.86 billion with fiscal Q3 guided to $33.5 billion (Micron 8-K, Mar 18, 2026). Intel is the only US-headquartered leading-edge foundry; its foundry revenue rose 16% to $5.4 billion in Q1 2026, the US government holds a 9.9% stake bought for $8.9 billion, and NVIDIA invested $5 billion in it (Intel 8-K, Apr 23, 2026; Intel 8-K, Aug 22, 2025).

What are the main risks of the US AI semiconductor theme?

Concentration and cyclicality. A handful of hyperscalers fund most of the demand, so a pause in their capex would hit the whole basket at once. Export policy toward China has shifted repeatedly, from the April 2025 H20 license requirement to a 15% revenue-share arrangement in August 2025 and an H200 approval debate in late 2025 (Council on Foreign Relations). Memory is historically the most cyclical part of semiconductors, valuations embed years of growth, and newly public names like Cerebras carry extreme customer concentration: two UAE-linked customers were 86% of its 2025 revenue (Cerebras S-1).

Is US AI semiconductor a good long-term investment for an individual investor?

It suits an investor who wants direct exposure to the AI build-out and can tolerate single-theme volatility, sized as a satellite rather than a core holding because the basket is one trade on hyperscaler capex. The cash flows behind it are real and dated: NVIDIA earned $81.6 billion of revenue in a single quarter (NVIDIA, May 20, 2026) and the big four cloud builders plan roughly $700 billion of 2026 capex (Alphabet Q1 2026 call). But the same concentration cuts both ways: if AI infrastructure spending slows, every layer of the basket, chips, memory, and foundry, would feel it together.

How does US AI semiconductor fit an institutional thematic mandate?

As a high-beta AI infrastructure sleeve in liquid US-listed large caps, with one new-issue exception in Cerebras. The names are easy to model and trade, and the thesis maps to disclosed customer commitments: OpenAI's 6-gigawatt agreement with AMD (AMD, Oct 6, 2025), Broadcom's $16.0 billion fiscal Q3 AI revenue guide (Broadcom 8-K), and Micron's $33.5 billion quarterly guide (Micron 8-K). The allocator's caveats are factor overlap with any growth or momentum book and a demand base concentrated in four capex budgets.

Is there an ETF that tracks the US AI semiconductor concept?

Yes. The KODEX US AI Semiconductor TOP3 Plus ETF (0151S0) from Samsung Asset Management listed on the Korea Exchange on January 13, 2026 and tracks the Akros U.S. AI Semiconductor Top 3 Plus Index behind this concept (Samsung Asset Management). The fund concentrates over 60% in NVIDIA, TSMC, and Broadcom, also holds memory names such as Micron and Sandisk, and charges a total fee of 0.49% a year (Samsung Asset Management newsroom). It passed 100 billion won of net assets within eight trading days of listing (Edaily, Jan 26, 2026). Always check fees, holdings, and risk before investing.

Sources & references

  1. NVIDIA Announces Financial Results for First Quarter Fiscal 2027 · NVIDIA Corporation (Newsroom), 2026-05-20
  2. Broadcom announces second quarter fiscal year 2026 financial results (SEC 8-K, Exhibit 99.1) · Broadcom Inc. / SEC EDGAR, 2026-06-03
  3. Intel Reports First-Quarter 2026 Financial Results (SEC 8-K, Exhibit 99.1) · Intel Corporation / SEC EDGAR, 2026-04-23
  4. Intel Announces Agreement with the Trump Administration: $8.9 Billion Investment in Intel Common Stock (SEC 8-K, Exhibit 99.1) · Intel Corporation / SEC EDGAR, 2025-08-22
  5. AMD Reports First Quarter 2026 Financial Results (SEC 8-K, Exhibit 99.1) · Advanced Micro Devices, Inc. / SEC EDGAR, 2026-05-05
  6. AMD and OpenAI Announce Strategic Partnership to Deploy 6 Gigawatts of AMD GPUs · Advanced Micro Devices, Inc., 2025-10-06
  7. Micron Technology, Inc. Reports Results for the Second Quarter of Fiscal 2026 (SEC 8-K, Exhibit 99.1) · Micron Technology, Inc. / SEC EDGAR, 2026-03-18
  8. TSMC Reports First Quarter EPS of NT$22.08 (Form 6-K, Exhibit 99.1) · Taiwan Semiconductor Manufacturing Company / SEC EDGAR, 2026-04-16
  9. Alphabet Announces First Quarter 2026 Results (SEC 8-K, Exhibit 99.1) · Alphabet Inc. / SEC EDGAR, 2026-04-29
  10. Alphabet (GOOGL) Q1 2026 earnings call transcript · The Motley Fool, 2026-04-29
  11. Meta Reports First Quarter 2026 Results (SEC 8-K, Exhibit 99.1) · Meta Platforms, Inc. / SEC EDGAR, 2026-04-29
  12. Amazon.com Announces First Quarter Results (SEC 8-K, Exhibit 99.1) · Amazon.com, Inc. / SEC EDGAR, 2026-04-29
  13. Amazon (AMZN) Q1 earnings report 2026 · CNBC, 2026-04-29
  14. Microsoft (MSFT) Q3 earnings report 2026 · CNBC, 2026-04-29
  15. Cerebras Systems Inc. Form S-1 Registration Statement · Cerebras Systems Inc. / SEC EDGAR, 2026-04-17
  16. Cerebras Systems Announces Pricing of Initial Public Offering · Cerebras Systems Inc., 2026-05-13
  17. Cerebras (CBRS) starts trading on Nasdaq after IPO · CNBC, 2026-05-14
  18. Sandisk Reports Fiscal Third Quarter 2026 Financial Results (SEC 8-K, Exhibit 99.1) · Sandisk Corporation / SEC EDGAR, 2026-04-30
  19. Arm Holdings plc Reports Results for the Fourth Quarter and Fiscal Year Ended 2026 · Arm Holdings plc, 2026-05-06
  20. Marvell Technology reports fiscal Q1 2027 results (Form 8-K, Exhibit 99.1) · Marvell Technology / SEC EDGAR, 2026-05-27
  21. NVIDIA and Intel to Develop AI Infrastructure and Personal Computing Products · NVIDIA Corporation (Newsroom), 2025-09-18
  22. The Consequences of Exporting Nvidia's H200 Chips to China · Council on Foreign Relations, 2025-12-15
  23. KODEX 미국AI반도체TOP3플러스 ETF (0151S0) — product page · Samsung Asset Management, 2026-01-13
  24. KODEX 미국AI반도체TOP3플러스 ETF (0151S0) — newsroom listing article (1/13 상장) · Samsung Asset Management, 2026-01-06
  25. KODEX 미국AI반도체TOP3플러스, 순자산 1000억 돌파 · Edaily, 2026-01-26