As of Friday's U.S. stock market close, NVIDIA (NVDA.US) shares fell nearly 3%, marking the first time in nearly two months that the world's highest market cap company faces significant risk of losing its milestone $4 trillion valuation. The "AI chip leader" NVIDIA saw intraday declines of nearly 5% on Friday, with weekly indicators showing the stock has declined for four consecutive weeks. For a stock with over $4 trillion market cap, a near 5% drop represents a massive decline, highlighting market panic selling pressure triggered by disappointing U.S. non-farm payroll data that sparked recession warnings, combined with Broadcom's surging AI ASIC market scale. Additionally, Taiwan Semiconductor's (TSM) U.S. ADR strong gains represent the second-best performance among recent U.S. chip stocks, trailing only Broadcom.

In the chip industry chain, Taiwan Semiconductor stands as the "eternal champion," with both high-demand AI GPUs and AI ASICs relying on TSMC. Through decades of accumulated chip manufacturing technology and long-term positioning at the global forefront of chip manufacturing technology improvement and innovation, TSMC dominates the vast majority of global chip foundry orders with leading advanced process and packaging technologies, plus ultra-high yields, especially for 5nm and below most advanced process chip foundry orders.

From a retrospective perspective, while extremely weak non-farm data has cooled expectations for a U.S. economic "soft landing," market expectations for Federal Reserve rate cuts have heated up significantly, with September potentially seeing a 50 basis point cut, which would be a positive factor for mega-cap growth stock NVIDIA on the denominator side. Siebert Financial Chief Investment Officer Mark Malek stated that a 50 basis point move "would provide tailwinds for the stock market." "This would undoubtedly boost mega-cap growth stocks and give investors the green light to take on more risk."

Therefore, Broadcom's exceptionally strong performance and future outlook represent the core logic behind market panic. NVIDIA's strongest competitor in the AI chip field - Broadcom's collaboration with tech giants like Google, Meta, and Microsoft in launching AI ASICs with such rapid revenue scale growth has panicked Wall Street's top investment institutions and retail investors who have long been bullish on NVIDIA's stock price and performance expansion trends. As AI ASIC revenue scale strengthens and may soon begin eroding AI GPU market share, they are beginning to adjust their long-term performance growth expectations for NVIDIA.

AI ASICs and NVIDIA AI GPUs represent two completely different technological approaches in AI chips. Currently, the two are largely market competitors, especially as AI ASICs offer significant cost-effectiveness and energy efficiency advantages in AI training/inference for hyperscale cloud computing giants and AI leaders like OpenAI.

OpenAI's $10 billion "super order" to Broadcom and Broadcom's performance highlighting explosive AI ASIC demand have caused Wall Street analysts' 2026-2030 performance expectations for NVIDIA to crack. Broadcom's strong rise has forced them to moderately downgrade their previously extremely high performance growth expectations, which is also the core logic behind NVIDIA's Friday stock plunge.

NVIDIA's "$4 Trillion Club" status under pressure - significant stock retreat in recent weeks

According to sources cited by media, Broadcom is helping OpenAI design and manufacture a customized artificial intelligence accelerator chip (AI ASIC chip). The companies plan to deliver the first batch of AI chips in this series starting next year at the earliest, which will put NVIDIA, the absolute leader in AI computing infrastructure, under the most direct and strongest long-term competitive pressure from Broadcom in the AI chip field.

Broadcom's third-quarter AI infrastructure-related semiconductor revenue was approximately $5.2 billion, up 63% year-over-year, exceeding the Wall Street average expectation of $5.11 billion. Broadcom management expects this category revenue to reach approximately $6.2 billion in the fourth quarter, implying potential year-over-year growth of nearly 70%, above analysts' previous expectations of approximately $5.82 billion.

Before earnings release, market expectations for Broadcom's performance and future outlook data were very high, so exceeding market expectations significantly boosted investor bullish sentiment toward Broadcom and the entire AI computing industry chain.

Broadcom CEO Hock Tan stated in a conference call following Broadcom's earnings announcement that after securing over $10 billion in AI infrastructure orders from an unnamed new hyperscale customer (later revealed by media to be OpenAI), the company expects fiscal 2026 AI-related revenue growth to be stronger than previously anticipated. In the last earnings call, Tan had indicated that 2026 AI-related revenue prospects would show growth trajectory similar to this year - expecting growth of approximately 50% to 60%.

In the Chinese stock market, Cambricon, which similarly bets on the ASIC route, represents the hottest stock in China's current market, with year-to-date gains of 95%. Wall Street giant Goldman Sachs raised its target price for the "China AI chip leader" and "domestic chip substitution" leader Cambricon for the second time in just one week. In its latest September 1 report, Goldman Sachs raised Cambricon's 12-month target price from RMB 1,835 to RMB 2,104, an increase of 14.7%, while maintaining a "buy" rating. The latest target price implies 41% upside potential for the stock, which has repeatedly hit new highs this year, compared to its August 29 closing price.

Cambricon has recently seen both stock price and performance resonance, highlighting the intense heat of China's AI investment boom and "domestic chip substitution" trend. Performance-wise, Cambricon achieved operating revenue of RMB 2.881 billion in the first half of 2025, a staggering 4,347.82% year-over-year increase, with the company achieving net profit attributable to shareholders of RMB 1.038 billion, compared to a loss of RMB 530 million in the same period last year.

AI computing industry chain bull market logic remains unassailable, but capital begins comprehensive focus on ASICs

Undoubtedly, Broadcom's strong performance and future outlook have reinforced the "long-term bull market narrative" for the AI computing sector, but global capital net long bets are beginning to shift massively from NVIDIA's AI GPU chain to AI ASIC chains.

Globally explosive AI computing demand growth, combined with increasingly massive U.S. government-led AI infrastructure investment projects and tech giants' continued massive investments in building large data centers, largely means that for investors who have long favored NVIDIA and the AI computing industry chain, the global "AI faith's" "super catalyst" effect on computing leaders' stock prices is far from over. They bet that stock prices of AI computing industry chain companies led by NVIDIA, TSMC, and Broadcom will continue演绎"bull market curves," driving global stock markets to continue bull market performance.

It is precisely under the leadership of epic stock price gains and continuously strong performance this year from AI computing industry chain leaders like NVIDIA, Google, TSMC, and Broadcom that an unprecedented AI investment boom has swept U.S. stock markets and global equity markets, driving the global benchmark index - MSCI World Index to surge significantly since April, recently hitting new historical highs.

NVIDIA, with over $4 trillion market cap, saw intraday declines of nearly 5% on Friday. In comparison, Broadcom's stock surged as much as 16% intraday Friday, hitting an intraday record high, adding nearly $150 billion to its market value at peak, reaching approximately $1.6 trillion, highlighting global capital flocking to Broadcom with rapidly heating bullish sentiment toward this AI ASIC leader.

Clearstead Advisors Senior Managing Director Jim Awad stated that while investors should prepare for NVIDIA facing stronger competition in this field, due to the extremely fast growth of the AI infrastructure market, the company can maintain continued growth even if it loses some market share, though future growth rates may continue to lag behind competitors like Broadcom.

With this continued retreat, NVIDIA has fallen approximately 10% from its August highs, with market cap evaporating nearly $470 billion. The company recently broke below its 50-day moving average, though it remains the world's highest market cap company. Microsoft ranks second with $3.7 trillion market cap.

Mizuho Securities Managing Director of Equity Trading Daniel O'Regan noted that relative to Broadcom, NVIDIA has touched 18-month relative lows, which is quite noteworthy. "The word OpenAI really generates a lot of momentum, and this accelerates a trend: capital is increasingly favoring Broadcom," he stated, noting Broadcom's significantly outperforming NVIDIA this year. "Today's divergence seems somewhat extreme, but NVIDIA has been synonymous with AI for about three years. While I don't think the market is cooling on it permanently, capital is indeed expanding to other AI winners," O'Regan emphasized. "In contrast, Broadcom is a 'shiny new thing.'"

On Wall Street, given Broadcom's ethernet switch chips and AI ASIC chip demand continuing explosive growth, Wall Street financial giants generally remain bullish on Broadcom's stock price prospects, optimistic about Broadcom continuing to hit new stock price highs. Therefore, after Broadcom announced earnings, they generally significantly raised target prices for the company's next 12 months, with Susquehanna, Bernstein, KeyBanc, and Barclays raising Broadcom target prices significantly to $400.

As of Friday's U.S. market close, Broadcom closed up nearly 10% at $334.89.

AI ASIC wave increasingly surging

With absolute technological leadership in inter-chip interconnect communication and high-speed data transmission between chips, Broadcom has become the most important participant in AI infrastructure AI ASIC customized chips in recent years. For example, Google's data center server AI chips - TPU AI accelerator chips - Broadcom is a core participant, with Broadcom and Google teams jointly participating in TPU AI accelerator chip development.

Beyond chip design, Broadcom also provides Google with critical inter-chip interconnect communication intellectual property and handles manufacturing, testing, and packaging of new chips, thereby supporting Google's expansion of new AI data centers.

As U.S. tech giants firmly invest massive resources in artificial intelligence, the biggest beneficiaries include not only NVIDIA but also AI ASIC giants like Broadcom, Marvell Technology, and Taiwan's Global Unichip. Microsoft, Amazon, Google, Meta, and even generative AI leader OpenAI are all partnering with Broadcom or other ASIC giants to update and iterate AI ASIC chips, primarily for massive inference-side AI computing deployment.

Therefore, AI ASIC future market share expansion is expected to significantly outpace AI GPUs, trending toward parity rather than the current NVIDIA AI GPU dominance holding up to 90% of AI chip market share.

According to reports, one of Broadcom's major customers, Google, disclosed latest details of Ironwood TPU (TPU v6) at a conference, showing remarkable performance improvements. Compared to TPU v5p, Ironwood's peak FLOPS performance improved 10-fold, with efficiency ratio improving 5.6-fold. Compared to Google's 2022-launched TPU v4, Ironwood's single-chip computing power improvement exceeds 16-fold.

Performance comparison shows: Google Ironwood's 4.2 TFLOPS/watt efficiency ratio is only slightly lower than NVIDIA B200/300 GPU's 4.5 TFLOPS/watt.

JPMorgan commented: This performance data prominently demonstrates that advanced AI's specialized AI ASIC chips are rapidly narrowing the performance gap with market-leading AI GPUs, driving hyperscale cloud service providers to increase investment in more cost-effective customized ASIC projects.

According to Wall Street financial giant JPMorgan's latest forecast, this chip adopts 3nm advanced process technology in cooperation with Broadcom, will enter mass production in the second half of 2025, with Ironwood expected to generate approximately $10 billion in revenue for Broadcom over the next 6-7 months.

AI ASIC cost-effectiveness advantages represent the biggest advantage over NVIDIA AI GPUs for giants like Google, Microsoft, and OpenAI. Meta's data center internal AI ASIC - MTIA v2, under third-party GEMM/inference testing models, consumes less energy than NVIDIA H100 while maintaining similar throughput (both approximately 5.5-5.7× relative to T4). Estimating H100 ≥$30k and MTIA v2 expected $2-3k maximum cost, MTIA's cost-performance ratio (perf/$) far exceeds NVIDIA.

Therefore, while AI ASICs cannot comprehensively replace NVIDIA on a large scale, market share will inevitably expand rather than maintaining the current NVIDIA AI GPU dominance. Especially in actual AI data center computing infrastructure configuration, AI ASIC and NVIDIA AI GPU "hybrid formations" (GPU for training/exploration, ASIC for large-scale inference/partial training) significantly improve efficiency ratios and substantially reduce TCO.

For standardizable mainstream inference and partial training (especially continuous long-tail training/fine-tuning), customized AI ASIC's "per-throughput cost/energy consumption" significantly outperforms pure GPU solutions. For rapid exploration, frontier large model training, and multimodal new operator experimentation, NVIDIA AI GPUs remain the mainstay.

Therefore, in current AI engineering practice, tech giants increasingly tend toward "ASIC handling normalization, GPU handling exploration peaks/new model development" hybrid architecture to minimize TCO.

TSMC demonstrates what "classics never go out of style" means! Both AI GPUs with explosive demand since 2023 and recently surging AI ASICs depend on TSMC. TSMC currently serves as the world's largest contract chip foundry. As the AI deployment frenzy shows no signs of cooling and continues sweeping globally, its customers NVIDIA, AMD, Broadcom and other chip giants continue benefiting from market trends toward explosive growth in AI chips - AI's most core infrastructure.

These chip giants' foundry contract scales with TSMC have surged, driving TSMC's performance to continue exceeding expectations with strong expansion since last year, which also represents important logical support for TSMC's Taiwan stock and U.S. ADR stock prices repeatedly hitting new highs this year.

TSMC, with decades of accumulated chip manufacturing technology and long-term positioning at the global forefront of chip manufacturing technology improvement and innovation (pioneering the FinFET era and driving the 2nm GAA era), dominates the vast majority of global chip foundry orders with leading advanced process and packaging technologies plus ultra-high yields, especially 5nm and below advanced process chip foundry orders.

More importantly, TSMC currently captures almost all 5nm and below process high-end chip packaging orders with its industry-leading 2.5D and 3D chiplet advanced packaging, with advanced packaging capacity far unable to meet demand. NVIDIA Blackwell has been in short supply since achieving mass production late last year, fully constrained by TSMC's 2.5D-level CoWoS packaging capacity.

Currently, chip giants like Apple, AMD, NVIDIA, and Broadcom are turning to TSMC's 3D-level advanced packaging capacity, potentially further driving TSMC's advanced packaging capacity shortage.

TSMC's latest earnings show explosive AI computing demand growth drove Q2 net profit up 61%. TSMC expects 2025 dollar-denominated sales to grow around 30%, above previous "close to mid-20%" growth expectations, mainly benefiting from continued explosive growth in AI chip orders based on 3nm and 5nm advanced process technologies.

Due to still extremely strong AI computing demand, TSMC is actively expanding backend capacity to improve CoWoS advanced packaging actual output, mainly for NVIDIA AI GPU capacity, indicating the company's confidence that extremely strong AI chip demand will continue through 2026.